diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,9 +1,9 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.30490251143384417,
+  "epoch": 0.4894487683543288,
   "eval_steps": 500,
-  "global_step": 7600,
+  "global_step": 12200,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -45607,13 +45607,27613 @@
       "learning_rate": 0.00023742596944938637,
       "loss": 0.7099,
       "step": 7600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023741054449087753,
+      "loss": 0.6735,
+      "step": 7601
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023739511813262,
+      "loss": 0.6141,
+      "step": 7602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002373796903748608,
+      "loss": 0.4585,
+      "step": 7603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000237364261217847,
+      "loss": 0.2934,
+      "step": 7604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023734883066182567,
+      "loss": 0.6636,
+      "step": 7605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002373333987070439,
+      "loss": 1.6048,
+      "step": 7606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023731796535374883,
+      "loss": 0.1613,
+      "step": 7607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023730253060218757,
+      "loss": 0.1716,
+      "step": 7608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023728709445260734,
+      "loss": 0.7061,
+      "step": 7609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002372716569052553,
+      "loss": 0.4302,
+      "step": 7610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002372562179603786,
+      "loss": 0.4265,
+      "step": 7611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002372407776182245,
+      "loss": 0.8683,
+      "step": 7612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023722533587904038,
+      "loss": 0.2322,
+      "step": 7613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023720989274307337,
+      "loss": 0.5941,
+      "step": 7614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002371944482105708,
+      "loss": 0.0499,
+      "step": 7615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023717900228177995,
+      "loss": 1.1117,
+      "step": 7616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002371635549569483,
+      "loss": 0.3098,
+      "step": 7617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002371481062363231,
+      "loss": 0.4054,
+      "step": 7618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023713265612015173,
+      "loss": 1.5695,
+      "step": 7619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023711720460868168,
+      "loss": 0.0198,
+      "step": 7620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023710175170216038,
+      "loss": 0.0683,
+      "step": 7621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002370862974008352,
+      "loss": 0.4361,
+      "step": 7622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023707084170495365,
+      "loss": 0.5569,
+      "step": 7623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023705538461476328,
+      "loss": 2.3641,
+      "step": 7624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002370399261305116,
+      "loss": 0.0051,
+      "step": 7625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023702446625244606,
+      "loss": 0.5533,
+      "step": 7626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023700900498081434,
+      "loss": 0.5173,
+      "step": 7627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000236993542315864,
+      "loss": 0.6658,
+      "step": 7628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023697807825784266,
+      "loss": 0.6016,
+      "step": 7629
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023696261280699787,
+      "loss": 0.4612,
+      "step": 7630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023694714596357735,
+      "loss": 1.8501,
+      "step": 7631
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023693167772782878,
+      "loss": 0.2923,
+      "step": 7632
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023691620809999993,
+      "loss": 0.6055,
+      "step": 7633
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023691620809999993,
+      "loss": 1.0909,
+      "step": 7634
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023690073708033837,
+      "loss": 0.3101,
+      "step": 7635
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023688526466909195,
+      "loss": 1.785,
+      "step": 7636
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023686979086650842,
+      "loss": 0.5883,
+      "step": 7637
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023685431567283557,
+      "loss": 0.1128,
+      "step": 7638
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002368388390883212,
+      "loss": 0.7366,
+      "step": 7639
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023682336111321314,
+      "loss": 1.0468,
+      "step": 7640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023680788174775926,
+      "loss": 0.2956,
+      "step": 7641
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023679240099220742,
+      "loss": 0.3053,
+      "step": 7642
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023677691884680557,
+      "loss": 0.9651,
+      "step": 7643
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023676143531180157,
+      "loss": 0.4367,
+      "step": 7644
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002367459503874434,
+      "loss": 1.0155,
+      "step": 7645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000236730464073979,
+      "loss": 0.2126,
+      "step": 7646
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023671497637165644,
+      "loss": 1.2277,
+      "step": 7647
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023669948728072363,
+      "loss": 0.456,
+      "step": 7648
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023668399680142865,
+      "loss": 0.4807,
+      "step": 7649
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023666850493401955,
+      "loss": 0.0134,
+      "step": 7650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023665301167874442,
+      "loss": 0.2415,
+      "step": 7651
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023663751703585138,
+      "loss": 0.2048,
+      "step": 7652
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023662202100558852,
+      "loss": 0.3441,
+      "step": 7653
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000236606523588204,
+      "loss": 0.2812,
+      "step": 7654
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023659102478394594,
+      "loss": 0.2055,
+      "step": 7655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023657552459306264,
+      "loss": 0.4024,
+      "step": 7656
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023656002301580218,
+      "loss": 0.3531,
+      "step": 7657
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023654452005241285,
+      "loss": 0.3804,
+      "step": 7658
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023652901570314302,
+      "loss": 0.0663,
+      "step": 7659
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002365135099682408,
+      "loss": 0.1609,
+      "step": 7660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023649800284795454,
+      "loss": 0.897,
+      "step": 7661
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002364824943425326,
+      "loss": 0.4787,
+      "step": 7662
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002364669844522233,
+      "loss": 0.2247,
+      "step": 7663
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023645147317727502,
+      "loss": 0.0204,
+      "step": 7664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023643596051793614,
+      "loss": 0.0708,
+      "step": 7665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023642044647445505,
+      "loss": 0.0082,
+      "step": 7666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023640493104708026,
+      "loss": 0.1325,
+      "step": 7667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023638941423606013,
+      "loss": 0.5926,
+      "step": 7668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002363738960416432,
+      "loss": 0.0382,
+      "step": 7669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023635837646407797,
+      "loss": 4.9678,
+      "step": 7670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023634285550361292,
+      "loss": 0.0136,
+      "step": 7671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023632733316049663,
+      "loss": 0.2609,
+      "step": 7672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023631180943497766,
+      "loss": 0.2398,
+      "step": 7673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002362962843273046,
+      "loss": 0.6406,
+      "step": 7674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023628075783772604,
+      "loss": 0.3611,
+      "step": 7675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002362652299664907,
+      "loss": 0.5287,
+      "step": 7676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002362497007138471,
+      "loss": 0.4862,
+      "step": 7677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023623417008004398,
+      "loss": 0.3209,
+      "step": 7678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002362186380653301,
+      "loss": 0.3434,
+      "step": 7679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023620310466995406,
+      "loss": 0.7303,
+      "step": 7680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023618756989416472,
+      "loss": 0.2127,
+      "step": 7681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023617203373821078,
+      "loss": 0.2098,
+      "step": 7682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023615649620234108,
+      "loss": 0.3163,
+      "step": 7683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023614095728680435,
+      "loss": 0.3524,
+      "step": 7684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002361254169918495,
+      "loss": 0.4196,
+      "step": 7685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002361098753177253,
+      "loss": 0.5007,
+      "step": 7686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023609433226468076,
+      "loss": 0.2489,
+      "step": 7687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002360787878329646,
+      "loss": 0.7034,
+      "step": 7688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023606324202282597,
+      "loss": 1.2387,
+      "step": 7689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023604769483451357,
+      "loss": 0.5309,
+      "step": 7690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023603214626827652,
+      "loss": 0.7718,
+      "step": 7691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023601659632436375,
+      "loss": 0.3333,
+      "step": 7692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023600104500302432,
+      "loss": 0.65,
+      "step": 7693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023598549230450716,
+      "loss": 0.4506,
+      "step": 7694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002359699382290614,
+      "loss": 0.9338,
+      "step": 7695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023595438277693615,
+      "loss": 0.4285,
+      "step": 7696
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023593882594838045,
+      "loss": 0.0068,
+      "step": 7697
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023592326774364342,
+      "loss": 0.3989,
+      "step": 7698
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002359077081629742,
+      "loss": 0.598,
+      "step": 7699
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023589214720662194,
+      "loss": 0.6105,
+      "step": 7700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002358765848748359,
+      "loss": 0.2842,
+      "step": 7701
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023586102116786517,
+      "loss": 1.6985,
+      "step": 7702
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023584545608595908,
+      "loss": 0.3299,
+      "step": 7703
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002358298896293669,
+      "loss": 0.4204,
+      "step": 7704
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002358143217983378,
+      "loss": 0.8234,
+      "step": 7705
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023579875259312113,
+      "loss": 0.4153,
+      "step": 7706
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023578318201396617,
+      "loss": 0.3945,
+      "step": 7707
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023576761006112231,
+      "loss": 1.1338,
+      "step": 7708
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023575203673483895,
+      "loss": 0.2372,
+      "step": 7709
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023573646203536532,
+      "loss": 1.065,
+      "step": 7710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000235720885962951,
+      "loss": 0.4438,
+      "step": 7711
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023570530851784534,
+      "loss": 0.6774,
+      "step": 7712
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023568972970029777,
+      "loss": 0.2739,
+      "step": 7713
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023567414951055778,
+      "loss": 0.6877,
+      "step": 7714
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023565856794887486,
+      "loss": 0.5252,
+      "step": 7715
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023564298501549852,
+      "loss": 0.7533,
+      "step": 7716
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023562740071067832,
+      "loss": 0.5141,
+      "step": 7717
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023561181503466377,
+      "loss": 0.3394,
+      "step": 7718
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023559622798770456,
+      "loss": 0.5429,
+      "step": 7719
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002355806395700502,
+      "loss": 0.8061,
+      "step": 7720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023556504978195032,
+      "loss": 0.496,
+      "step": 7721
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023554945862365456,
+      "loss": 0.3299,
+      "step": 7722
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023553386609541263,
+      "loss": 0.2558,
+      "step": 7723
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023551827219747424,
+      "loss": 0.7946,
+      "step": 7724
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000235502676930089,
+      "loss": 0.3981,
+      "step": 7725
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023548708029350678,
+      "loss": 0.0031,
+      "step": 7726
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002354714822879772,
+      "loss": 0.2986,
+      "step": 7727
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023545588291375017,
+      "loss": 0.7299,
+      "step": 7728
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023544028217107536,
+      "loss": 0.4322,
+      "step": 7729
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023542468006020268,
+      "loss": 0.7465,
+      "step": 7730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000235409076581382,
+      "loss": 0.3179,
+      "step": 7731
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002353934717348631,
+      "loss": 0.3525,
+      "step": 7732
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023537786552089591,
+      "loss": 0.837,
+      "step": 7733
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023536225793973037,
+      "loss": 0.9162,
+      "step": 7734
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023534664899161633,
+      "loss": 0.6503,
+      "step": 7735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002353310386768038,
+      "loss": 0.4194,
+      "step": 7736
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023531542699554276,
+      "loss": 0.8075,
+      "step": 7737
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023529981394808323,
+      "loss": 0.2568,
+      "step": 7738
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023528419953467515,
+      "loss": 0.4549,
+      "step": 7739
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023526858375556866,
+      "loss": 0.1999,
+      "step": 7740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002352529666110137,
+      "loss": 0.0053,
+      "step": 7741
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023523734810126045,
+      "loss": 0.0081,
+      "step": 7742
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023522172822655905,
+      "loss": 0.9512,
+      "step": 7743
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023520610698715949,
+      "loss": 0.4819,
+      "step": 7744
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023519048438331206,
+      "loss": 0.0367,
+      "step": 7745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023517486041526686,
+      "loss": 0.3327,
+      "step": 7746
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002351592350832741,
+      "loss": 0.8725,
+      "step": 7747
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023514360838758399,
+      "loss": 0.2306,
+      "step": 7748
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002351279803284468,
+      "loss": 0.441,
+      "step": 7749
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002351123509061127,
+      "loss": 0.6864,
+      "step": 7750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023509672012083215,
+      "loss": 0.375,
+      "step": 7751
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002350810879728552,
+      "loss": 1.8397,
+      "step": 7752
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023506545446243238,
+      "loss": 0.3072,
+      "step": 7753
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023504981958981402,
+      "loss": 0.3751,
+      "step": 7754
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023503418335525043,
+      "loss": 0.2204,
+      "step": 7755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023501854575899193,
+      "loss": 0.9357,
+      "step": 7756
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023500290680128908,
+      "loss": 0.524,
+      "step": 7757
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002349872664823922,
+      "loss": 0.2864,
+      "step": 7758
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023497162480255184,
+      "loss": 0.7406,
+      "step": 7759
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023495598176201843,
+      "loss": 0.5459,
+      "step": 7760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023494033736104247,
+      "loss": 2.3047,
+      "step": 7761
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023492469159987447,
+      "loss": 0.7821,
+      "step": 7762
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023490904447876501,
+      "loss": 0.3717,
+      "step": 7763
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023489339599796463,
+      "loss": 0.0032,
+      "step": 7764
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023487774615772389,
+      "loss": 0.3187,
+      "step": 7765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023486209495829347,
+      "loss": 0.2391,
+      "step": 7766
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023484644239992394,
+      "loss": 0.384,
+      "step": 7767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023483078848286597,
+      "loss": 0.0272,
+      "step": 7768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023481513320737024,
+      "loss": 0.3317,
+      "step": 7769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023479947657368747,
+      "loss": 0.4565,
+      "step": 7770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023478381858206828,
+      "loss": 0.3565,
+      "step": 7771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023476815923276353,
+      "loss": 0.3743,
+      "step": 7772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023475249852602392,
+      "loss": 0.2033,
+      "step": 7773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023473683646210022,
+      "loss": 1.0538,
+      "step": 7774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023472117304124324,
+      "loss": 0.2401,
+      "step": 7775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023470550826370387,
+      "loss": 0.7846,
+      "step": 7776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023468984212973292,
+      "loss": 0.7935,
+      "step": 7777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002346741746395812,
+      "loss": 0.0077,
+      "step": 7778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002346585057934997,
+      "loss": 0.6067,
+      "step": 7779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023464283559173927,
+      "loss": 0.5806,
+      "step": 7780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023462716403455084,
+      "loss": 0.0229,
+      "step": 7781
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023461149112218537,
+      "loss": 0.477,
+      "step": 7782
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023459581685489388,
+      "loss": 0.6163,
+      "step": 7783
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023458014123292734,
+      "loss": 0.2025,
+      "step": 7784
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023456446425653685,
+      "loss": 0.9318,
+      "step": 7785
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023454878592597326,
+      "loss": 0.0023,
+      "step": 7786
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023453310624148781,
+      "loss": 0.6564,
+      "step": 7787
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023451742520333153,
+      "loss": 1.4454,
+      "step": 7788
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023450174281175552,
+      "loss": 0.2335,
+      "step": 7789
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002344860590670109,
+      "loss": 0.8064,
+      "step": 7790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002344703739693488,
+      "loss": 0.5763,
+      "step": 7791
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023445468751902054,
+      "loss": 0.0031,
+      "step": 7792
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023443899971627716,
+      "loss": 0.3259,
+      "step": 7793
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002344233105613699,
+      "loss": 0.555,
+      "step": 7794
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023440762005455008,
+      "loss": 0.3423,
+      "step": 7795
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002343919281960689,
+      "loss": 0.5474,
+      "step": 7796
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023437623498617758,
+      "loss": 0.2077,
+      "step": 7797
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002343605404251275,
+      "loss": 0.1996,
+      "step": 7798
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023434484451317,
+      "loss": 2.3302,
+      "step": 7799
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002343291472505564,
+      "loss": 0.5741,
+      "step": 7800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023431344863753803,
+      "loss": 0.5579,
+      "step": 7801
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023429774867436638,
+      "loss": 0.5672,
+      "step": 7802
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002342820473612928,
+      "loss": 0.4267,
+      "step": 7803
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023426634469856866,
+      "loss": 0.3916,
+      "step": 7804
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023425064068644547,
+      "loss": 0.6572,
+      "step": 7805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023423493532517475,
+      "loss": 0.4014,
+      "step": 7806
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023421922861500795,
+      "loss": 0.2069,
+      "step": 7807
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002342035205561966,
+      "loss": 0.3528,
+      "step": 7808
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023418781114899225,
+      "loss": 0.3909,
+      "step": 7809
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023417210039364643,
+      "loss": 0.2846,
+      "step": 7810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023415638829041077,
+      "loss": 0.3549,
+      "step": 7811
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023414067483953682,
+      "loss": 0.0064,
+      "step": 7812
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023412496004127626,
+      "loss": 0.3028,
+      "step": 7813
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023410924389588076,
+      "loss": 0.0031,
+      "step": 7814
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002340935264036019,
+      "loss": 0.4889,
+      "step": 7815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023407780756469143,
+      "loss": 0.7892,
+      "step": 7816
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023406208737940105,
+      "loss": 0.7342,
+      "step": 7817
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023404636584798254,
+      "loss": 0.0017,
+      "step": 7818
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002340306429706876,
+      "loss": 0.4654,
+      "step": 7819
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023401491874776802,
+      "loss": 0.7408,
+      "step": 7820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023399919317947557,
+      "loss": 0.4494,
+      "step": 7821
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023398346626606214,
+      "loss": 0.3647,
+      "step": 7822
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023396773800777953,
+      "loss": 0.41,
+      "step": 7823
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002339520084048796,
+      "loss": 0.238,
+      "step": 7824
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023393627745761425,
+      "loss": 2.6368,
+      "step": 7825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002339205451662354,
+      "loss": 0.7405,
+      "step": 7826
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023390481153099497,
+      "loss": 0.8017,
+      "step": 7827
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023388907655214492,
+      "loss": 0.5932,
+      "step": 7828
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002338733402299372,
+      "loss": 1.7862,
+      "step": 7829
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023385760256462382,
+      "loss": 0.5356,
+      "step": 7830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023384186355645674,
+      "loss": 0.0054,
+      "step": 7831
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002338261232056881,
+      "loss": 0.4376,
+      "step": 7832
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023381038151256987,
+      "loss": 0.0043,
+      "step": 7833
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002337946384773542,
+      "loss": 0.4203,
+      "step": 7834
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002337788941002931,
+      "loss": 0.3647,
+      "step": 7835
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023376314838163877,
+      "loss": 1.5787,
+      "step": 7836
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002337474013216433,
+      "loss": 0.8582,
+      "step": 7837
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023373165292055897,
+      "loss": 0.0075,
+      "step": 7838
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002337159031786378,
+      "loss": 0.3231,
+      "step": 7839
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023370015209613212,
+      "loss": 0.6742,
+      "step": 7840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023368439967329412,
+      "loss": 0.0137,
+      "step": 7841
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000233668645910376,
+      "loss": 0.7334,
+      "step": 7842
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023365289080763013,
+      "loss": 0.2147,
+      "step": 7843
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023363713436530876,
+      "loss": 2.7736,
+      "step": 7844
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023362137658366423,
+      "loss": 0.2986,
+      "step": 7845
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002336056174629488,
+      "loss": 0.4038,
+      "step": 7846
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002335898570034149,
+      "loss": 0.6761,
+      "step": 7847
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002335740952053149,
+      "loss": 0.317,
+      "step": 7848
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002335583320689012,
+      "loss": 0.7167,
+      "step": 7849
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023354256759442618,
+      "loss": 0.3028,
+      "step": 7850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023352680178214235,
+      "loss": 0.6488,
+      "step": 7851
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023351103463230215,
+      "loss": 0.0349,
+      "step": 7852
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023349526614515806,
+      "loss": 0.5471,
+      "step": 7853
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023347949632096258,
+      "loss": 0.3552,
+      "step": 7854
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002334637251599682,
+      "loss": 0.4068,
+      "step": 7855
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002334479526624276,
+      "loss": 0.6091,
+      "step": 7856
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023343217882859323,
+      "loss": 0.4604,
+      "step": 7857
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023341640365871777,
+      "loss": 0.4952,
+      "step": 7858
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023340062715305378,
+      "loss": 0.5021,
+      "step": 7859
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023338484931185392,
+      "loss": 0.1859,
+      "step": 7860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023336907013537083,
+      "loss": 0.2335,
+      "step": 7861
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023335328962385718,
+      "loss": 0.0788,
+      "step": 7862
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023333750777756573,
+      "loss": 0.0329,
+      "step": 7863
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023332172459674918,
+      "loss": 0.0201,
+      "step": 7864
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023330594008166023,
+      "loss": 0.3086,
+      "step": 7865
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023330594008166023,
+      "loss": 0.7881,
+      "step": 7866
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023329015423255165,
+      "loss": 0.9456,
+      "step": 7867
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023327436704967627,
+      "loss": 0.0063,
+      "step": 7868
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023325857853328686,
+      "loss": 0.5269,
+      "step": 7869
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002332427886836363,
+      "loss": 0.3371,
+      "step": 7870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023322699750097739,
+      "loss": 3.0603,
+      "step": 7871
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023321120498556307,
+      "loss": 0.2778,
+      "step": 7872
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023319541113764615,
+      "loss": 0.8711,
+      "step": 7873
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023317961595747957,
+      "loss": 0.2694,
+      "step": 7874
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002331638194453163,
+      "loss": 0.3206,
+      "step": 7875
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023314802160140922,
+      "loss": 0.8477,
+      "step": 7876
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023313222242601138,
+      "loss": 0.2279,
+      "step": 7877
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023311642191937583,
+      "loss": 0.1992,
+      "step": 7878
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023310062008175547,
+      "loss": 0.5106,
+      "step": 7879
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023308481691340337,
+      "loss": 0.3873,
+      "step": 7880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023306901241457265,
+      "loss": 0.8626,
+      "step": 7881
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023305320658551637,
+      "loss": 0.4442,
+      "step": 7882
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023303739942648764,
+      "loss": 0.5291,
+      "step": 7883
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023302159093773958,
+      "loss": 0.6908,
+      "step": 7884
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023300578111952529,
+      "loss": 1.0708,
+      "step": 7885
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000232989969972098,
+      "loss": 1.8012,
+      "step": 7886
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023297415749571094,
+      "loss": 0.037,
+      "step": 7887
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023295834369061727,
+      "loss": 0.32,
+      "step": 7888
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002329425285570702,
+      "loss": 0.0787,
+      "step": 7889
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023292671209532303,
+      "loss": 0.0176,
+      "step": 7890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000232910894305629,
+      "loss": 0.3443,
+      "step": 7891
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023289507518824146,
+      "loss": 0.163,
+      "step": 7892
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023287925474341366,
+      "loss": 0.4588,
+      "step": 7893
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000232863432971399,
+      "loss": 0.2093,
+      "step": 7894
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023284760987245083,
+      "loss": 0.3915,
+      "step": 7895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002328317854468225,
+      "loss": 0.0116,
+      "step": 7896
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023281595969476746,
+      "loss": 0.9757,
+      "step": 7897
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023280013261653914,
+      "loss": 0.2825,
+      "step": 7898
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023278430421239097,
+      "loss": 2.2528,
+      "step": 7899
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023276847448257634,
+      "loss": 0.0307,
+      "step": 7900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023275264342734883,
+      "loss": 0.6015,
+      "step": 7901
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023273681104696195,
+      "loss": 0.8403,
+      "step": 7902
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023272097734166922,
+      "loss": 0.2333,
+      "step": 7903
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023270514231172416,
+      "loss": 0.3299,
+      "step": 7904
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002326893059573804,
+      "loss": 0.9014,
+      "step": 7905
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002326734682788915,
+      "loss": 1.134,
+      "step": 7906
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023265762927651105,
+      "loss": 0.3536,
+      "step": 7907
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002326417889504927,
+      "loss": 0.6289,
+      "step": 7908
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023262594730109014,
+      "loss": 0.6983,
+      "step": 7909
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002326101043285571,
+      "loss": 0.4723,
+      "step": 7910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023259426003314717,
+      "loss": 0.5532,
+      "step": 7911
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023257841441511412,
+      "loss": 0.006,
+      "step": 7912
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002325625674747117,
+      "loss": 0.6182,
+      "step": 7913
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002325467192121937,
+      "loss": 1.6756,
+      "step": 7914
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023253086962781385,
+      "loss": 0.2043,
+      "step": 7915
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023251501872182602,
+      "loss": 0.2337,
+      "step": 7916
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023249916649448394,
+      "loss": 0.4206,
+      "step": 7917
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023248331294604155,
+      "loss": 0.0133,
+      "step": 7918
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002324674580767527,
+      "loss": 0.0019,
+      "step": 7919
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023245160188687126,
+      "loss": 0.9314,
+      "step": 7920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023243574437665119,
+      "loss": 0.7386,
+      "step": 7921
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023241988554634641,
+      "loss": 0.2608,
+      "step": 7922
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002324040253962108,
+      "loss": 0.4291,
+      "step": 7923
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002323881639264984,
+      "loss": 0.0057,
+      "step": 7924
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023237230113746322,
+      "loss": 1.2833,
+      "step": 7925
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023235643702935923,
+      "loss": 0.1263,
+      "step": 7926
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023234057160244055,
+      "loss": 0.0053,
+      "step": 7927
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002323247048569612,
+      "loss": 0.0469,
+      "step": 7928
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023230883679317522,
+      "loss": 0.3204,
+      "step": 7929
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023229296741133674,
+      "loss": 0.5613,
+      "step": 7930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002322770967116999,
+      "loss": 0.0039,
+      "step": 7931
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023226122469451884,
+      "loss": 0.9054,
+      "step": 7932
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023224535136004773,
+      "loss": 1.5791,
+      "step": 7933
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023222947670854075,
+      "loss": 0.4615,
+      "step": 7934
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023221360074025208,
+      "loss": 1.7045,
+      "step": 7935
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023219772345543603,
+      "loss": 0.9077,
+      "step": 7936
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023218184485434678,
+      "loss": 1.1407,
+      "step": 7937
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002321659649372386,
+      "loss": 0.0076,
+      "step": 7938
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023215008370436582,
+      "loss": 0.63,
+      "step": 7939
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023213420115598276,
+      "loss": 0.0075,
+      "step": 7940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002321183172923437,
+      "loss": 0.5192,
+      "step": 7941
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023210243211370304,
+      "loss": 2.6271,
+      "step": 7942
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023208654562031518,
+      "loss": 0.0043,
+      "step": 7943
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023207065781243445,
+      "loss": 0.4213,
+      "step": 7944
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002320547686903153,
+      "loss": 0.0381,
+      "step": 7945
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023203887825421222,
+      "loss": 0.2594,
+      "step": 7946
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002320229865043796,
+      "loss": 1.3711,
+      "step": 7947
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023200709344107194,
+      "loss": 1.2018,
+      "step": 7948
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023199119906454376,
+      "loss": 0.4849,
+      "step": 7949
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002319753033750496,
+      "loss": 0.1663,
+      "step": 7950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000231959406372844,
+      "loss": 3.2451,
+      "step": 7951
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023194350805818146,
+      "loss": 0.3193,
+      "step": 7952
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023192760843131663,
+      "loss": 0.0105,
+      "step": 7953
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002319117074925041,
+      "loss": 0.3962,
+      "step": 7954
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023189580524199854,
+      "loss": 0.0375,
+      "step": 7955
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023187990168005456,
+      "loss": 0.2834,
+      "step": 7956
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002318639968069268,
+      "loss": 0.2779,
+      "step": 7957
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023184809062287001,
+      "loss": 1.668,
+      "step": 7958
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002318321831281389,
+      "loss": 0.179,
+      "step": 7959
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023181627432298818,
+      "loss": 0.6479,
+      "step": 7960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023180036420767264,
+      "loss": 0.392,
+      "step": 7961
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023178445278244702,
+      "loss": 0.0061,
+      "step": 7962
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002317685400475661,
+      "loss": 0.5788,
+      "step": 7963
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023175262600328475,
+      "loss": 0.0128,
+      "step": 7964
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023173671064985783,
+      "loss": 1.6807,
+      "step": 7965
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023172079398754011,
+      "loss": 0.2647,
+      "step": 7966
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023170487601658653,
+      "loss": 0.1967,
+      "step": 7967
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023168895673725197,
+      "loss": 1.0022,
+      "step": 7968
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002316730361497914,
+      "loss": 0.254,
+      "step": 7969
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023165711425445973,
+      "loss": 0.0125,
+      "step": 7970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023164119105151186,
+      "loss": 1.3149,
+      "step": 7971
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023162526654120288,
+      "loss": 0.0155,
+      "step": 7972
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023160934072378775,
+      "loss": 0.0935,
+      "step": 7973
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023159341359952156,
+      "loss": 0.3731,
+      "step": 7974
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023157748516865924,
+      "loss": 0.2497,
+      "step": 7975
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023156155543145596,
+      "loss": 0.466,
+      "step": 7976
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002315456243881668,
+      "loss": 0.3005,
+      "step": 7977
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002315296920390468,
+      "loss": 0.4746,
+      "step": 7978
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023151375838435114,
+      "loss": 0.1716,
+      "step": 7979
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000231497823424335,
+      "loss": 0.3821,
+      "step": 7980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023148188715925352,
+      "loss": 0.0972,
+      "step": 7981
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002314659495893619,
+      "loss": 0.6158,
+      "step": 7982
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023145001071491535,
+      "loss": 0.3769,
+      "step": 7983
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023143407053616914,
+      "loss": 0.5215,
+      "step": 7984
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002314181290533785,
+      "loss": 0.5605,
+      "step": 7985
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023140218626679874,
+      "loss": 0.0029,
+      "step": 7986
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002313862421766851,
+      "loss": 1.249,
+      "step": 7987
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023137029678329296,
+      "loss": 0.4497,
+      "step": 7988
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023135435008687759,
+      "loss": 0.4881,
+      "step": 7989
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023133840208769445,
+      "loss": 0.5678,
+      "step": 7990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023132245278599886,
+      "loss": 0.3002,
+      "step": 7991
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023130650218204622,
+      "loss": 0.4953,
+      "step": 7992
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000231290550276092,
+      "loss": 0.6557,
+      "step": 7993
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023127459706839156,
+      "loss": 0.0048,
+      "step": 7994
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023125864255920046,
+      "loss": 0.0043,
+      "step": 7995
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023124268674877416,
+      "loss": 0.0076,
+      "step": 7996
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002312267296373682,
+      "loss": 0.6705,
+      "step": 7997
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023121077122523797,
+      "loss": 0.6016,
+      "step": 7998
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002311948115126392,
+      "loss": 0.1583,
+      "step": 7999
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023117885049982732,
+      "loss": 1.2713,
+      "step": 8000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023116288818705806,
+      "loss": 0.4147,
+      "step": 8001
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023114692457458687,
+      "loss": 0.8815,
+      "step": 8002
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002311309596626695,
+      "loss": 0.1502,
+      "step": 8003
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002311149934515616,
+      "loss": 0.4656,
+      "step": 8004
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023109902594151878,
+      "loss": 0.4908,
+      "step": 8005
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023108305713279677,
+      "loss": 0.3918,
+      "step": 8006
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002310670870256513,
+      "loss": 1.158,
+      "step": 8007
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002310511156203381,
+      "loss": 0.0757,
+      "step": 8008
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023103514291711292,
+      "loss": 0.4513,
+      "step": 8009
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023101916891623157,
+      "loss": 0.0064,
+      "step": 8010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023100319361794982,
+      "loss": 0.5257,
+      "step": 8011
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023098721702252345,
+      "loss": 0.5287,
+      "step": 8012
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023097123913020837,
+      "loss": 0.4415,
+      "step": 8013
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002309552599412604,
+      "loss": 0.2616,
+      "step": 8014
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023093927945593544,
+      "loss": 0.387,
+      "step": 8015
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023092329767448943,
+      "loss": 0.3309,
+      "step": 8016
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023090731459717823,
+      "loss": 0.1674,
+      "step": 8017
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023089133022425782,
+      "loss": 0.4729,
+      "step": 8018
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023087534455598415,
+      "loss": 0.6591,
+      "step": 8019
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002308593575926132,
+      "loss": 0.4868,
+      "step": 8020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023084336933440095,
+      "loss": 0.0115,
+      "step": 8021
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023082737978160356,
+      "loss": 0.6212,
+      "step": 8022
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023081138893447697,
+      "loss": 0.4365,
+      "step": 8023
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023079539679327725,
+      "loss": 0.7832,
+      "step": 8024
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002307794033582605,
+      "loss": 0.4735,
+      "step": 8025
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002307634086296828,
+      "loss": 0.2249,
+      "step": 8026
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002307474126078004,
+      "loss": 0.8291,
+      "step": 8027
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023073141529286932,
+      "loss": 0.7453,
+      "step": 8028
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002307154166851458,
+      "loss": 0.0047,
+      "step": 8029
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023069941678488602,
+      "loss": 2.0215,
+      "step": 8030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002306834155923462,
+      "loss": 0.0888,
+      "step": 8031
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002306674131077825,
+      "loss": 4.7136,
+      "step": 8032
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023065140933145137,
+      "loss": 0.0337,
+      "step": 8033
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023063540426360892,
+      "loss": 0.6726,
+      "step": 8034
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023061939790451142,
+      "loss": 0.2983,
+      "step": 8035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002306033902544153,
+      "loss": 0.7096,
+      "step": 8036
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023058738131357686,
+      "loss": 0.6242,
+      "step": 8037
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023057137108225244,
+      "loss": 0.6193,
+      "step": 8038
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023055535956069843,
+      "loss": 0.3548,
+      "step": 8039
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023053934674917124,
+      "loss": 0.2577,
+      "step": 8040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023052333264792727,
+      "loss": 0.2002,
+      "step": 8041
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023050731725722304,
+      "loss": 0.6309,
+      "step": 8042
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023049130057731484,
+      "loss": 0.0048,
+      "step": 8043
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023047528260845932,
+      "loss": 0.4908,
+      "step": 8044
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000230459263350913,
+      "loss": 0.2844,
+      "step": 8045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023044324280493224,
+      "loss": 0.1815,
+      "step": 8046
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023042722097077364,
+      "loss": 0.6087,
+      "step": 8047
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002304111978486939,
+      "loss": 0.547,
+      "step": 8048
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023039517343894942,
+      "loss": 0.4853,
+      "step": 8049
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023037914774179694,
+      "loss": 0.5914,
+      "step": 8050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000230363120757493,
+      "loss": 0.8468,
+      "step": 8051
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023034709248629433,
+      "loss": 0.5012,
+      "step": 8052
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023033106292845757,
+      "loss": 0.1789,
+      "step": 8053
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023031503208423943,
+      "loss": 0.8467,
+      "step": 8054
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023029899995389647,
+      "loss": 0.2088,
+      "step": 8055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002302829665376856,
+      "loss": 0.3905,
+      "step": 8056
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023026693183586354,
+      "loss": 0.0022,
+      "step": 8057
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023025089584868697,
+      "loss": 0.3804,
+      "step": 8058
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002302348585764128,
+      "loss": 0.3116,
+      "step": 8059
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023021882001929777,
+      "loss": 0.6545,
+      "step": 8060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002302027801775987,
+      "loss": 0.0046,
+      "step": 8061
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002301867390515725,
+      "loss": 0.3614,
+      "step": 8062
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023017069664147598,
+      "loss": 0.0047,
+      "step": 8063
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023015465294756612,
+      "loss": 2.2529,
+      "step": 8064
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023013860797009979,
+      "loss": 0.2838,
+      "step": 8065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023012256170933386,
+      "loss": 0.7769,
+      "step": 8066
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023012256170933386,
+      "loss": 2.9376,
+      "step": 8067
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023010651416552542,
+      "loss": 0.8598,
+      "step": 8068
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023009046533893138,
+      "loss": 0.0567,
+      "step": 8069
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023007441522980871,
+      "loss": 0.4985,
+      "step": 8070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023005836383841442,
+      "loss": 0.1695,
+      "step": 8071
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023004231116500567,
+      "loss": 3.2652,
+      "step": 8072
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023002625720983935,
+      "loss": 0.4181,
+      "step": 8073
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023001020197317262,
+      "loss": 0.5507,
+      "step": 8074
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022999414545526266,
+      "loss": 0.2148,
+      "step": 8075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022997808765636647,
+      "loss": 0.1808,
+      "step": 8076
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022996202857674127,
+      "loss": 0.4143,
+      "step": 8077
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022994596821664413,
+      "loss": 0.0274,
+      "step": 8078
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002299299065763323,
+      "loss": 0.349,
+      "step": 8079
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000229913843656063,
+      "loss": 0.4298,
+      "step": 8080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022989777945609349,
+      "loss": 0.1367,
+      "step": 8081
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022988171397668081,
+      "loss": 0.8581,
+      "step": 8082
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022986564721808248,
+      "loss": 1.3616,
+      "step": 8083
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002298495791805556,
+      "loss": 0.3377,
+      "step": 8084
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002298335098643576,
+      "loss": 0.0221,
+      "step": 8085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022981743926974574,
+      "loss": 0.4434,
+      "step": 8086
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022980136739697736,
+      "loss": 0.4765,
+      "step": 8087
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022978529424630986,
+      "loss": 0.0082,
+      "step": 8088
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022976921981800064,
+      "loss": 0.8037,
+      "step": 8089
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022975314411230704,
+      "loss": 0.8975,
+      "step": 8090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022973706712948655,
+      "loss": 1.1475,
+      "step": 8091
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002297209888697966,
+      "loss": 1.3627,
+      "step": 8092
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022970490933349463,
+      "loss": 0.2742,
+      "step": 8093
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022968882852083817,
+      "loss": 0.7156,
+      "step": 8094
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022967274643208475,
+      "loss": 0.0889,
+      "step": 8095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022965666306749182,
+      "loss": 0.5415,
+      "step": 8096
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022964057842731698,
+      "loss": 0.3956,
+      "step": 8097
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022962449251181786,
+      "loss": 0.0129,
+      "step": 8098
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022962449251181786,
+      "loss": 0.5752,
+      "step": 8099
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022960840532125195,
+      "loss": 0.624,
+      "step": 8100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022959231685587688,
+      "loss": 0.7648,
+      "step": 8101
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022957622711595037,
+      "loss": 0.0415,
+      "step": 8102
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022956013610172997,
+      "loss": 0.4063,
+      "step": 8103
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002295440438134734,
+      "loss": 0.4597,
+      "step": 8104
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022952795025143837,
+      "loss": 0.1874,
+      "step": 8105
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022951185541588247,
+      "loss": 0.3829,
+      "step": 8106
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022949575930706365,
+      "loss": 0.5823,
+      "step": 8107
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022947966192523954,
+      "loss": 0.2673,
+      "step": 8108
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022946356327066788,
+      "loss": 1.0564,
+      "step": 8109
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022944746334360657,
+      "loss": 0.1057,
+      "step": 8110
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002294313621443133,
+      "loss": 0.2694,
+      "step": 8111
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000229415259673046,
+      "loss": 2.6759,
+      "step": 8112
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002293991559300625,
+      "loss": 0.619,
+      "step": 8113
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002293830509156207,
+      "loss": 0.1164,
+      "step": 8114
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022936694462997843,
+      "loss": 0.8767,
+      "step": 8115
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002293508370733937,
+      "loss": 0.2279,
+      "step": 8116
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002293347282461244,
+      "loss": 0.7005,
+      "step": 8117
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022931861814842847,
+      "loss": 0.4098,
+      "step": 8118
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022930250678056396,
+      "loss": 0.2264,
+      "step": 8119
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002292863941427888,
+      "loss": 0.2973,
+      "step": 8120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022927028023536096,
+      "loss": 1.2573,
+      "step": 8121
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022925416505853864,
+      "loss": 0.5366,
+      "step": 8122
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002292380486125798,
+      "loss": 0.252,
+      "step": 8123
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022922193089774248,
+      "loss": 0.1055,
+      "step": 8124
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022920581191428487,
+      "loss": 0.3118,
+      "step": 8125
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022918969166246502,
+      "loss": 0.0162,
+      "step": 8126
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002291735701425411,
+      "loss": 3.1779,
+      "step": 8127
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022915744735477132,
+      "loss": 0.1358,
+      "step": 8128
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022914132329941383,
+      "loss": 0.6493,
+      "step": 8129
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022912519797672678,
+      "loss": 0.4387,
+      "step": 8130
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022910907138696848,
+      "loss": 0.0158,
+      "step": 8131
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002290929435303971,
+      "loss": 0.308,
+      "step": 8132
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022907681440727088,
+      "loss": 0.431,
+      "step": 8133
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022906068401784823,
+      "loss": 0.9619,
+      "step": 8134
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022904455236238738,
+      "loss": 0.4237,
+      "step": 8135
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002290284194411466,
+      "loss": 1.2997,
+      "step": 8136
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022901228525438433,
+      "loss": 0.9286,
+      "step": 8137
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022899614980235886,
+      "loss": 0.348,
+      "step": 8138
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022898001308532863,
+      "loss": 0.1763,
+      "step": 8139
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000228963875103552,
+      "loss": 0.5453,
+      "step": 8140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022894773585728742,
+      "loss": 0.8753,
+      "step": 8141
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022893159534679338,
+      "loss": 0.3542,
+      "step": 8142
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022891545357232827,
+      "loss": 0.283,
+      "step": 8143
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002288993105341506,
+      "loss": 0.6846,
+      "step": 8144
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022888316623251892,
+      "loss": 0.0219,
+      "step": 8145
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022886702066769168,
+      "loss": 0.6596,
+      "step": 8146
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022885087383992745,
+      "loss": 0.0111,
+      "step": 8147
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002288347257494849,
+      "loss": 0.321,
+      "step": 8148
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022881857639662245,
+      "loss": 0.1339,
+      "step": 8149
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022880242578159882,
+      "loss": 0.0097,
+      "step": 8150
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022878627390467257,
+      "loss": 0.2885,
+      "step": 8151
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002287701207661024,
+      "loss": 0.8259,
+      "step": 8152
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022875396636614699,
+      "loss": 0.3255,
+      "step": 8153
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022873781070506503,
+      "loss": 0.6498,
+      "step": 8154
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022872165378311508,
+      "loss": 0.0983,
+      "step": 8155
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002287054956005561,
+      "loss": 0.4694,
+      "step": 8156
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022868933615764671,
+      "loss": 1.3575,
+      "step": 8157
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022867317545464567,
+      "loss": 0.627,
+      "step": 8158
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022865701349181184,
+      "loss": 0.4974,
+      "step": 8159
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022864085026940394,
+      "loss": 0.1758,
+      "step": 8160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022862468578768085,
+      "loss": 0.4333,
+      "step": 8161
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022860852004690145,
+      "loss": 0.0938,
+      "step": 8162
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022859235304732457,
+      "loss": 0.1669,
+      "step": 8163
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022857618478920908,
+      "loss": 0.1461,
+      "step": 8164
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022856001527281393,
+      "loss": 2.9603,
+      "step": 8165
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022854384449839802,
+      "loss": 0.342,
+      "step": 8166
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022852767246622034,
+      "loss": 0.536,
+      "step": 8167
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022851149917653985,
+      "loss": 2.2305,
+      "step": 8168
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002284953246296155,
+      "loss": 0.241,
+      "step": 8169
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022847914882570634,
+      "loss": 0.245,
+      "step": 8170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022846297176507143,
+      "loss": 0.567,
+      "step": 8171
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022844679344796972,
+      "loss": 0.4776,
+      "step": 8172
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022843061387466037,
+      "loss": 0.0365,
+      "step": 8173
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022841443304540246,
+      "loss": 0.5153,
+      "step": 8174
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022839825096045507,
+      "loss": 0.551,
+      "step": 8175
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022838206762007735,
+      "loss": 0.2012,
+      "step": 8176
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022836588302452848,
+      "loss": 0.5615,
+      "step": 8177
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022834969717406761,
+      "loss": 2.1484,
+      "step": 8178
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002283335100689539,
+      "loss": 0.5051,
+      "step": 8179
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022831732170944662,
+      "loss": 0.4393,
+      "step": 8180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022830113209580492,
+      "loss": 0.0632,
+      "step": 8181
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022828494122828808,
+      "loss": 1.7911,
+      "step": 8182
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022826874910715552,
+      "loss": 0.218,
+      "step": 8183
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022825255573266632,
+      "loss": 0.6315,
+      "step": 8184
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022823636110507983,
+      "loss": 0.3232,
+      "step": 8185
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022822016522465552,
+      "loss": 0.3466,
+      "step": 8186
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022820396809165262,
+      "loss": 0.2463,
+      "step": 8187
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022818776970633055,
+      "loss": 0.5721,
+      "step": 8188
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002281715700689487,
+      "loss": 0.4684,
+      "step": 8189
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022815536917976647,
+      "loss": 0.8092,
+      "step": 8190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002281391670390433,
+      "loss": 1.2417,
+      "step": 8191
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002281229636470386,
+      "loss": 1.1328,
+      "step": 8192
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022810675900401193,
+      "loss": 0.1725,
+      "step": 8193
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022809055311022274,
+      "loss": 0.5785,
+      "step": 8194
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002280743459659305,
+      "loss": 0.2857,
+      "step": 8195
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022805813757139476,
+      "loss": 0.4374,
+      "step": 8196
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022804192792687518,
+      "loss": 0.665,
+      "step": 8197
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002280257170326312,
+      "loss": 0.0222,
+      "step": 8198
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022800950488892252,
+      "loss": 0.0231,
+      "step": 8199
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002279932914960086,
+      "loss": 0.3295,
+      "step": 8200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022797707685414921,
+      "loss": 0.5841,
+      "step": 8201
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000227960860963604,
+      "loss": 0.0113,
+      "step": 8202
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022794464382463255,
+      "loss": 0.9371,
+      "step": 8203
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002279284254374946,
+      "loss": 0.2064,
+      "step": 8204
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022791220580244993,
+      "loss": 0.7778,
+      "step": 8205
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022789598491975818,
+      "loss": 0.0896,
+      "step": 8206
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022787976278967914,
+      "loss": 1.3944,
+      "step": 8207
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022786353941247258,
+      "loss": 0.4054,
+      "step": 8208
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002278473147883983,
+      "loss": 0.3591,
+      "step": 8209
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002278310889177161,
+      "loss": 0.8179,
+      "step": 8210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022781486180068578,
+      "loss": 0.4179,
+      "step": 8211
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002277986334375673,
+      "loss": 0.566,
+      "step": 8212
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022778240382862044,
+      "loss": 2.4319,
+      "step": 8213
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022776617297410508,
+      "loss": 0.0111,
+      "step": 8214
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022774994087428123,
+      "loss": 0.7219,
+      "step": 8215
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022773370752940871,
+      "loss": 0.4527,
+      "step": 8216
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022771747293974756,
+      "loss": 0.532,
+      "step": 8217
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022770123710555776,
+      "loss": 0.5432,
+      "step": 8218
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022768500002709919,
+      "loss": 0.0362,
+      "step": 8219
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022766876170463195,
+      "loss": 0.6962,
+      "step": 8220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022765252213841606,
+      "loss": 0.8875,
+      "step": 8221
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022763628132871152,
+      "loss": 0.506,
+      "step": 8222
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022762003927577853,
+      "loss": 0.2842,
+      "step": 8223
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022760379597987706,
+      "loss": 2.237,
+      "step": 8224
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022758755144126728,
+      "loss": 0.005,
+      "step": 8225
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002275713056602093,
+      "loss": 1.1768,
+      "step": 8226
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022755505863696329,
+      "loss": 0.0047,
+      "step": 8227
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002275388103717894,
+      "loss": 0.6771,
+      "step": 8228
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022752256086494788,
+      "loss": 0.5765,
+      "step": 8229
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002275063101166988,
+      "loss": 0.4892,
+      "step": 8230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274900581273025,
+      "loss": 0.3037,
+      "step": 8231
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274738048970193,
+      "loss": 0.1683,
+      "step": 8232
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022745755042610931,
+      "loss": 0.6786,
+      "step": 8233
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022744129471483285,
+      "loss": 0.2885,
+      "step": 8234
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274250377634504,
+      "loss": 0.2211,
+      "step": 8235
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274087795722221,
+      "loss": 0.3625,
+      "step": 8236
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022739252014140833,
+      "loss": 0.8017,
+      "step": 8237
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022737625947126964,
+      "loss": 0.3341,
+      "step": 8238
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022735999756206616,
+      "loss": 0.3047,
+      "step": 8239
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022734373441405845,
+      "loss": 0.7293,
+      "step": 8240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022732747002750692,
+      "loss": 1.2222,
+      "step": 8241
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022731120440267199,
+      "loss": 0.0212,
+      "step": 8242
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022729493753981416,
+      "loss": 0.4959,
+      "step": 8243
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022727866943919394,
+      "loss": 0.1933,
+      "step": 8244
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022726240010107176,
+      "loss": 0.2662,
+      "step": 8245
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022724612952570823,
+      "loss": 0.4346,
+      "step": 8246
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022722985771336386,
+      "loss": 0.8405,
+      "step": 8247
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002272135846642992,
+      "loss": 0.6859,
+      "step": 8248
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022719731037877487,
+      "loss": 0.7272,
+      "step": 8249
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022718103485705148,
+      "loss": 0.2338,
+      "step": 8250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022716475809938966,
+      "loss": 0.0026,
+      "step": 8251
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022714848010605004,
+      "loss": 0.4619,
+      "step": 8252
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002271322008772933,
+      "loss": 0.3761,
+      "step": 8253
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002271159204133801,
+      "loss": 0.9796,
+      "step": 8254
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002270996387145712,
+      "loss": 0.3778,
+      "step": 8255
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022708335578112726,
+      "loss": 0.0132,
+      "step": 8256
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022706707161330907,
+      "loss": 0.0131,
+      "step": 8257
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022705078621137743,
+      "loss": 0.084,
+      "step": 8258
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002270344995755931,
+      "loss": 0.6443,
+      "step": 8259
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022701821170621677,
+      "loss": 0.288,
+      "step": 8260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022700192260350946,
+      "loss": 0.3199,
+      "step": 8261
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002269856322677319,
+      "loss": 0.5678,
+      "step": 8262
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022696934069914496,
+      "loss": 0.3081,
+      "step": 8263
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022695304789800954,
+      "loss": 0.1836,
+      "step": 8264
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022693675386458657,
+      "loss": 0.6103,
+      "step": 8265
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022692045859913694,
+      "loss": 0.0025,
+      "step": 8266
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022690416210192162,
+      "loss": 0.2223,
+      "step": 8267
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022688786437320154,
+      "loss": 0.3882,
+      "step": 8268
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002268715654132377,
+      "loss": 1.116,
+      "step": 8269
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022685526522229113,
+      "loss": 1.1182,
+      "step": 8270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022683896380062282,
+      "loss": 0.6209,
+      "step": 8271
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022682266114849387,
+      "loss": 0.8998,
+      "step": 8272
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022680635726616524,
+      "loss": 0.5881,
+      "step": 8273
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022679005215389808,
+      "loss": 0.325,
+      "step": 8274
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022677374581195347,
+      "loss": 0.8704,
+      "step": 8275
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002267574382405926,
+      "loss": 0.7984,
+      "step": 8276
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002267411294400765,
+      "loss": 0.2424,
+      "step": 8277
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022672481941066636,
+      "loss": 0.2984,
+      "step": 8278
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002267085081526234,
+      "loss": 0.4074,
+      "step": 8279
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022669219566620884,
+      "loss": 0.302,
+      "step": 8280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002266758819516839,
+      "loss": 0.0058,
+      "step": 8281
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022665956700930972,
+      "loss": 0.0107,
+      "step": 8282
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022664325083934762,
+      "loss": 0.5092,
+      "step": 8283
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022662693344205893,
+      "loss": 0.5312,
+      "step": 8284
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022661061481770489,
+      "loss": 0.9081,
+      "step": 8285
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022659429496654676,
+      "loss": 1.2432,
+      "step": 8286
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022657797388884603,
+      "loss": 0.5364,
+      "step": 8287
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022656165158486393,
+      "loss": 2.379,
+      "step": 8288
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022654532805486193,
+      "loss": 0.3409,
+      "step": 8289
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022652900329910133,
+      "loss": 0.6104,
+      "step": 8290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022651267731784363,
+      "loss": 0.326,
+      "step": 8291
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002264963501113502,
+      "loss": 0.2083,
+      "step": 8292
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022648002167988257,
+      "loss": 3.417,
+      "step": 8293
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002264636920237021,
+      "loss": 0.2194,
+      "step": 8294
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022644736114307045,
+      "loss": 1.0635,
+      "step": 8295
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022643102903824905,
+      "loss": 0.8968,
+      "step": 8296
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022641469570949932,
+      "loss": 0.2488,
+      "step": 8297
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000226398361157083,
+      "loss": 0.5761,
+      "step": 8298
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022638202538126153,
+      "loss": 0.7806,
+      "step": 8299
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022636568838229663,
+      "loss": 0.341,
+      "step": 8300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002263493501604498,
+      "loss": 0.3419,
+      "step": 8301
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022633301071598268,
+      "loss": 0.5054,
+      "step": 8302
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022631667004915698,
+      "loss": 0.1123,
+      "step": 8303
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002263003281602344,
+      "loss": 1.1137,
+      "step": 8304
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022628398504947648,
+      "loss": 0.775,
+      "step": 8305
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022626764071714505,
+      "loss": 0.0239,
+      "step": 8306
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002262512951635019,
+      "loss": 0.1598,
+      "step": 8307
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002262349483888086,
+      "loss": 0.7802,
+      "step": 8308
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022621860039332702,
+      "loss": 0.0754,
+      "step": 8309
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022620225117731902,
+      "loss": 0.0683,
+      "step": 8310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022618590074104623,
+      "loss": 0.8892,
+      "step": 8311
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022616954908477062,
+      "loss": 0.5877,
+      "step": 8312
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000226153196208754,
+      "loss": 0.2703,
+      "step": 8313
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022613684211325827,
+      "loss": 0.0707,
+      "step": 8314
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022612048679854522,
+      "loss": 0.351,
+      "step": 8315
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022610413026487688,
+      "loss": 0.095,
+      "step": 8316
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022608777251251504,
+      "loss": 0.4292,
+      "step": 8317
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022607141354172172,
+      "loss": 1.8748,
+      "step": 8318
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022605505335275894,
+      "loss": 0.478,
+      "step": 8319
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022603869194588857,
+      "loss": 0.0036,
+      "step": 8320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022602232932137265,
+      "loss": 0.9056,
+      "step": 8321
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022600596547947327,
+      "loss": 0.3001,
+      "step": 8322
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022598960042045243,
+      "loss": 0.0333,
+      "step": 8323
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022597323414457217,
+      "loss": 0.2961,
+      "step": 8324
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022595686665209455,
+      "loss": 0.2897,
+      "step": 8325
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022594049794328176,
+      "loss": 0.5942,
+      "step": 8326
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022592412801839578,
+      "loss": 0.1688,
+      "step": 8327
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002259077568776989,
+      "loss": 0.5017,
+      "step": 8328
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022589138452145323,
+      "loss": 0.6402,
+      "step": 8329
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022587501094992093,
+      "loss": 0.1794,
+      "step": 8330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022585863616336418,
+      "loss": 0.0133,
+      "step": 8331
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022584226016204517,
+      "loss": 0.2524,
+      "step": 8332
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022582588294622625,
+      "loss": 1.0031,
+      "step": 8333
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022580950451616964,
+      "loss": 1.5944,
+      "step": 8334
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022579312487213752,
+      "loss": 0.2871,
+      "step": 8335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002257767440143923,
+      "loss": 0.3276,
+      "step": 8336
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022576036194319623,
+      "loss": 0.0488,
+      "step": 8337
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002257439786588117,
+      "loss": 0.7216,
+      "step": 8338
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022572759416150095,
+      "loss": 0.2291,
+      "step": 8339
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002257112084515265,
+      "loss": 0.4775,
+      "step": 8340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022569482152915064,
+      "loss": 4.6282,
+      "step": 8341
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022567843339463587,
+      "loss": 0.4318,
+      "step": 8342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022566204404824447,
+      "loss": 0.0174,
+      "step": 8343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022564565349023907,
+      "loss": 0.0064,
+      "step": 8344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000225629261720882,
+      "loss": 0.5026,
+      "step": 8345
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022561286874043585,
+      "loss": 1.5622,
+      "step": 8346
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022559647454916304,
+      "loss": 0.1373,
+      "step": 8347
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022558007914732622,
+      "loss": 0.6722,
+      "step": 8348
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022556368253518781,
+      "loss": 0.4274,
+      "step": 8349
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022554728471301042,
+      "loss": 0.3487,
+      "step": 8350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022553088568105667,
+      "loss": 1.0663,
+      "step": 8351
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022551448543958914,
+      "loss": 0.0121,
+      "step": 8352
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022549808398887044,
+      "loss": 0.3284,
+      "step": 8353
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022548168132916326,
+      "loss": 0.1342,
+      "step": 8354
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022546527746073022,
+      "loss": 0.2962,
+      "step": 8355
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022544887238383404,
+      "loss": 0.0332,
+      "step": 8356
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022543246609873734,
+      "loss": 0.5122,
+      "step": 8357
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022541605860570293,
+      "loss": 0.371,
+      "step": 8358
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022539964990499355,
+      "loss": 0.7154,
+      "step": 8359
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022538323999687194,
+      "loss": 0.0236,
+      "step": 8360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022536682888160087,
+      "loss": 0.7173,
+      "step": 8361
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022535041655944314,
+      "loss": 0.312,
+      "step": 8362
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022533400303066155,
+      "loss": 0.9727,
+      "step": 8363
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000225317588295519,
+      "loss": 0.0157,
+      "step": 8364
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002253011723542783,
+      "loss": 1.1089,
+      "step": 8365
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002252847552072023,
+      "loss": 0.4488,
+      "step": 8366
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022526833685455395,
+      "loss": 0.1595,
+      "step": 8367
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022525191729659617,
+      "loss": 0.2178,
+      "step": 8368
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022523549653359184,
+      "loss": 0.4394,
+      "step": 8369
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022521907456580394,
+      "loss": 0.3636,
+      "step": 8370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022520265139349552,
+      "loss": 0.5689,
+      "step": 8371
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022518622701692945,
+      "loss": 0.0771,
+      "step": 8372
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022516980143636882,
+      "loss": 0.4296,
+      "step": 8373
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002251533746520766,
+      "loss": 0.6157,
+      "step": 8374
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002251369466643159,
+      "loss": 0.3743,
+      "step": 8375
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022512051747334975,
+      "loss": 0.8456,
+      "step": 8376
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002251040870794412,
+      "loss": 0.843,
+      "step": 8377
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022508765548285347,
+      "loss": 1.7217,
+      "step": 8378
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022507122268384963,
+      "loss": 0.94,
+      "step": 8379
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022505478868269283,
+      "loss": 0.3808,
+      "step": 8380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022503835347964618,
+      "loss": 0.1911,
+      "step": 8381
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000225021917074973,
+      "loss": 0.3331,
+      "step": 8382
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022500547946893636,
+      "loss": 0.1291,
+      "step": 8383
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002249890406617995,
+      "loss": 0.3341,
+      "step": 8384
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022497260065382576,
+      "loss": 0.5862,
+      "step": 8385
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002249561594452783,
+      "loss": 0.2757,
+      "step": 8386
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022493971703642046,
+      "loss": 0.6178,
+      "step": 8387
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002249232734275155,
+      "loss": 0.2776,
+      "step": 8388
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002249068286188268,
+      "loss": 0.486,
+      "step": 8389
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002248903826106176,
+      "loss": 0.5368,
+      "step": 8390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002248739354031514,
+      "loss": 0.4487,
+      "step": 8391
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022485748699669142,
+      "loss": 0.2867,
+      "step": 8392
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022484103739150112,
+      "loss": 0.4129,
+      "step": 8393
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000224824586587844,
+      "loss": 0.0028,
+      "step": 8394
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022480813458598337,
+      "loss": 0.4261,
+      "step": 8395
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022479168138618273,
+      "loss": 0.4608,
+      "step": 8396
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002247752269887056,
+      "loss": 0.3801,
+      "step": 8397
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022475877139381542,
+      "loss": 0.2191,
+      "step": 8398
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002247423146017757,
+      "loss": 0.2041,
+      "step": 8399
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022472585661284996,
+      "loss": 2.2702,
+      "step": 8400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002247093974273018,
+      "loss": 0.3579,
+      "step": 8401
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022469293704539478,
+      "loss": 0.7816,
+      "step": 8402
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022467647546739244,
+      "loss": 0.3357,
+      "step": 8403
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022466001269355845,
+      "loss": 0.4364,
+      "step": 8404
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022464354872415637,
+      "loss": 0.0021,
+      "step": 8405
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022462708355944997,
+      "loss": 0.0041,
+      "step": 8406
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002246106171997027,
+      "loss": 0.6824,
+      "step": 8407
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022459414964517844,
+      "loss": 0.0615,
+      "step": 8408
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022457768089614082,
+      "loss": 0.2604,
+      "step": 8409
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022456121095285357,
+      "loss": 0.0013,
+      "step": 8410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022454473981558038,
+      "loss": 1.1085,
+      "step": 8411
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002245282674845851,
+      "loss": 1.0031,
+      "step": 8412
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002245117939601315,
+      "loss": 0.5749,
+      "step": 8413
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022449531924248332,
+      "loss": 0.8032,
+      "step": 8414
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022447884333190432,
+      "loss": 0.0088,
+      "step": 8415
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022446236622865852,
+      "loss": 0.7959,
+      "step": 8416
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022444588793300965,
+      "loss": 0.387,
+      "step": 8417
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002244294084452216,
+      "loss": 0.3502,
+      "step": 8418
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002244129277655583,
+      "loss": 0.5992,
+      "step": 8419
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022439644589428365,
+      "loss": 0.1507,
+      "step": 8420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022437996283166154,
+      "loss": 0.4296,
+      "step": 8421
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022436347857795593,
+      "loss": 0.2942,
+      "step": 8422
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022434699313343088,
+      "loss": 2.1213,
+      "step": 8423
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022433050649835028,
+      "loss": 0.4303,
+      "step": 8424
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022431401867297818,
+      "loss": 0.1737,
+      "step": 8425
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002242975296575786,
+      "loss": 0.7613,
+      "step": 8426
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002242810394524156,
+      "loss": 0.1616,
+      "step": 8427
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002242645480577532,
+      "loss": 0.9463,
+      "step": 8428
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022424805547385557,
+      "loss": 0.239,
+      "step": 8429
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002242315617009867,
+      "loss": 0.0216,
+      "step": 8430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022421506673941079,
+      "loss": 0.5002,
+      "step": 8431
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022419857058939203,
+      "loss": 0.3459,
+      "step": 8432
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022418207325119443,
+      "loss": 0.0089,
+      "step": 8433
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022416557472508232,
+      "loss": 0.8542,
+      "step": 8434
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002241490750113198,
+      "loss": 0.5155,
+      "step": 8435
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022413257411017115,
+      "loss": 0.253,
+      "step": 8436
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022411607202190055,
+      "loss": 0.0006,
+      "step": 8437
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022409956874677234,
+      "loss": 0.0039,
+      "step": 8438
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022408306428505073,
+      "loss": 0.9001,
+      "step": 8439
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022406655863700005,
+      "loss": 0.2478,
+      "step": 8440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002240500518028845,
+      "loss": 1.5464,
+      "step": 8441
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022403354378296855,
+      "loss": 0.4493,
+      "step": 8442
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002240170345775166,
+      "loss": 0.2694,
+      "step": 8443
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002240005241867928,
+      "loss": 0.2769,
+      "step": 8444
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022398401261106168,
+      "loss": 0.5093,
+      "step": 8445
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002239674998505877,
+      "loss": 0.9106,
+      "step": 8446
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022395098590563515,
+      "loss": 0.3439,
+      "step": 8447
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002239344707764686,
+      "loss": 0.4817,
+      "step": 8448
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022391795446335242,
+      "loss": 0.0969,
+      "step": 8449
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022390143696655114,
+      "loss": 0.7231,
+      "step": 8450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022388491828632931,
+      "loss": 1.0555,
+      "step": 8451
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022386839842295132,
+      "loss": 0.4746,
+      "step": 8452
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022385187737668181,
+      "loss": 0.7461,
+      "step": 8453
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022383535514778537,
+      "loss": 1.161,
+      "step": 8454
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022381883173652653,
+      "loss": 0.0184,
+      "step": 8455
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022380230714316983,
+      "loss": 0.7398,
+      "step": 8456
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022378578136797998,
+      "loss": 0.6886,
+      "step": 8457
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022376925441122157,
+      "loss": 0.4809,
+      "step": 8458
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022375272627315926,
+      "loss": 2.9492,
+      "step": 8459
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022373619695405768,
+      "loss": 0.5613,
+      "step": 8460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022371966645418163,
+      "loss": 0.7093,
+      "step": 8461
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022370313477379573,
+      "loss": 0.312,
+      "step": 8462
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022368660191316475,
+      "loss": 0.4926,
+      "step": 8463
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022367006787255345,
+      "loss": 0.0191,
+      "step": 8464
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022365353265222653,
+      "loss": 0.3213,
+      "step": 8465
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022363699625244888,
+      "loss": 1.0947,
+      "step": 8466
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022362045867348518,
+      "loss": 0.489,
+      "step": 8467
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022360391991560038,
+      "loss": 0.6292,
+      "step": 8468
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022358737997905926,
+      "loss": 0.2414,
+      "step": 8469
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022357083886412666,
+      "loss": 0.4484,
+      "step": 8470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022355429657106748,
+      "loss": 0.0056,
+      "step": 8471
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002235377531001467,
+      "loss": 0.4446,
+      "step": 8472
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002235212084516291,
+      "loss": 0.0004,
+      "step": 8473
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002235046626257797,
+      "loss": 0.002,
+      "step": 8474
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022348811562286348,
+      "loss": 1.2784,
+      "step": 8475
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022347156744314536,
+      "loss": 0.4509,
+      "step": 8476
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022345501808689034,
+      "loss": 0.0079,
+      "step": 8477
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022343846755436346,
+      "loss": 0.0038,
+      "step": 8478
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002234219158458297,
+      "loss": 0.3577,
+      "step": 8479
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022340536296155416,
+      "loss": 0.1863,
+      "step": 8480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022338880890180194,
+      "loss": 0.8362,
+      "step": 8481
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022337225366683805,
+      "loss": 0.4398,
+      "step": 8482
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022335569725692766,
+      "loss": 0.5465,
+      "step": 8483
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022333913967233584,
+      "loss": 0.0034,
+      "step": 8484
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022332258091332783,
+      "loss": 0.3299,
+      "step": 8485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022330602098016863,
+      "loss": 0.4515,
+      "step": 8486
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002232894598731236,
+      "loss": 0.3426,
+      "step": 8487
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002232728975924578,
+      "loss": 0.0568,
+      "step": 8488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022325633413843656,
+      "loss": 0.4477,
+      "step": 8489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022323976951132507,
+      "loss": 0.4705,
+      "step": 8490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022322320371138857,
+      "loss": 0.3596,
+      "step": 8491
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022320663673889237,
+      "loss": 0.0013,
+      "step": 8492
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022319006859410174,
+      "loss": 0.0186,
+      "step": 8493
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000223173499277282,
+      "loss": 0.6151,
+      "step": 8494
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002231569287886985,
+      "loss": 2.2647,
+      "step": 8495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022314035712861656,
+      "loss": 0.3678,
+      "step": 8496
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022312378429730157,
+      "loss": 0.3718,
+      "step": 8497
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022310721029501892,
+      "loss": 0.1739,
+      "step": 8498
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022309063512203402,
+      "loss": 0.6113,
+      "step": 8499
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022307405877861232,
+      "loss": 1.065,
+      "step": 8500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002230574812650192,
+      "loss": 0.2482,
+      "step": 8501
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002230409025815202,
+      "loss": 0.0022,
+      "step": 8502
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022302432272838072,
+      "loss": 2.0723,
+      "step": 8503
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022300774170586635,
+      "loss": 0.3693,
+      "step": 8504
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002229911595142425,
+      "loss": 0.3078,
+      "step": 8505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022297457615377483,
+      "loss": 0.1379,
+      "step": 8506
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022295799162472884,
+      "loss": 0.0022,
+      "step": 8507
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002229414059273701,
+      "loss": 0.5237,
+      "step": 8508
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002229248190619642,
+      "loss": 0.1726,
+      "step": 8509
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022290823102877677,
+      "loss": 0.3022,
+      "step": 8510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022289164182807345,
+      "loss": 0.2743,
+      "step": 8511
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022287505146011987,
+      "loss": 0.3797,
+      "step": 8512
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022285845992518172,
+      "loss": 0.746,
+      "step": 8513
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022284186722352467,
+      "loss": 0.2947,
+      "step": 8514
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022282527335541446,
+      "loss": 0.8549,
+      "step": 8515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022280867832111676,
+      "loss": 1.0635,
+      "step": 8516
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022279208212089738,
+      "loss": 0.4261,
+      "step": 8517
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022277548475502205,
+      "loss": 0.0096,
+      "step": 8518
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022275888622375653,
+      "loss": 0.6491,
+      "step": 8519
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022274228652736661,
+      "loss": 0.6269,
+      "step": 8520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002227256856661182,
+      "loss": 0.3261,
+      "step": 8521
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002227090836402771,
+      "loss": 0.2577,
+      "step": 8522
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002226924804501091,
+      "loss": 0.6193,
+      "step": 8523
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022267587609588012,
+      "loss": 1.1055,
+      "step": 8524
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002226592705778561,
+      "loss": 0.3768,
+      "step": 8525
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022264266389630284,
+      "loss": 0.0045,
+      "step": 8526
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002226260560514864,
+      "loss": 0.6156,
+      "step": 8527
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002226094470436726,
+      "loss": 0.0155,
+      "step": 8528
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022259283687312752,
+      "loss": 0.9287,
+      "step": 8529
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022257622554011715,
+      "loss": 1.0075,
+      "step": 8530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022255961304490734,
+      "loss": 0.6903,
+      "step": 8531
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002225429993877643,
+      "loss": 0.971,
+      "step": 8532
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022252638456895404,
+      "loss": 0.6342,
+      "step": 8533
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002225097685887425,
+      "loss": 0.0088,
+      "step": 8534
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022249315144739587,
+      "loss": 0.4427,
+      "step": 8535
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002224765331451802,
+      "loss": 0.2513,
+      "step": 8536
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022245991368236161,
+      "loss": 1.331,
+      "step": 8537
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022244329305920628,
+      "loss": 0.4004,
+      "step": 8538
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022242667127598032,
+      "loss": 0.2837,
+      "step": 8539
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002224100483329499,
+      "loss": 0.0059,
+      "step": 8540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022239342423038128,
+      "loss": 1.2305,
+      "step": 8541
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002223767989685406,
+      "loss": 1.8194,
+      "step": 8542
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022236017254769407,
+      "loss": 0.2008,
+      "step": 8543
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000222343544968108,
+      "loss": 0.4602,
+      "step": 8544
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022232691623004866,
+      "loss": 1.8731,
+      "step": 8545
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022231028633378226,
+      "loss": 0.5607,
+      "step": 8546
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022229365527957516,
+      "loss": 0.6926,
+      "step": 8547
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022227702306769364,
+      "loss": 0.2218,
+      "step": 8548
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022226038969840408,
+      "loss": 0.2936,
+      "step": 8549
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022224375517197287,
+      "loss": 0.7998,
+      "step": 8550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002222271194886663,
+      "loss": 0.0002,
+      "step": 8551
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022221048264875083,
+      "loss": 0.7073,
+      "step": 8552
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022219384465249286,
+      "loss": 3.5508,
+      "step": 8553
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002221772055001588,
+      "loss": 0.5191,
+      "step": 8554
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022216056519201512,
+      "loss": 0.0006,
+      "step": 8555
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022214392372832832,
+      "loss": 0.0039,
+      "step": 8556
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022212728110936479,
+      "loss": 1.1046,
+      "step": 8557
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002221106373353911,
+      "loss": 0.0073,
+      "step": 8558
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022209399240667382,
+      "loss": 0.2758,
+      "step": 8559
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022207734632347944,
+      "loss": 0.8091,
+      "step": 8560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022206069908607454,
+      "loss": 0.0023,
+      "step": 8561
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002220440506947257,
+      "loss": 0.3643,
+      "step": 8562
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022202740114969947,
+      "loss": 0.2454,
+      "step": 8563
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022201075045126254,
+      "loss": 1.0036,
+      "step": 8564
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002219940985996815,
+      "loss": 0.283,
+      "step": 8565
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000221977445595223,
+      "loss": 3.5047,
+      "step": 8566
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002219607914381538,
+      "loss": 0.1272,
+      "step": 8567
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022194413612874048,
+      "loss": 0.3959,
+      "step": 8568
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022192747966724976,
+      "loss": 0.0545,
+      "step": 8569
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022191082205394848,
+      "loss": 0.3236,
+      "step": 8570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022189416328910323,
+      "loss": 0.428,
+      "step": 8571
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022187750337298087,
+      "loss": 0.4686,
+      "step": 8572
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002218608423058482,
+      "loss": 0.3532,
+      "step": 8573
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022184418008797195,
+      "loss": 0.6764,
+      "step": 8574
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000221827516719619,
+      "loss": 0.0027,
+      "step": 8575
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002218108522010562,
+      "loss": 0.5252,
+      "step": 8576
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002217941865325503,
+      "loss": 0.001,
+      "step": 8577
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002217775197143683,
+      "loss": 0.4286,
+      "step": 8578
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022176085174677705,
+      "loss": 0.3665,
+      "step": 8579
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022174418263004342,
+      "loss": 0.7016,
+      "step": 8580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002217275123644344,
+      "loss": 0.6655,
+      "step": 8581
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022171084095021695,
+      "loss": 0.449,
+      "step": 8582
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022169416838765797,
+      "loss": 0.001,
+      "step": 8583
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022167749467702448,
+      "loss": 6.2202,
+      "step": 8584
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002216608198185835,
+      "loss": 0.3454,
+      "step": 8585
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000221644143812602,
+      "loss": 0.3937,
+      "step": 8586
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002216274666593471,
+      "loss": 2.238,
+      "step": 8587
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002216107883590858,
+      "loss": 0.3427,
+      "step": 8588
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002215941089120852,
+      "loss": 0.5727,
+      "step": 8589
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002215774283186124,
+      "loss": 4.7527,
+      "step": 8590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022156074657893454,
+      "loss": 0.4382,
+      "step": 8591
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022154406369331866,
+      "loss": 0.7356,
+      "step": 8592
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022152737966203202,
+      "loss": 0.9355,
+      "step": 8593
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022151069448534172,
+      "loss": 0.759,
+      "step": 8594
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022149400816351498,
+      "loss": 0.3712,
+      "step": 8595
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000221477320696819,
+      "loss": 0.1523,
+      "step": 8596
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022146063208552095,
+      "loss": 1.0869,
+      "step": 8597
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022144394232988818,
+      "loss": 0.2291,
+      "step": 8598
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002214272514301879,
+      "loss": 0.0769,
+      "step": 8599
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022141055938668734,
+      "loss": 0.4531,
+      "step": 8600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022139386619965387,
+      "loss": 0.241,
+      "step": 8601
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002213771718693548,
+      "loss": 3.2272,
+      "step": 8602
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022136047639605743,
+      "loss": 1.2544,
+      "step": 8603
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022134377978002908,
+      "loss": 1.0365,
+      "step": 8604
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022132708202153727,
+      "loss": 0.0521,
+      "step": 8605
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022131038312084923,
+      "loss": 0.0147,
+      "step": 8606
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002212936830782324,
+      "loss": 0.3607,
+      "step": 8607
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002212769818939543,
+      "loss": 0.2033,
+      "step": 8608
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022126027956828224,
+      "loss": 0.0003,
+      "step": 8609
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022124357610148377,
+      "loss": 0.017,
+      "step": 8610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022122687149382634,
+      "loss": 0.4037,
+      "step": 8611
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022121016574557745,
+      "loss": 0.5847,
+      "step": 8612
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022119345885700467,
+      "loss": 0.4934,
+      "step": 8613
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022117675082837548,
+      "loss": 0.0082,
+      "step": 8614
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022116004165995744,
+      "loss": 3.5371,
+      "step": 8615
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022114333135201812,
+      "loss": 0.4355,
+      "step": 8616
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022112661990482518,
+      "loss": 0.0013,
+      "step": 8617
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002211099073186461,
+      "loss": 0.2367,
+      "step": 8618
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002210931935937486,
+      "loss": 0.3131,
+      "step": 8619
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002210764787304003,
+      "loss": 0.3583,
+      "step": 8620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022105976272886885,
+      "loss": 0.8522,
+      "step": 8621
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022104304558942194,
+      "loss": 0.2203,
+      "step": 8622
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002210263273123273,
+      "loss": 0.0052,
+      "step": 8623
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002210096078978526,
+      "loss": 0.1248,
+      "step": 8624
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022099288734626563,
+      "loss": 1.1251,
+      "step": 8625
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022097616565783408,
+      "loss": 0.4189,
+      "step": 8626
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022095944283282576,
+      "loss": 0.0047,
+      "step": 8627
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022094271887150853,
+      "loss": 0.7627,
+      "step": 8628
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022092599377415004,
+      "loss": 0.1832,
+      "step": 8629
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022090926754101824,
+      "loss": 0.8068,
+      "step": 8630
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022089254017238093,
+      "loss": 0.2928,
+      "step": 8631
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000220875811668506,
+      "loss": 0.8651,
+      "step": 8632
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000220875811668506,
+      "loss": 0.9307,
+      "step": 8633
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022085908202966128,
+      "loss": 1.01,
+      "step": 8634
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022084235125611475,
+      "loss": 0.607,
+      "step": 8635
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022082561934813426,
+      "loss": 0.0008,
+      "step": 8636
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022080888630598777,
+      "loss": 0.6828,
+      "step": 8637
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022079215212994323,
+      "loss": 0.0026,
+      "step": 8638
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022077541682026863,
+      "loss": 1.2218,
+      "step": 8639
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022075868037723195,
+      "loss": 0.8746,
+      "step": 8640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022074194280110123,
+      "loss": 0.0443,
+      "step": 8641
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002207252040921444,
+      "loss": 0.6831,
+      "step": 8642
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002207084642506296,
+      "loss": 0.5846,
+      "step": 8643
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002206917232768249,
+      "loss": 0.0024,
+      "step": 8644
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002206749811709983,
+      "loss": 1.669,
+      "step": 8645
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022065823793341794,
+      "loss": 0.4099,
+      "step": 8646
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022064149356435198,
+      "loss": 0.0023,
+      "step": 8647
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022062474806406847,
+      "loss": 0.4506,
+      "step": 8648
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022060800143283563,
+      "loss": 0.3121,
+      "step": 8649
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022059125367092163,
+      "loss": 0.9235,
+      "step": 8650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022057450477859465,
+      "loss": 0.1637,
+      "step": 8651
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022055775475612288,
+      "loss": 0.2242,
+      "step": 8652
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022054100360377455,
+      "loss": 0.3471,
+      "step": 8653
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022052425132181792,
+      "loss": 0.5369,
+      "step": 8654
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022050749791052127,
+      "loss": 0.3118,
+      "step": 8655
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022049074337015284,
+      "loss": 0.3271,
+      "step": 8656
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022047398770098093,
+      "loss": 4.5941,
+      "step": 8657
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022045723090327393,
+      "loss": 4.4245,
+      "step": 8658
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022044047297730006,
+      "loss": 0.4845,
+      "step": 8659
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022042371392332776,
+      "loss": 0.4597,
+      "step": 8660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022040695374162537,
+      "loss": 0.5926,
+      "step": 8661
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022039019243246126,
+      "loss": 1.0066,
+      "step": 8662
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022037342999610388,
+      "loss": 0.237,
+      "step": 8663
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022035666643282164,
+      "loss": 1.4674,
+      "step": 8664
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022033990174288295,
+      "loss": 0.8776,
+      "step": 8665
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022032313592655633,
+      "loss": 0.3582,
+      "step": 8666
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002203063689841102,
+      "loss": 0.0019,
+      "step": 8667
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002202896009158131,
+      "loss": 0.2968,
+      "step": 8668
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002202728317219335,
+      "loss": 0.4145,
+      "step": 8669
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022025606140274004,
+      "loss": 0.5611,
+      "step": 8670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022023928995850114,
+      "loss": 0.3518,
+      "step": 8671
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022022251738948546,
+      "loss": 0.0004,
+      "step": 8672
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002202057436959615,
+      "loss": 1.3623,
+      "step": 8673
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022018896887819798,
+      "loss": 0.2495,
+      "step": 8674
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002201721929364634,
+      "loss": 0.9903,
+      "step": 8675
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002201554158710265,
+      "loss": 0.2304,
+      "step": 8676
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022013863768215593,
+      "loss": 0.1785,
+      "step": 8677
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002201218583701203,
+      "loss": 0.1229,
+      "step": 8678
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002201050779351883,
+      "loss": 0.0051,
+      "step": 8679
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022008829637762878,
+      "loss": 0.1697,
+      "step": 8680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022007151369771036,
+      "loss": 0.7345,
+      "step": 8681
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022005472989570184,
+      "loss": 0.1829,
+      "step": 8682
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022003794497187185,
+      "loss": 0.2765,
+      "step": 8683
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022002115892648936,
+      "loss": 0.4769,
+      "step": 8684
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002200043717598231,
+      "loss": 0.0015,
+      "step": 8685
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021998758347214188,
+      "loss": 0.2227,
+      "step": 8686
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021997079406371455,
+      "loss": 0.6967,
+      "step": 8687
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021995400353480994,
+      "loss": 0.5009,
+      "step": 8688
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021993721188569697,
+      "loss": 0.5382,
+      "step": 8689
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021992041911664454,
+      "loss": 4.1333,
+      "step": 8690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021990362522792146,
+      "loss": 3.004,
+      "step": 8691
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002198868302197968,
+      "loss": 0.473,
+      "step": 8692
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021987003409253944,
+      "loss": 0.5361,
+      "step": 8693
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021985323684641832,
+      "loss": 0.5987,
+      "step": 8694
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002198364384817024,
+      "loss": 0.1023,
+      "step": 8695
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002198196389986608,
+      "loss": 0.5614,
+      "step": 8696
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021980283839756246,
+      "loss": 0.2642,
+      "step": 8697
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002197860366786764,
+      "loss": 0.2508,
+      "step": 8698
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002197692338422717,
+      "loss": 0.0028,
+      "step": 8699
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002197524298886174,
+      "loss": 0.2894,
+      "step": 8700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021973562481798265,
+      "loss": 0.5811,
+      "step": 8701
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021971881863063652,
+      "loss": 0.7871,
+      "step": 8702
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021970201132684818,
+      "loss": 0.3349,
+      "step": 8703
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002196852029068867,
+      "loss": 0.2077,
+      "step": 8704
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002196683933710213,
+      "loss": 0.7759,
+      "step": 8705
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002196515827195211,
+      "loss": 0.3862,
+      "step": 8706
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021963477095265536,
+      "loss": 0.7271,
+      "step": 8707
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021961795807069328,
+      "loss": 0.4517,
+      "step": 8708
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021960114407390405,
+      "loss": 0.0465,
+      "step": 8709
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021958432896255697,
+      "loss": 0.0053,
+      "step": 8710
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021956751273692127,
+      "loss": 0.7042,
+      "step": 8711
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021955069539726632,
+      "loss": 0.3268,
+      "step": 8712
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021953387694386133,
+      "loss": 0.3316,
+      "step": 8713
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021951705737697564,
+      "loss": 0.3373,
+      "step": 8714
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021950023669687862,
+      "loss": 1.6725,
+      "step": 8715
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021948341490383964,
+      "loss": 0.5886,
+      "step": 8716
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000219466591998128,
+      "loss": 0.0623,
+      "step": 8717
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021944976798001317,
+      "loss": 0.2652,
+      "step": 8718
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021943294284976458,
+      "loss": 3.9247,
+      "step": 8719
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021941611660765157,
+      "loss": 0.0016,
+      "step": 8720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002193992892539436,
+      "loss": 0.8008,
+      "step": 8721
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002193824607889102,
+      "loss": 0.4626,
+      "step": 8722
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021936563121282085,
+      "loss": 0.0047,
+      "step": 8723
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000219348800525945,
+      "loss": 2.7266,
+      "step": 8724
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021933196872855215,
+      "loss": 0.6241,
+      "step": 8725
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002193151358209119,
+      "loss": 0.5346,
+      "step": 8726
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021929830180329374,
+      "loss": 0.5356,
+      "step": 8727
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021928146667596736,
+      "loss": 0.6409,
+      "step": 8728
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021926463043920222,
+      "loss": 0.4471,
+      "step": 8729
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021924779309326798,
+      "loss": 0.6351,
+      "step": 8730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021923095463843425,
+      "loss": 1.2669,
+      "step": 8731
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021921411507497068,
+      "loss": 0.7464,
+      "step": 8732
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021919727440314698,
+      "loss": 0.7743,
+      "step": 8733
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021918043262323273,
+      "loss": 1.2121,
+      "step": 8734
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002191635897354977,
+      "loss": 0.294,
+      "step": 8735
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021914674574021156,
+      "loss": 0.1488,
+      "step": 8736
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002191299006376441,
+      "loss": 0.2684,
+      "step": 8737
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021911305442806497,
+      "loss": 0.9114,
+      "step": 8738
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021909620711174407,
+      "loss": 0.6119,
+      "step": 8739
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021907935868895103,
+      "loss": 0.3226,
+      "step": 8740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021906250915995577,
+      "loss": 0.5663,
+      "step": 8741
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021904565852502811,
+      "loss": 0.1303,
+      "step": 8742
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002190288067844378,
+      "loss": 0.4891,
+      "step": 8743
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021901195393845478,
+      "loss": 0.1841,
+      "step": 8744
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002189950999873489,
+      "loss": 0.49,
+      "step": 8745
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021897824493139005,
+      "loss": 0.0593,
+      "step": 8746
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021896138877084808,
+      "loss": 0.1767,
+      "step": 8747
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021894453150599297,
+      "loss": 2.6641,
+      "step": 8748
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002189276731370947,
+      "loss": 0.3046,
+      "step": 8749
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021891081366442317,
+      "loss": 0.0885,
+      "step": 8750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021889395308824836,
+      "loss": 0.787,
+      "step": 8751
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002188770914088403,
+      "loss": 0.1924,
+      "step": 8752
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021886022862646897,
+      "loss": 0.2731,
+      "step": 8753
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021884336474140448,
+      "loss": 1.9686,
+      "step": 8754
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021882649975391675,
+      "loss": 1.9004,
+      "step": 8755
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000218809633664276,
+      "loss": 0.2148,
+      "step": 8756
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002187927664727522,
+      "loss": 0.4486,
+      "step": 8757
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021877589817961547,
+      "loss": 2.2168,
+      "step": 8758
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021875902878513596,
+      "loss": 0.1676,
+      "step": 8759
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021874215828958382,
+      "loss": 0.3999,
+      "step": 8760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021872528669322915,
+      "loss": 0.3261,
+      "step": 8761
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021870841399634217,
+      "loss": 0.8384,
+      "step": 8762
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021869154019919307,
+      "loss": 0.6857,
+      "step": 8763
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021867466530205203,
+      "loss": 0.2105,
+      "step": 8764
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002186577893051893,
+      "loss": 0.0018,
+      "step": 8765
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002186409122088751,
+      "loss": 0.4115,
+      "step": 8766
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021862403401337968,
+      "loss": 0.5456,
+      "step": 8767
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021860715471897342,
+      "loss": 0.4421,
+      "step": 8768
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021859027432592654,
+      "loss": 0.4596,
+      "step": 8769
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021857339283450926,
+      "loss": 0.2352,
+      "step": 8770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021855651024499208,
+      "loss": 1.3201,
+      "step": 8771
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021853962655764527,
+      "loss": 0.8194,
+      "step": 8772
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002185227417727392,
+      "loss": 0.3445,
+      "step": 8773
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021850585589054422,
+      "loss": 0.4853,
+      "step": 8774
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021848896891133086,
+      "loss": 0.0436,
+      "step": 8775
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021847208083536937,
+      "loss": 0.6966,
+      "step": 8776
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021845519166293029,
+      "loss": 0.8579,
+      "step": 8777
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002184383013942841,
+      "loss": 0.3759,
+      "step": 8778
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021842141002970114,
+      "loss": 0.8152,
+      "step": 8779
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021840451756945206,
+      "loss": 0.6419,
+      "step": 8780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021838762401380724,
+      "loss": 0.011,
+      "step": 8781
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002183707293630373,
+      "loss": 0.002,
+      "step": 8782
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002183538336174127,
+      "loss": 0.0452,
+      "step": 8783
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021833693677720406,
+      "loss": 0.1787,
+      "step": 8784
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021832003884268193,
+      "loss": 0.186,
+      "step": 8785
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021830313981411692,
+      "loss": 0.2287,
+      "step": 8786
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021828623969177963,
+      "loss": 0.0235,
+      "step": 8787
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021826933847594074,
+      "loss": 0.3764,
+      "step": 8788
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021825243616687077,
+      "loss": 0.324,
+      "step": 8789
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021823553276484046,
+      "loss": 0.7625,
+      "step": 8790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021821862827012058,
+      "loss": 0.4861,
+      "step": 8791
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021820172268298172,
+      "loss": 0.0133,
+      "step": 8792
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021818481600369458,
+      "loss": 0.6518,
+      "step": 8793
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021816790823253002,
+      "loss": 0.2534,
+      "step": 8794
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021815099936975868,
+      "loss": 0.6811,
+      "step": 8795
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021813408941565132,
+      "loss": 0.2067,
+      "step": 8796
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021811717837047883,
+      "loss": 0.3014,
+      "step": 8797
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000218100266234512,
+      "loss": 0.2895,
+      "step": 8798
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021808335300802152,
+      "loss": 0.5243,
+      "step": 8799
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021806643869127837,
+      "loss": 1.4847,
+      "step": 8800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002180495232845533,
+      "loss": 0.0038,
+      "step": 8801
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002180326067881173,
+      "loss": 0.3624,
+      "step": 8802
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021801568920224124,
+      "loss": 1.219,
+      "step": 8803
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021799877052719588,
+      "loss": 0.339,
+      "step": 8804
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021798185076325237,
+      "loss": 0.4835,
+      "step": 8805
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021796492991068151,
+      "loss": 0.1312,
+      "step": 8806
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021794800796975426,
+      "loss": 0.0247,
+      "step": 8807
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021793108494074168,
+      "loss": 0.3711,
+      "step": 8808
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021791416082391472,
+      "loss": 0.4936,
+      "step": 8809
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021789723561954437,
+      "loss": 0.5875,
+      "step": 8810
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021788030932790172,
+      "loss": 0.5939,
+      "step": 8811
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002178633819492578,
+      "loss": 0.2771,
+      "step": 8812
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021784645348388366,
+      "loss": 0.1685,
+      "step": 8813
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021782952393205042,
+      "loss": 0.2139,
+      "step": 8814
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021781259329402907,
+      "loss": 0.5387,
+      "step": 8815
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021779566157009085,
+      "loss": 0.5607,
+      "step": 8816
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002177787287605069,
+      "loss": 0.0012,
+      "step": 8817
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002177617948655483,
+      "loss": 0.1029,
+      "step": 8818
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002177448598854862,
+      "loss": 0.6446,
+      "step": 8819
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002177279238205919,
+      "loss": 0.0366,
+      "step": 8820
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021771098667113653,
+      "loss": 0.4341,
+      "step": 8821
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021769404843739135,
+      "loss": 0.4939,
+      "step": 8822
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021767710911962753,
+      "loss": 0.4859,
+      "step": 8823
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002176601687181164,
+      "loss": 1.713,
+      "step": 8824
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021764322723312923,
+      "loss": 0.3344,
+      "step": 8825
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002176262846649373,
+      "loss": 0.9678,
+      "step": 8826
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021760934101381187,
+      "loss": 0.1818,
+      "step": 8827
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021759239628002435,
+      "loss": 0.4437,
+      "step": 8828
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021757545046384606,
+      "loss": 0.0013,
+      "step": 8829
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021755850356554825,
+      "loss": 0.0369,
+      "step": 8830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002175415555854025,
+      "loss": 0.0222,
+      "step": 8831
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021752460652368008,
+      "loss": 0.259,
+      "step": 8832
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021750765638065242,
+      "loss": 0.0355,
+      "step": 8833
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021749070515659092,
+      "loss": 0.4716,
+      "step": 8834
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021747375285176713,
+      "loss": 0.6401,
+      "step": 8835
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002174567994664524,
+      "loss": 0.3464,
+      "step": 8836
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021743984500091827,
+      "loss": 0.4707,
+      "step": 8837
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021742288945543627,
+      "loss": 0.3783,
+      "step": 8838
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021740593283027786,
+      "loss": 0.0338,
+      "step": 8839
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002173889751257146,
+      "loss": 0.3263,
+      "step": 8840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021737201634201806,
+      "loss": 0.2198,
+      "step": 8841
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021735505647945978,
+      "loss": 0.2699,
+      "step": 8842
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021733809553831134,
+      "loss": 0.8909,
+      "step": 8843
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021732113351884434,
+      "loss": 0.0567,
+      "step": 8844
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021730417042133043,
+      "loss": 1.5946,
+      "step": 8845
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021728720624604128,
+      "loss": 0.0012,
+      "step": 8846
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021727024099324854,
+      "loss": 0.2405,
+      "step": 8847
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021725327466322377,
+      "loss": 0.1895,
+      "step": 8848
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021723630725623875,
+      "loss": 0.6809,
+      "step": 8849
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002172193387725652,
+      "loss": 0.4959,
+      "step": 8850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021720236921247477,
+      "loss": 0.7347,
+      "step": 8851
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021718539857623934,
+      "loss": 0.3306,
+      "step": 8852
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002171684268641305,
+      "loss": 0.0016,
+      "step": 8853
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002171514540764202,
+      "loss": 1.7875,
+      "step": 8854
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021713448021338006,
+      "loss": 3.408,
+      "step": 8855
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021711750527528205,
+      "loss": 0.0271,
+      "step": 8856
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021710052926239787,
+      "loss": 0.3248,
+      "step": 8857
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002170835521749994,
+      "loss": 0.0031,
+      "step": 8858
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021706657401335854,
+      "loss": 1.0043,
+      "step": 8859
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021704959477774718,
+      "loss": 0.5513,
+      "step": 8860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021703261446843713,
+      "loss": 0.0016,
+      "step": 8861
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021701563308570044,
+      "loss": 1.5444,
+      "step": 8862
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021699865062980888,
+      "loss": 0.4013,
+      "step": 8863
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021698166710103453,
+      "loss": 0.0248,
+      "step": 8864
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002169646824996493,
+      "loss": 0.0143,
+      "step": 8865
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002169476968259252,
+      "loss": 0.5314,
+      "step": 8866
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021693071008013418,
+      "loss": 0.3135,
+      "step": 8867
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021691372226254827,
+      "loss": 0.0084,
+      "step": 8868
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021689673337343956,
+      "loss": 0.0548,
+      "step": 8869
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021687974341308002,
+      "loss": 0.6688,
+      "step": 8870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002168627523817418,
+      "loss": 0.0711,
+      "step": 8871
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002168457602796969,
+      "loss": 0.5319,
+      "step": 8872
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021682876710721753,
+      "loss": 0.62,
+      "step": 8873
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021681177286457572,
+      "loss": 3.2071,
+      "step": 8874
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021679477755204364,
+      "loss": 0.2888,
+      "step": 8875
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021677778116989345,
+      "loss": 0.037,
+      "step": 8876
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002167607837183973,
+      "loss": 0.9983,
+      "step": 8877
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021674378519782745,
+      "loss": 0.3645,
+      "step": 8878
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021672678560845596,
+      "loss": 0.2312,
+      "step": 8879
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021670978495055517,
+      "loss": 0.0016,
+      "step": 8880
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002166927832243973,
+      "loss": 0.8856,
+      "step": 8881
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021667578043025462,
+      "loss": 0.7018,
+      "step": 8882
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021665877656839937,
+      "loss": 0.4788,
+      "step": 8883
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021664177163910385,
+      "loss": 0.4066,
+      "step": 8884
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021662476564264038,
+      "loss": 0.0032,
+      "step": 8885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021660775857928127,
+      "loss": 1.0489,
+      "step": 8886
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021659075044929888,
+      "loss": 0.693,
+      "step": 8887
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021657374125296554,
+      "loss": 0.6306,
+      "step": 8888
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021655673099055364,
+      "loss": 0.5645,
+      "step": 8889
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021653971966233563,
+      "loss": 0.027,
+      "step": 8890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021652270726858384,
+      "loss": 0.0703,
+      "step": 8891
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021650569380957075,
+      "loss": 0.0022,
+      "step": 8892
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021648867928556874,
+      "loss": 0.3016,
+      "step": 8893
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002164716636968503,
+      "loss": 1.2921,
+      "step": 8894
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000216454647043688,
+      "loss": 0.362,
+      "step": 8895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021643762932635423,
+      "loss": 0.4816,
+      "step": 8896
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021642061054512153,
+      "loss": 0.8981,
+      "step": 8897
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021640359070026246,
+      "loss": 0.426,
+      "step": 8898
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002163865697920495,
+      "loss": 2.1494,
+      "step": 8899
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021636954782075529,
+      "loss": 1.7334,
+      "step": 8900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021635252478665237,
+      "loss": 0.7385,
+      "step": 8901
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021633550069001336,
+      "loss": 2.4922,
+      "step": 8902
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002163184755311108,
+      "loss": 0.2596,
+      "step": 8903
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002163014493102175,
+      "loss": 0.884,
+      "step": 8904
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002162844220276059,
+      "loss": 0.1098,
+      "step": 8905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021626739368354875,
+      "loss": 0.883,
+      "step": 8906
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021625036427831882,
+      "loss": 0.6871,
+      "step": 8907
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002162333338121887,
+      "loss": 0.815,
+      "step": 8908
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021621630228543112,
+      "loss": 0.2252,
+      "step": 8909
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021619926969831886,
+      "loss": 0.2396,
+      "step": 8910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021618223605112462,
+      "loss": 0.4883,
+      "step": 8911
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002161652013441212,
+      "loss": 2.4012,
+      "step": 8912
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002161481655775814,
+      "loss": 0.5144,
+      "step": 8913
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000216131128751778,
+      "loss": 1.0138,
+      "step": 8914
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021611409086698386,
+      "loss": 0.0075,
+      "step": 8915
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021609705192347174,
+      "loss": 0.3058,
+      "step": 8916
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021608001192151448,
+      "loss": 0.5885,
+      "step": 8917
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021606297086138507,
+      "loss": 0.3098,
+      "step": 8918
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021604592874335628,
+      "loss": 1.9968,
+      "step": 8919
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021602888556770112,
+      "loss": 0.3168,
+      "step": 8920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021601184133469243,
+      "loss": 0.035,
+      "step": 8921
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021599479604460313,
+      "loss": 0.4557,
+      "step": 8922
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021597774969770627,
+      "loss": 0.8574,
+      "step": 8923
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021596070229427475,
+      "loss": 0.8847,
+      "step": 8924
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021594365383458154,
+      "loss": 2.4142,
+      "step": 8925
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002159266043188997,
+      "loss": 0.0624,
+      "step": 8926
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002159095537475023,
+      "loss": 0.4962,
+      "step": 8927
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021589250212066221,
+      "loss": 0.8207,
+      "step": 8928
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021587544943865264,
+      "loss": 0.7511,
+      "step": 8929
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021585839570174666,
+      "loss": 1.1817,
+      "step": 8930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021584134091021725,
+      "loss": 0.8309,
+      "step": 8931
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021582428506433754,
+      "loss": 1.3637,
+      "step": 8932
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002158072281643808,
+      "loss": 0.4771,
+      "step": 8933
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021579017021062,
+      "loss": 0.4775,
+      "step": 8934
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021577311120332836,
+      "loss": 0.3964,
+      "step": 8935
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021575605114277908,
+      "loss": 0.5249,
+      "step": 8936
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002157389900292453,
+      "loss": 1.3635,
+      "step": 8937
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002157219278630003,
+      "loss": 0.3457,
+      "step": 8938
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021570486464431724,
+      "loss": 0.0032,
+      "step": 8939
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021568780037346935,
+      "loss": 0.3031,
+      "step": 8940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021567073505072995,
+      "loss": 0.8308,
+      "step": 8941
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002156536686763723,
+      "loss": 0.4194,
+      "step": 8942
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002156366012506696,
+      "loss": 0.3766,
+      "step": 8943
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021561953277389535,
+      "loss": 0.0022,
+      "step": 8944
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002156024632463227,
+      "loss": 0.8364,
+      "step": 8945
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021558539266822505,
+      "loss": 0.0265,
+      "step": 8946
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002155683210398758,
+      "loss": 0.2467,
+      "step": 8947
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021555124836154826,
+      "loss": 0.3188,
+      "step": 8948
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021553417463351587,
+      "loss": 0.1051,
+      "step": 8949
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021551709985605203,
+      "loss": 3.7021,
+      "step": 8950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021550002402943015,
+      "loss": 0.0177,
+      "step": 8951
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002154829471539237,
+      "loss": 0.6594,
+      "step": 8952
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021546586922980613,
+      "loss": 1.9882,
+      "step": 8953
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002154487902573509,
+      "loss": 0.5961,
+      "step": 8954
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002154317102368315,
+      "loss": 0.7456,
+      "step": 8955
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021541462916852154,
+      "loss": 0.9529,
+      "step": 8956
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002153975470526944,
+      "loss": 0.2378,
+      "step": 8957
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002153804638896237,
+      "loss": 0.5849,
+      "step": 8958
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000215363379679583,
+      "loss": 0.5953,
+      "step": 8959
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021534629442284585,
+      "loss": 0.4211,
+      "step": 8960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021532920811968593,
+      "loss": 0.4184,
+      "step": 8961
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021531212077037672,
+      "loss": 0.3962,
+      "step": 8962
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021529503237519194,
+      "loss": 0.671,
+      "step": 8963
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002152779429344052,
+      "loss": 1.4483,
+      "step": 8964
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002152608524482902,
+      "loss": 0.4339,
+      "step": 8965
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021524376091712054,
+      "loss": 2.2735,
+      "step": 8966
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021522666834117003,
+      "loss": 0.5837,
+      "step": 8967
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002152095747207123,
+      "loss": 0.3568,
+      "step": 8968
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021519248005602107,
+      "loss": 0.0032,
+      "step": 8969
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021517538434737016,
+      "loss": 0.2561,
+      "step": 8970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021515828759503325,
+      "loss": 0.3216,
+      "step": 8971
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021514118979928414,
+      "loss": 0.4556,
+      "step": 8972
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021512409096039665,
+      "loss": 0.8539,
+      "step": 8973
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002151069910786446,
+      "loss": 0.0055,
+      "step": 8974
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021508989015430176,
+      "loss": 1.046,
+      "step": 8975
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021507278818764207,
+      "loss": 0.2066,
+      "step": 8976
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021505568517893927,
+      "loss": 0.435,
+      "step": 8977
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021503858112846734,
+      "loss": 0.8406,
+      "step": 8978
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021502147603650018,
+      "loss": 0.4208,
+      "step": 8979
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021500436990331163,
+      "loss": 0.6046,
+      "step": 8980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002149872627291756,
+      "loss": 0.1415,
+      "step": 8981
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021497015451436618,
+      "loss": 0.4458,
+      "step": 8982
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002149530452591572,
+      "loss": 0.3192,
+      "step": 8983
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002149359349638227,
+      "loss": 0.4329,
+      "step": 8984
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021491882362863665,
+      "loss": 0.3902,
+      "step": 8985
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021490171125387305,
+      "loss": 0.1105,
+      "step": 8986
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000214884597839806,
+      "loss": 0.4591,
+      "step": 8987
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021486748338670945,
+      "loss": 0.9493,
+      "step": 8988
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021485036789485747,
+      "loss": 0.2979,
+      "step": 8989
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021483325136452421,
+      "loss": 0.8238,
+      "step": 8990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021481613379598376,
+      "loss": 0.0041,
+      "step": 8991
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021479901518951016,
+      "loss": 0.5041,
+      "step": 8992
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021478189554537764,
+      "loss": 0.8948,
+      "step": 8993
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021476477486386024,
+      "loss": 0.2814,
+      "step": 8994
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002147476531452322,
+      "loss": 0.0033,
+      "step": 8995
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021473053038976761,
+      "loss": 0.3757,
+      "step": 8996
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021471340659774075,
+      "loss": 0.8917,
+      "step": 8997
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002146962817694258,
+      "loss": 0.9264,
+      "step": 8998
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000214679155905097,
+      "loss": 0.1641,
+      "step": 8999
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021466202900502862,
+      "loss": 0.3644,
+      "step": 9000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021464490106949484,
+      "loss": 1.1357,
+      "step": 9001
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021462777209877007,
+      "loss": 0.3453,
+      "step": 9002
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021461064209312843,
+      "loss": 0.6496,
+      "step": 9003
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002145935110528443,
+      "loss": 0.9006,
+      "step": 9004
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021457637897819213,
+      "loss": 0.0896,
+      "step": 9005
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021455924586944613,
+      "loss": 0.333,
+      "step": 9006
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021454211172688067,
+      "loss": 0.0968,
+      "step": 9007
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021452497655077022,
+      "loss": 0.0003,
+      "step": 9008
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021450784034138907,
+      "loss": 0.7602,
+      "step": 9009
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021449070309901162,
+      "loss": 0.0616,
+      "step": 9010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002144735648239124,
+      "loss": 0.2598,
+      "step": 9011
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002144564255163658,
+      "loss": 0.0002,
+      "step": 9012
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021443928517664627,
+      "loss": 0.5736,
+      "step": 9013
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021442214380502832,
+      "loss": 0.0066,
+      "step": 9014
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021440500140178638,
+      "loss": 1.4947,
+      "step": 9015
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021438785796719503,
+      "loss": 0.3512,
+      "step": 9016
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021437071350152877,
+      "loss": 0.3533,
+      "step": 9017
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021435356800506212,
+      "loss": 0.057,
+      "step": 9018
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021433642147806968,
+      "loss": 0.2083,
+      "step": 9019
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021431927392082597,
+      "loss": 0.0092,
+      "step": 9020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021430212533360567,
+      "loss": 0.1803,
+      "step": 9021
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002142849757166833,
+      "loss": 0.8858,
+      "step": 9022
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021426782507033353,
+      "loss": 0.659,
+      "step": 9023
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000214250673394831,
+      "loss": 0.6197,
+      "step": 9024
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021423352069045037,
+      "loss": 1.8838,
+      "step": 9025
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002142163669574663,
+      "loss": 0.8262,
+      "step": 9026
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021419921219615346,
+      "loss": 0.4514,
+      "step": 9027
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021418205640678662,
+      "loss": 0.7931,
+      "step": 9028
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021416489958964047,
+      "loss": 1.2194,
+      "step": 9029
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002141477417449897,
+      "loss": 0.6198,
+      "step": 9030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002141305828731092,
+      "loss": 0.2123,
+      "step": 9031
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002141134229742736,
+      "loss": 0.0029,
+      "step": 9032
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002140962620487578,
+      "loss": 1.4451,
+      "step": 9033
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021407910009683652,
+      "loss": 0.5844,
+      "step": 9034
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021406193711878463,
+      "loss": 0.4032,
+      "step": 9035
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021404477311487695,
+      "loss": 0.9313,
+      "step": 9036
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002140276080853884,
+      "loss": 0.2678,
+      "step": 9037
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021401044203059373,
+      "loss": 0.9572,
+      "step": 9038
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021399327495076793,
+      "loss": 1.2918,
+      "step": 9039
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002139761068461859,
+      "loss": 1.9935,
+      "step": 9040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021395893771712247,
+      "loss": 0.4647,
+      "step": 9041
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021394176756385266,
+      "loss": 1.1199,
+      "step": 9042
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021392459638665146,
+      "loss": 2.4825,
+      "step": 9043
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021390742418579375,
+      "loss": 0.3454,
+      "step": 9044
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021389025096155454,
+      "loss": 0.697,
+      "step": 9045
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021387307671420886,
+      "loss": 0.1293,
+      "step": 9046
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002138559014440317,
+      "loss": 0.376,
+      "step": 9047
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002138387251512981,
+      "loss": 0.5472,
+      "step": 9048
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021382154783628314,
+      "loss": 0.7305,
+      "step": 9049
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002138043694992619,
+      "loss": 0.6377,
+      "step": 9050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021378719014050943,
+      "loss": 2.4215,
+      "step": 9051
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021377000976030087,
+      "loss": 0.6503,
+      "step": 9052
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021375282835891125,
+      "loss": 0.2292,
+      "step": 9053
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021373564593661577,
+      "loss": 0.4012,
+      "step": 9054
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002137184624936896,
+      "loss": 0.4121,
+      "step": 9055
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021370127803040784,
+      "loss": 0.3092,
+      "step": 9056
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021368409254704577,
+      "loss": 0.3129,
+      "step": 9057
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021366690604387854,
+      "loss": 0.856,
+      "step": 9058
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002136497185211813,
+      "loss": 0.0059,
+      "step": 9059
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021363252997922937,
+      "loss": 1.2601,
+      "step": 9060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021361534041829794,
+      "loss": 0.1896,
+      "step": 9061
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021359814983866238,
+      "loss": 0.3544,
+      "step": 9062
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021358095824059787,
+      "loss": 0.6087,
+      "step": 9063
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021356376562437968,
+      "loss": 0.441,
+      "step": 9064
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021354657199028327,
+      "loss": 0.559,
+      "step": 9065
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002135293773385838,
+      "loss": 0.889,
+      "step": 9066
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021351218166955677,
+      "loss": 0.0031,
+      "step": 9067
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021349498498347743,
+      "loss": 0.2293,
+      "step": 9068
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002134777872806212,
+      "loss": 0.7364,
+      "step": 9069
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021346058856126345,
+      "loss": 0.9582,
+      "step": 9070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021344338882567962,
+      "loss": 0.2162,
+      "step": 9071
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021342618807414516,
+      "loss": 1.2114,
+      "step": 9072
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021340898630693546,
+      "loss": 1.027,
+      "step": 9073
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021339178352432602,
+      "loss": 0.0396,
+      "step": 9074
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021337457972659231,
+      "loss": 3.4205,
+      "step": 9075
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021335737491400981,
+      "loss": 0.7665,
+      "step": 9076
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002133401690868541,
+      "loss": 0.3617,
+      "step": 9077
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021332296224540058,
+      "loss": 0.5602,
+      "step": 9078
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021330575438992483,
+      "loss": 0.4802,
+      "step": 9079
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021328854552070246,
+      "loss": 0.1281,
+      "step": 9080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021327133563800904,
+      "loss": 0.8357,
+      "step": 9081
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002132541247421201,
+      "loss": 2.4047,
+      "step": 9082
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002132369128333113,
+      "loss": 0.3213,
+      "step": 9083
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021321969991185824,
+      "loss": 0.5001,
+      "step": 9084
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021320248597803658,
+      "loss": 0.7632,
+      "step": 9085
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021318527103212197,
+      "loss": 0.5498,
+      "step": 9086
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021316805507439002,
+      "loss": 0.561,
+      "step": 9087
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002131508381051165,
+      "loss": 0.4074,
+      "step": 9088
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021313362012457713,
+      "loss": 0.9874,
+      "step": 9089
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002131164011330475,
+      "loss": 0.2726,
+      "step": 9090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021309918113080343,
+      "loss": 0.4103,
+      "step": 9091
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021308196011812078,
+      "loss": 0.0124,
+      "step": 9092
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021306473809527514,
+      "loss": 0.6493,
+      "step": 9093
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021304751506254238,
+      "loss": 0.3836,
+      "step": 9094
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021303029102019827,
+      "loss": 0.3147,
+      "step": 9095
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021301306596851864,
+      "loss": 1.0122,
+      "step": 9096
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021299583990777934,
+      "loss": 0.0163,
+      "step": 9097
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002129786128382562,
+      "loss": 0.781,
+      "step": 9098
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002129613847602251,
+      "loss": 0.3264,
+      "step": 9099
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021294415567396193,
+      "loss": 0.0902,
+      "step": 9100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021292692557974257,
+      "loss": 0.7039,
+      "step": 9101
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002129096944778429,
+      "loss": 0.3091,
+      "step": 9102
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021289246236853894,
+      "loss": 0.3412,
+      "step": 9103
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021287522925210654,
+      "loss": 0.3556,
+      "step": 9104
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002128579951288217,
+      "loss": 2.1956,
+      "step": 9105
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021284075999896044,
+      "loss": 0.4556,
+      "step": 9106
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002128235238627987,
+      "loss": 0.2194,
+      "step": 9107
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021280628672061254,
+      "loss": 0.0016,
+      "step": 9108
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021278904857267793,
+      "loss": 0.0099,
+      "step": 9109
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021277180941927095,
+      "loss": 0.3515,
+      "step": 9110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021275456926066763,
+      "loss": 0.0306,
+      "step": 9111
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002127373280971441,
+      "loss": 0.2177,
+      "step": 9112
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021272008592897637,
+      "loss": 0.9241,
+      "step": 9113
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002127028427564406,
+      "loss": 1.3506,
+      "step": 9114
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000212685598579813,
+      "loss": 1.2843,
+      "step": 9115
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021266835339936953,
+      "loss": 0.2953,
+      "step": 9116
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002126511072153864,
+      "loss": 1.213,
+      "step": 9117
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021263386002813987,
+      "loss": 0.2124,
+      "step": 9118
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021261661183790605,
+      "loss": 0.1859,
+      "step": 9119
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002125993626449612,
+      "loss": 0.0065,
+      "step": 9120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021258211244958147,
+      "loss": 0.8204,
+      "step": 9121
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021256486125204317,
+      "loss": 0.3842,
+      "step": 9122
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021254760905262246,
+      "loss": 0.6784,
+      "step": 9123
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021253035585159567,
+      "loss": 0.0071,
+      "step": 9124
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021251310164923913,
+      "loss": 0.0027,
+      "step": 9125
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021249584644582903,
+      "loss": 1.0836,
+      "step": 9126
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002124785902416418,
+      "loss": 0.3084,
+      "step": 9127
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021246133303695364,
+      "loss": 0.3485,
+      "step": 9128
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000212444074832041,
+      "loss": 0.0027,
+      "step": 9129
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021242681562718026,
+      "loss": 0.0044,
+      "step": 9130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021240955542264773,
+      "loss": 0.2899,
+      "step": 9131
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021239229421871983,
+      "loss": 0.0096,
+      "step": 9132
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000212375032015673,
+      "loss": 0.7339,
+      "step": 9133
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002123577688137836,
+      "loss": 0.0021,
+      "step": 9134
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021234050461332816,
+      "loss": 0.3097,
+      "step": 9135
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021232323941458301,
+      "loss": 0.7287,
+      "step": 9136
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002123059732178248,
+      "loss": 0.0077,
+      "step": 9137
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021228870602332998,
+      "loss": 0.011,
+      "step": 9138
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021227143783137488,
+      "loss": 0.8697,
+      "step": 9139
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021225416864223626,
+      "loss": 0.2438,
+      "step": 9140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021223689845619053,
+      "loss": 0.4059,
+      "step": 9141
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021221962727351428,
+      "loss": 0.0131,
+      "step": 9142
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021220235509448405,
+      "loss": 0.5369,
+      "step": 9143
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021218508191937652,
+      "loss": 0.8985,
+      "step": 9144
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021216780774846819,
+      "loss": 0.0004,
+      "step": 9145
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021215053258203572,
+      "loss": 0.6833,
+      "step": 9146
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021213325642035574,
+      "loss": 0.0549,
+      "step": 9147
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002121159792637049,
+      "loss": 0.0028,
+      "step": 9148
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002120987011123599,
+      "loss": 0.614,
+      "step": 9149
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002120814219665974,
+      "loss": 0.4839,
+      "step": 9150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021206414182669406,
+      "loss": 0.5064,
+      "step": 9151
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021204686069292665,
+      "loss": 0.5543,
+      "step": 9152
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021202957856557192,
+      "loss": 0.8864,
+      "step": 9153
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021201229544490652,
+      "loss": 0.6914,
+      "step": 9154
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021199501133120735,
+      "loss": 0.0007,
+      "step": 9155
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021197772622475106,
+      "loss": 0.5917,
+      "step": 9156
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021196044012581453,
+      "loss": 0.2317,
+      "step": 9157
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021194315303467453,
+      "loss": 0.2549,
+      "step": 9158
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021192586495160788,
+      "loss": 0.2528,
+      "step": 9159
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002119085758768915,
+      "loss": 1.14,
+      "step": 9160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021189128581080216,
+      "loss": 0.6738,
+      "step": 9161
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021187399475361673,
+      "loss": 0.9824,
+      "step": 9162
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021185670270561214,
+      "loss": 0.6082,
+      "step": 9163
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021183940966706534,
+      "loss": 0.7769,
+      "step": 9164
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021182211563825314,
+      "loss": 0.5359,
+      "step": 9165
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021180482061945255,
+      "loss": 1.2738,
+      "step": 9166
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021178752461094058,
+      "loss": 0.509,
+      "step": 9167
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021177022761299406,
+      "loss": 1.0427,
+      "step": 9168
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021175292962589005,
+      "loss": 0.4413,
+      "step": 9169
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021173563064990557,
+      "loss": 0.0852,
+      "step": 9170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021171833068531762,
+      "loss": 0.1263,
+      "step": 9171
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021170102973240326,
+      "loss": 0.8219,
+      "step": 9172
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021168372779143945,
+      "loss": 0.137,
+      "step": 9173
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021166642486270334,
+      "loss": 0.2536,
+      "step": 9174
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021164912094647203,
+      "loss": 0.0221,
+      "step": 9175
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002116318160430225,
+      "loss": 0.358,
+      "step": 9176
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021161451015263197,
+      "loss": 0.7368,
+      "step": 9177
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021159720327557755,
+      "loss": 0.3421,
+      "step": 9178
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021157989541213632,
+      "loss": 0.2476,
+      "step": 9179
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021156258656258552,
+      "loss": 0.381,
+      "step": 9180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021154527672720226,
+      "loss": 0.5243,
+      "step": 9181
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021152796590626375,
+      "loss": 0.0013,
+      "step": 9182
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021151065410004722,
+      "loss": 0.4645,
+      "step": 9183
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002114933413088299,
+      "loss": 0.0006,
+      "step": 9184
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021147602753288894,
+      "loss": 0.0702,
+      "step": 9185
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021145871277250175,
+      "loss": 0.2238,
+      "step": 9186
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021144139702794547,
+      "loss": 0.6434,
+      "step": 9187
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021142408029949743,
+      "loss": 0.2152,
+      "step": 9188
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002114067625874349,
+      "loss": 0.9778,
+      "step": 9189
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002113894438920353,
+      "loss": 0.5711,
+      "step": 9190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021137212421357585,
+      "loss": 0.0516,
+      "step": 9191
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021135480355233388,
+      "loss": 0.4661,
+      "step": 9192
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002113374819085869,
+      "loss": 0.7426,
+      "step": 9193
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021132015928261217,
+      "loss": 0.7083,
+      "step": 9194
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002113028356746871,
+      "loss": 0.3507,
+      "step": 9195
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021128551108508912,
+      "loss": 0.4457,
+      "step": 9196
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021126818551409567,
+      "loss": 0.0008,
+      "step": 9197
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021125085896198418,
+      "loss": 0.8676,
+      "step": 9198
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021123353142903214,
+      "loss": 0.5687,
+      "step": 9199
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021121620291551693,
+      "loss": 0.0066,
+      "step": 9200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021119887342171612,
+      "loss": 0.002,
+      "step": 9201
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021118154294790724,
+      "loss": 0.4833,
+      "step": 9202
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002111642114943677,
+      "loss": 0.416,
+      "step": 9203
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021114687906137518,
+      "loss": 2.0938,
+      "step": 9204
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002111295456492071,
+      "loss": 0.0769,
+      "step": 9205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021111221125814117,
+      "loss": 1.4835,
+      "step": 9206
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021109487588845483,
+      "loss": 0.1313,
+      "step": 9207
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021107753954042576,
+      "loss": 1.6448,
+      "step": 9208
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021106020221433152,
+      "loss": 0.4019,
+      "step": 9209
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021104286391044984,
+      "loss": 0.4035,
+      "step": 9210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002110255246290583,
+      "loss": 0.3521,
+      "step": 9211
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021100818437043452,
+      "loss": 0.3545,
+      "step": 9212
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002109908431348563,
+      "loss": 0.245,
+      "step": 9213
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002109735009226012,
+      "loss": 0.4867,
+      "step": 9214
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021095615773394696,
+      "loss": 0.6104,
+      "step": 9215
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021093881356917146,
+      "loss": 0.4958,
+      "step": 9216
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021092146842855222,
+      "loss": 0.1723,
+      "step": 9217
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021090412231236714,
+      "loss": 0.6923,
+      "step": 9218
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021088677522089388,
+      "loss": 0.7958,
+      "step": 9219
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021086942715441034,
+      "loss": 0.3909,
+      "step": 9220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021085207811319428,
+      "loss": 0.3049,
+      "step": 9221
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002108347280975235,
+      "loss": 0.3806,
+      "step": 9222
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021081737710767587,
+      "loss": 0.3332,
+      "step": 9223
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021080002514392923,
+      "loss": 0.6267,
+      "step": 9224
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002107826722065614,
+      "loss": 0.0049,
+      "step": 9225
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021076531829585028,
+      "loss": 0.308,
+      "step": 9226
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002107479634120738,
+      "loss": 1.4031,
+      "step": 9227
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021073060755550993,
+      "loss": 0.7058,
+      "step": 9228
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021071325072643645,
+      "loss": 0.002,
+      "step": 9229
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002106958929251314,
+      "loss": 0.2248,
+      "step": 9230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021067853415187267,
+      "loss": 0.5887,
+      "step": 9231
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021066117440693834,
+      "loss": 0.5663,
+      "step": 9232
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021064381369060632,
+      "loss": 0.5073,
+      "step": 9233
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021062645200315466,
+      "loss": 0.6783,
+      "step": 9234
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002106090893448613,
+      "loss": 1.4884,
+      "step": 9235
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021059172571600437,
+      "loss": 0.8443,
+      "step": 9236
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021057436111686188,
+      "loss": 0.1837,
+      "step": 9237
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002105569955477119,
+      "loss": 0.9456,
+      "step": 9238
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021053962900883258,
+      "loss": 0.1517,
+      "step": 9239
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002105222615005019,
+      "loss": 0.306,
+      "step": 9240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021050489302299803,
+      "loss": 1.2798,
+      "step": 9241
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021048752357659913,
+      "loss": 0.5211,
+      "step": 9242
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021047015316158333,
+      "loss": 0.114,
+      "step": 9243
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021045278177822874,
+      "loss": 0.5166,
+      "step": 9244
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002104354094268136,
+      "loss": 0.3508,
+      "step": 9245
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021041803610761606,
+      "loss": 0.0027,
+      "step": 9246
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002104006618209144,
+      "loss": 0.5069,
+      "step": 9247
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021038328656698674,
+      "loss": 0.3698,
+      "step": 9248
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021036591034611136,
+      "loss": 0.2176,
+      "step": 9249
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021034853315856654,
+      "loss": 0.6715,
+      "step": 9250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021033115500463057,
+      "loss": 0.3047,
+      "step": 9251
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021031377588458163,
+      "loss": 0.2508,
+      "step": 9252
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021029639579869813,
+      "loss": 0.26,
+      "step": 9253
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021027901474725836,
+      "loss": 0.7236,
+      "step": 9254
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021026163273054057,
+      "loss": 0.5549,
+      "step": 9255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002102442497488232,
+      "loss": 0.3592,
+      "step": 9256
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002102268658023846,
+      "loss": 0.876,
+      "step": 9257
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021020948089150312,
+      "loss": 0.6061,
+      "step": 9258
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002101920950164572,
+      "loss": 0.5437,
+      "step": 9259
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021017470817752518,
+      "loss": 0.0024,
+      "step": 9260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002101573203749855,
+      "loss": 0.0006,
+      "step": 9261
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021013993160911667,
+      "loss": 0.5604,
+      "step": 9262
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021012254188019708,
+      "loss": 1.419,
+      "step": 9263
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021010515118850514,
+      "loss": 0.9877,
+      "step": 9264
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021008775953431952,
+      "loss": 0.0007,
+      "step": 9265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021007036691791855,
+      "loss": 0.2373,
+      "step": 9266
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021005297333958077,
+      "loss": 0.1784,
+      "step": 9267
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002100355787995848,
+      "loss": 0.202,
+      "step": 9268
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021001818329820912,
+      "loss": 0.0004,
+      "step": 9269
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021000078683573235,
+      "loss": 2.3468,
+      "step": 9270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020998338941243295,
+      "loss": 0.1374,
+      "step": 9271
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020996599102858968,
+      "loss": 0.0022,
+      "step": 9272
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020994859168448098,
+      "loss": 0.2586,
+      "step": 9273
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002099311913803856,
+      "loss": 0.7401,
+      "step": 9274
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020991379011658213,
+      "loss": 0.2294,
+      "step": 9275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002098963878933492,
+      "loss": 0.2065,
+      "step": 9276
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020987898471096556,
+      "loss": 1.0898,
+      "step": 9277
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020986158056970984,
+      "loss": 0.5324,
+      "step": 9278
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020984417546986067,
+      "loss": 0.5918,
+      "step": 9279
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002098267694116969,
+      "loss": 0.5467,
+      "step": 9280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020980936239549722,
+      "loss": 0.2269,
+      "step": 9281
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020979195442154033,
+      "loss": 0.3135,
+      "step": 9282
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020977454549010507,
+      "loss": 0.5431,
+      "step": 9283
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020975713560147013,
+      "loss": 0.0003,
+      "step": 9284
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020973972475591435,
+      "loss": 0.2228,
+      "step": 9285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020972231295371657,
+      "loss": 0.3433,
+      "step": 9286
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020970490019515554,
+      "loss": 0.22,
+      "step": 9287
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020968748648051019,
+      "loss": 0.1715,
+      "step": 9288
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020967007181005927,
+      "loss": 0.4868,
+      "step": 9289
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020965265618408168,
+      "loss": 0.5396,
+      "step": 9290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002096352396028564,
+      "loss": 0.1267,
+      "step": 9291
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002096178220666622,
+      "loss": 0.2261,
+      "step": 9292
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020960040357577807,
+      "loss": 0.5811,
+      "step": 9293
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020958298413048293,
+      "loss": 0.6331,
+      "step": 9294
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020956556373105572,
+      "loss": 0.0087,
+      "step": 9295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020954814237777542,
+      "loss": 0.013,
+      "step": 9296
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020953072007092098,
+      "loss": 0.0005,
+      "step": 9297
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020951329681077137,
+      "loss": 0.5883,
+      "step": 9298
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020949587259760564,
+      "loss": 0.0345,
+      "step": 9299
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020947844743170283,
+      "loss": 0.2885,
+      "step": 9300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020946102131334194,
+      "loss": 0.0005,
+      "step": 9301
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020944359424280199,
+      "loss": 0.2031,
+      "step": 9302
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020942616622036215,
+      "loss": 0.0865,
+      "step": 9303
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020940873724630143,
+      "loss": 0.6815,
+      "step": 9304
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020939130732089891,
+      "loss": 0.4656,
+      "step": 9305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020937387644443374,
+      "loss": 0.4232,
+      "step": 9306
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002093564446171851,
+      "loss": 1.0914,
+      "step": 9307
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020933901183943203,
+      "loss": 0.0103,
+      "step": 9308
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020932157811145375,
+      "loss": 0.5171,
+      "step": 9309
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020930414343352943,
+      "loss": 1.8729,
+      "step": 9310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020928670780593825,
+      "loss": 0.2456,
+      "step": 9311
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020926927122895948,
+      "loss": 0.0006,
+      "step": 9312
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002092518337028722,
+      "loss": 0.0041,
+      "step": 9313
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002092343952279558,
+      "loss": 0.99,
+      "step": 9314
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020921695580448937,
+      "loss": 0.0025,
+      "step": 9315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020919951543275233,
+      "loss": 1.0552,
+      "step": 9316
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020918207411302385,
+      "loss": 0.3274,
+      "step": 9317
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020916463184558333,
+      "loss": 0.0003,
+      "step": 9318
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020914718863071,
+      "loss": 0.3032,
+      "step": 9319
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002091297444686832,
+      "loss": 0.4354,
+      "step": 9320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020911229935978228,
+      "loss": 0.1673,
+      "step": 9321
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020909485330428663,
+      "loss": 1.5785,
+      "step": 9322
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020907740630247558,
+      "loss": 1.1118,
+      "step": 9323
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020905995835462853,
+      "loss": 0.0003,
+      "step": 9324
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020904250946102485,
+      "loss": 0.5645,
+      "step": 9325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020902505962194404,
+      "loss": 0.3682,
+      "step": 9326
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020900760883766546,
+      "loss": 0.678,
+      "step": 9327
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020899015710846859,
+      "loss": 0.1062,
+      "step": 9328
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002089727044346329,
+      "loss": 0.5255,
+      "step": 9329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020895525081643783,
+      "loss": 2.8633,
+      "step": 9330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020893779625416291,
+      "loss": 0.0024,
+      "step": 9331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020892034074808762,
+      "loss": 0.3125,
+      "step": 9332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020890288429849154,
+      "loss": 0.3525,
+      "step": 9333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020888542690565413,
+      "loss": 1.2269,
+      "step": 9334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000208867968569855,
+      "loss": 0.3844,
+      "step": 9335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020885050929137368,
+      "loss": 0.2174,
+      "step": 9336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002088330490704898,
+      "loss": 0.8116,
+      "step": 9337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020881558790748292,
+      "loss": 1.2049,
+      "step": 9338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020879812580263266,
+      "loss": 1.0383,
+      "step": 9339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002087806627562187,
+      "loss": 0.9491,
+      "step": 9340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020876319876852062,
+      "loss": 0.226,
+      "step": 9341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020874573383981813,
+      "loss": 0.2743,
+      "step": 9342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020872826797039082,
+      "loss": 0.5968,
+      "step": 9343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020871080116051846,
+      "loss": 0.0274,
+      "step": 9344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020869333341048076,
+      "loss": 0.419,
+      "step": 9345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020867586472055738,
+      "loss": 0.0003,
+      "step": 9346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020865839509102806,
+      "loss": 0.7989,
+      "step": 9347
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020864092452217264,
+      "loss": 1.9853,
+      "step": 9348
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002086234530142708,
+      "loss": 0.0773,
+      "step": 9349
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020860598056760235,
+      "loss": 0.9304,
+      "step": 9350
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020858850718244703,
+      "loss": 0.1812,
+      "step": 9351
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020857103285908476,
+      "loss": 0.7201,
+      "step": 9352
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002085535575977953,
+      "loss": 0.7057,
+      "step": 9353
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020853608139885838,
+      "loss": 0.2738,
+      "step": 9354
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020851860426255412,
+      "loss": 0.4395,
+      "step": 9355
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020850112618916213,
+      "loss": 0.9214,
+      "step": 9356
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020848364717896244,
+      "loss": 0.1088,
+      "step": 9357
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002084661672322349,
+      "loss": 0.6383,
+      "step": 9358
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020844868634925946,
+      "loss": 0.2791,
+      "step": 9359
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000208431204530316,
+      "loss": 0.0065,
+      "step": 9360
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020841372177568456,
+      "loss": 0.4655,
+      "step": 9361
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002083962380856449,
+      "loss": 0.0859,
+      "step": 9362
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020837875346047725,
+      "loss": 0.3594,
+      "step": 9363
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020836126790046142,
+      "loss": 0.4906,
+      "step": 9364
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002083437814058775,
+      "loss": 2.5137,
+      "step": 9365
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020832629397700547,
+      "loss": 0.2823,
+      "step": 9366
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002083088056141254,
+      "loss": 0.5471,
+      "step": 9367
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020829131631751728,
+      "loss": 0.74,
+      "step": 9368
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020827382608746122,
+      "loss": 0.2772,
+      "step": 9369
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002082563349242373,
+      "loss": 0.3198,
+      "step": 9370
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020823884282812562,
+      "loss": 0.5536,
+      "step": 9371
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002082213497994063,
+      "loss": 0.4709,
+      "step": 9372
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020820385583835934,
+      "loss": 0.5406,
+      "step": 9373
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020818636094526504,
+      "loss": 0.5871,
+      "step": 9374
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020816886512040354,
+      "loss": 1.0389,
+      "step": 9375
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002081513683640549,
+      "loss": 0.3455,
+      "step": 9376
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020813387067649937,
+      "loss": 0.0047,
+      "step": 9377
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002081163720580172,
+      "loss": 0.703,
+      "step": 9378
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002080988725088885,
+      "loss": 0.5464,
+      "step": 9379
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020808137202939358,
+      "loss": 0.4074,
+      "step": 9380
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020806387061981266,
+      "loss": 1.2854,
+      "step": 9381
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000208046368280426,
+      "loss": 0.3573,
+      "step": 9382
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020802886501151383,
+      "loss": 0.5062,
+      "step": 9383
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020801136081335649,
+      "loss": 0.4281,
+      "step": 9384
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020799385568623428,
+      "loss": 0.7632,
+      "step": 9385
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020797634963042746,
+      "loss": 0.2285,
+      "step": 9386
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002079588426462165,
+      "loss": 0.5334,
+      "step": 9387
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020794133473388157,
+      "loss": 0.4968,
+      "step": 9388
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020792382589370317,
+      "loss": 0.0004,
+      "step": 9389
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020790631612596164,
+      "loss": 0.0013,
+      "step": 9390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020788880543093737,
+      "loss": 0.3573,
+      "step": 9391
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002078712938089107,
+      "loss": 0.1322,
+      "step": 9392
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020785378126016214,
+      "loss": 0.2358,
+      "step": 9393
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002078362677849721,
+      "loss": 0.0009,
+      "step": 9394
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020781875338362108,
+      "loss": 2.1485,
+      "step": 9395
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020780123805638946,
+      "loss": 0.0014,
+      "step": 9396
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020778372180355775,
+      "loss": 0.1749,
+      "step": 9397
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020776620462540648,
+      "loss": 0.0905,
+      "step": 9398
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002077486865222161,
+      "loss": 0.4094,
+      "step": 9399
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002077311674942672,
+      "loss": 0.8222,
+      "step": 9400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002077136475418403,
+      "loss": 0.2269,
+      "step": 9401
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020769612666521594,
+      "loss": 0.0042,
+      "step": 9402
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020767860486467466,
+      "loss": 0.0004,
+      "step": 9403
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020766108214049717,
+      "loss": 0.5762,
+      "step": 9404
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020764355849296393,
+      "loss": 0.2873,
+      "step": 9405
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020762603392235563,
+      "loss": 0.6332,
+      "step": 9406
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020760850842895288,
+      "loss": 0.3726,
+      "step": 9407
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020759098201303633,
+      "loss": 0.326,
+      "step": 9408
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020757345467488662,
+      "loss": 0.6414,
+      "step": 9409
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020755592641478449,
+      "loss": 2.4278,
+      "step": 9410
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002075383972330105,
+      "loss": 0.3233,
+      "step": 9411
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020752086712984543,
+      "loss": 1.0372,
+      "step": 9412
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002075033361055701,
+      "loss": 0.1715,
+      "step": 9413
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020748580416046506,
+      "loss": 0.0047,
+      "step": 9414
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020746827129481117,
+      "loss": 1.5037,
+      "step": 9415
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020745073750888914,
+      "loss": 0.1819,
+      "step": 9416
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002074332028029798,
+      "loss": 0.2751,
+      "step": 9417
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020741566717736388,
+      "loss": 0.6835,
+      "step": 9418
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020739813063232223,
+      "loss": 0.2816,
+      "step": 9419
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020738059316813566,
+      "loss": 0.5158,
+      "step": 9420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000207363054785085,
+      "loss": 0.3504,
+      "step": 9421
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020734551548345115,
+      "loss": 0.4541,
+      "step": 9422
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002073279752635149,
+      "loss": 0.0009,
+      "step": 9423
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020731043412555716,
+      "loss": 0.533,
+      "step": 9424
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020729289206985885,
+      "loss": 0.3113,
+      "step": 9425
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002072753490967008,
+      "loss": 0.6952,
+      "step": 9426
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020725780520636407,
+      "loss": 0.1473,
+      "step": 9427
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002072402603991295,
+      "loss": 0.0075,
+      "step": 9428
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020722271467527804,
+      "loss": 1.4014,
+      "step": 9429
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002072051680350907,
+      "loss": 1.1384,
+      "step": 9430
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020718762047884843,
+      "loss": 0.0591,
+      "step": 9431
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020717007200683226,
+      "loss": 1.6963,
+      "step": 9432
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020715252261932317,
+      "loss": 0.409,
+      "step": 9433
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020713497231660222,
+      "loss": 0.832,
+      "step": 9434
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020711742109895043,
+      "loss": 0.4683,
+      "step": 9435
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002070998689666489,
+      "loss": 3.3341,
+      "step": 9436
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002070823159199786,
+      "loss": 0.2854,
+      "step": 9437
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002070647619592207,
+      "loss": 0.0032,
+      "step": 9438
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020704720708465633,
+      "loss": 1.3369,
+      "step": 9439
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002070296512965665,
+      "loss": 0.3379,
+      "step": 9440
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020701209459523243,
+      "loss": 0.2734,
+      "step": 9441
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020699453698093517,
+      "loss": 0.4544,
+      "step": 9442
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000206976978453956,
+      "loss": 1.95,
+      "step": 9443
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000206959419014576,
+      "loss": 0.0136,
+      "step": 9444
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020694185866307636,
+      "loss": 0.0125,
+      "step": 9445
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020692429739973837,
+      "loss": 0.8176,
+      "step": 9446
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020690673522484317,
+      "loss": 0.1536,
+      "step": 9447
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020688917213867195,
+      "loss": 2.1677,
+      "step": 9448
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020687160814150608,
+      "loss": 0.2728,
+      "step": 9449
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002068540432336268,
+      "loss": 1.5293,
+      "step": 9450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020683647741531527,
+      "loss": 0.4409,
+      "step": 9451
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020681891068685283,
+      "loss": 1.1875,
+      "step": 9452
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020680134304852088,
+      "loss": 2.4179,
+      "step": 9453
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020678377450060066,
+      "loss": 0.0949,
+      "step": 9454
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020676620504337347,
+      "loss": 0.7979,
+      "step": 9455
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020674863467712073,
+      "loss": 0.0035,
+      "step": 9456
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020673106340212376,
+      "loss": 0.4891,
+      "step": 9457
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020671349121866397,
+      "loss": 0.2974,
+      "step": 9458
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020669591812702274,
+      "loss": 0.0012,
+      "step": 9459
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020667834412748144,
+      "loss": 0.0029,
+      "step": 9460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002066607692203215,
+      "loss": 0.3522,
+      "step": 9461
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020664319340582446,
+      "loss": 0.9249,
+      "step": 9462
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020662561668427162,
+      "loss": 0.2476,
+      "step": 9463
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020660803905594448,
+      "loss": 0.5954,
+      "step": 9464
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020659046052112464,
+      "loss": 0.0249,
+      "step": 9465
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020657288108009343,
+      "loss": 1.8975,
+      "step": 9466
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020655530073313252,
+      "loss": 0.3758,
+      "step": 9467
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020653771948052325,
+      "loss": 0.3039,
+      "step": 9468
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020652013732254732,
+      "loss": 0.0022,
+      "step": 9469
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020650255425948616,
+      "loss": 0.4241,
+      "step": 9470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020648497029162143,
+      "loss": 0.2885,
+      "step": 9471
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020646738541923467,
+      "loss": 1.0821,
+      "step": 9472
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002064497996426075,
+      "loss": 1.2149,
+      "step": 9473
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002064322129620215,
+      "loss": 0.248,
+      "step": 9474
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020641462537775824,
+      "loss": 0.0079,
+      "step": 9475
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020639703689009952,
+      "loss": 0.4876,
+      "step": 9476
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020637944749932682,
+      "loss": 0.4629,
+      "step": 9477
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020636185720572192,
+      "loss": 1.607,
+      "step": 9478
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020634426600956644,
+      "loss": 0.3027,
+      "step": 9479
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020632667391114212,
+      "loss": 0.4438,
+      "step": 9480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020630908091073062,
+      "loss": 0.9127,
+      "step": 9481
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020629148700861371,
+      "loss": 0.8299,
+      "step": 9482
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020627389220507312,
+      "loss": 0.9252,
+      "step": 9483
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002062562965003906,
+      "loss": 0.3658,
+      "step": 9484
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020623869989484797,
+      "loss": 0.3129,
+      "step": 9485
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002062211023887269,
+      "loss": 0.8663,
+      "step": 9486
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020620350398230927,
+      "loss": 0.3037,
+      "step": 9487
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002061859046758769,
+      "loss": 0.4116,
+      "step": 9488
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020616830446971155,
+      "loss": 0.0031,
+      "step": 9489
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002061507033640951,
+      "loss": 0.4472,
+      "step": 9490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020613310135930943,
+      "loss": 0.3487,
+      "step": 9491
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002061154984556364,
+      "loss": 1.0985,
+      "step": 9492
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020609789465335786,
+      "loss": 0.2972,
+      "step": 9493
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002060802899527557,
+      "loss": 2.1041,
+      "step": 9494
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020606268435411187,
+      "loss": 0.0879,
+      "step": 9495
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020604507785770831,
+      "loss": 0.0628,
+      "step": 9496
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020602747046382695,
+      "loss": 3.3086,
+      "step": 9497
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002060098621727497,
+      "loss": 0.0031,
+      "step": 9498
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002059922529847586,
+      "loss": 0.5881,
+      "step": 9499
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002059746429001356,
+      "loss": 0.6696,
+      "step": 9500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020595703191916267,
+      "loss": 0.4462,
+      "step": 9501
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002059394200421219,
+      "loss": 0.1248,
+      "step": 9502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020592180726929524,
+      "loss": 0.4036,
+      "step": 9503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020590419360096477,
+      "loss": 0.0021,
+      "step": 9504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020588657903741254,
+      "loss": 0.939,
+      "step": 9505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002058689635789206,
+      "loss": 2.0246,
+      "step": 9506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020585134722577106,
+      "loss": 0.009,
+      "step": 9507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020583372997824606,
+      "loss": 0.4551,
+      "step": 9508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020581611183662763,
+      "loss": 0.0018,
+      "step": 9509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020579849280119794,
+      "loss": 0.5921,
+      "step": 9510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020578087287223917,
+      "loss": 0.7241,
+      "step": 9511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020576325205003338,
+      "loss": 0.7244,
+      "step": 9512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020574563033486277,
+      "loss": 0.9465,
+      "step": 9513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020572800772700967,
+      "loss": 0.6222,
+      "step": 9514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020571038422675605,
+      "loss": 0.6144,
+      "step": 9515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020569275983438426,
+      "loss": 0.002,
+      "step": 9516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020567513455017655,
+      "loss": 0.399,
+      "step": 9517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002056575083744151,
+      "loss": 0.4679,
+      "step": 9518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002056398813073822,
+      "loss": 0.2285,
+      "step": 9519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020562225334936005,
+      "loss": 0.3041,
+      "step": 9520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020560462450063102,
+      "loss": 0.9227,
+      "step": 9521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020558699476147735,
+      "loss": 0.0319,
+      "step": 9522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020556936413218145,
+      "loss": 0.5452,
+      "step": 9523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002055517326130255,
+      "loss": 0.3989,
+      "step": 9524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020553410020429202,
+      "loss": 0.0128,
+      "step": 9525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020551646690626323,
+      "loss": 0.0282,
+      "step": 9526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002054988327192215,
+      "loss": 0.0303,
+      "step": 9527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020548119764344928,
+      "loss": 0.3569,
+      "step": 9528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020546356167922895,
+      "loss": 0.6366,
+      "step": 9529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002054459248268429,
+      "loss": 0.9035,
+      "step": 9530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002054282870865736,
+      "loss": 0.1822,
+      "step": 9531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020541064845870348,
+      "loss": 0.6328,
+      "step": 9532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020539300894351496,
+      "loss": 0.6594,
+      "step": 9533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020537536854129058,
+      "loss": 0.9429,
+      "step": 9534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020535772725231268,
+      "loss": 0.3737,
+      "step": 9535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020534008507686394,
+      "loss": 1.158,
+      "step": 9536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020532244201522677,
+      "loss": 0.2146,
+      "step": 9537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020530479806768375,
+      "loss": 0.2531,
+      "step": 9538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020528715323451735,
+      "loss": 0.5065,
+      "step": 9539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020526950751601018,
+      "loss": 0.5022,
+      "step": 9540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002052518609124448,
+      "loss": 0.5908,
+      "step": 9541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020523421342410377,
+      "loss": 0.1597,
+      "step": 9542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002052165650512697,
+      "loss": 1.5467,
+      "step": 9543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020519891579422523,
+      "loss": 0.2864,
+      "step": 9544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020518126565325298,
+      "loss": 0.1875,
+      "step": 9545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020516361462863554,
+      "loss": 0.8984,
+      "step": 9546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020514596272065564,
+      "loss": 0.1326,
+      "step": 9547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020512830992959588,
+      "loss": 0.081,
+      "step": 9548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020511065625573898,
+      "loss": 0.3972,
+      "step": 9549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002050930016993676,
+      "loss": 1.5226,
+      "step": 9550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020507534626076454,
+      "loss": 0.0082,
+      "step": 9551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020505768994021246,
+      "loss": 0.2671,
+      "step": 9552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020504003273799407,
+      "loss": 0.3426,
+      "step": 9553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020502237465439216,
+      "loss": 0.7035,
+      "step": 9554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020500471568968954,
+      "loss": 0.4371,
+      "step": 9555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002049870558441689,
+      "loss": 0.5076,
+      "step": 9556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020496939511811308,
+      "loss": 0.7673,
+      "step": 9557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020495173351180494,
+      "loss": 1.1882,
+      "step": 9558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020493407102552722,
+      "loss": 0.3218,
+      "step": 9559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020491640765956285,
+      "loss": 1.1925,
+      "step": 9560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020489874341419459,
+      "loss": 0.7035,
+      "step": 9561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002048810782897053,
+      "loss": 3.8065,
+      "step": 9562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000204863412286378,
+      "loss": 0.8936,
+      "step": 9563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020484574540449542,
+      "loss": 0.7915,
+      "step": 9564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020482807764434055,
+      "loss": 0.3567,
+      "step": 9565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020481040900619633,
+      "loss": 1.9805,
+      "step": 9566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020479273949034564,
+      "loss": 0.2852,
+      "step": 9567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020477506909707147,
+      "loss": 0.0017,
+      "step": 9568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002047573978266568,
+      "loss": 0.3399,
+      "step": 9569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020473972567938456,
+      "loss": 0.0024,
+      "step": 9570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020472205265553776,
+      "loss": 0.879,
+      "step": 9571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020470437875539946,
+      "loss": 0.4751,
+      "step": 9572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020468670397925256,
+      "loss": 0.0059,
+      "step": 9573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002046690283273802,
+      "loss": 0.3701,
+      "step": 9574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002046513518000654,
+      "loss": 0.5657,
+      "step": 9575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020463367439759126,
+      "loss": 0.6732,
+      "step": 9576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020461599612024077,
+      "loss": 0.4121,
+      "step": 9577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020459831696829705,
+      "loss": 0.2079,
+      "step": 9578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020458063694204327,
+      "loss": 0.3166,
+      "step": 9579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020456295604176248,
+      "loss": 0.2934,
+      "step": 9580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020454527426773785,
+      "loss": 0.6405,
+      "step": 9581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020452759162025252,
+      "loss": 0.2296,
+      "step": 9582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020450990809958963,
+      "loss": 0.411,
+      "step": 9583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020449222370603234,
+      "loss": 0.7859,
+      "step": 9584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002044745384398639,
+      "loss": 0.3637,
+      "step": 9585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020445685230136753,
+      "loss": 0.3804,
+      "step": 9586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020443916529082633,
+      "loss": 0.5501,
+      "step": 9587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020442147740852362,
+      "loss": 0.7571,
+      "step": 9588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020440378865474264,
+      "loss": 0.1559,
+      "step": 9589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020438609902976662,
+      "loss": 0.4119,
+      "step": 9590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020436840853387884,
+      "loss": 1.7938,
+      "step": 9591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020435071716736259,
+      "loss": 0.6835,
+      "step": 9592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020433302493050117,
+      "loss": 0.7096,
+      "step": 9593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020431533182357792,
+      "loss": 0.4274,
+      "step": 9594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020429763784687613,
+      "loss": 0.3384,
+      "step": 9595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020427994300067917,
+      "loss": 0.8201,
+      "step": 9596
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020426224728527036,
+      "loss": 0.5158,
+      "step": 9597
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020424455070093314,
+      "loss": 0.3597,
+      "step": 9598
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020422685324795082,
+      "loss": 0.3935,
+      "step": 9599
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020420915492660684,
+      "loss": 0.6771,
+      "step": 9600
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020419145573718461,
+      "loss": 1.5822,
+      "step": 9601
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020417375567996753,
+      "loss": 0.288,
+      "step": 9602
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020415605475523907,
+      "loss": 0.0021,
+      "step": 9603
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020413835296328268,
+      "loss": 0.2935,
+      "step": 9604
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020412065030438176,
+      "loss": 0.1668,
+      "step": 9605
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002041029467788199,
+      "loss": 0.0066,
+      "step": 9606
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020408524238688051,
+      "loss": 0.6077,
+      "step": 9607
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020406753712884717,
+      "loss": 0.2899,
+      "step": 9608
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020404983100500337,
+      "loss": 0.0064,
+      "step": 9609
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002040321240156326,
+      "loss": 0.934,
+      "step": 9610
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020401441616101846,
+      "loss": 0.4455,
+      "step": 9611
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020399670744144455,
+      "loss": 0.0031,
+      "step": 9612
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002039789978571944,
+      "loss": 0.2317,
+      "step": 9613
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020396128740855156,
+      "loss": 0.5406,
+      "step": 9614
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020394357609579975,
+      "loss": 0.362,
+      "step": 9615
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002039258639192225,
+      "loss": 0.2659,
+      "step": 9616
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020390815087910347,
+      "loss": 0.3092,
+      "step": 9617
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002038904369757263,
+      "loss": 0.3004,
+      "step": 9618
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002038727222093747,
+      "loss": 0.3902,
+      "step": 9619
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020385500658033225,
+      "loss": 0.1848,
+      "step": 9620
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020383729008888275,
+      "loss": 0.6251,
+      "step": 9621
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020381957273530978,
+      "loss": 0.5019,
+      "step": 9622
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020380185451989718,
+      "loss": 0.0005,
+      "step": 9623
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020378413544292861,
+      "loss": 0.3665,
+      "step": 9624
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002037664155046878,
+      "loss": 0.1158,
+      "step": 9625
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020374869470545857,
+      "loss": 0.9019,
+      "step": 9626
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002037309730455247,
+      "loss": 0.2016,
+      "step": 9627
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020371325052516987,
+      "loss": 0.0017,
+      "step": 9628
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020369552714467797,
+      "loss": 3.02,
+      "step": 9629
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020367780290433276,
+      "loss": 0.5912,
+      "step": 9630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020366007780441814,
+      "loss": 0.2401,
+      "step": 9631
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020364235184521786,
+      "loss": 0.2487,
+      "step": 9632
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020362462502701583,
+      "loss": 1.8651,
+      "step": 9633
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020360689735009595,
+      "loss": 0.9066,
+      "step": 9634
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020358916881474206,
+      "loss": 0.4509,
+      "step": 9635
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020357143942123805,
+      "loss": 1.1571,
+      "step": 9636
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002035537091698678,
+      "loss": 0.928,
+      "step": 9637
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002035359780609153,
+      "loss": 0.5673,
+      "step": 9638
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020351824609466445,
+      "loss": 0.915,
+      "step": 9639
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002035005132713992,
+      "loss": 0.848,
+      "step": 9640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020348277959140353,
+      "loss": 0.4863,
+      "step": 9641
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020346504505496142,
+      "loss": 0.4466,
+      "step": 9642
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020344730966235684,
+      "loss": 0.2029,
+      "step": 9643
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002034295734138738,
+      "loss": 0.0092,
+      "step": 9644
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002034118363097964,
+      "loss": 2.2321,
+      "step": 9645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020339409835040852,
+      "loss": 0.5428,
+      "step": 9646
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002033763595359943,
+      "loss": 1.038,
+      "step": 9647
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020335861986683782,
+      "loss": 0.2949,
+      "step": 9648
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020334087934322306,
+      "loss": 0.5396,
+      "step": 9649
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020332313796543426,
+      "loss": 0.3204,
+      "step": 9650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020330539573375537,
+      "loss": 0.002,
+      "step": 9651
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020328765264847055,
+      "loss": 0.6824,
+      "step": 9652
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020326990870986404,
+      "loss": 0.3423,
+      "step": 9653
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020325216391821983,
+      "loss": 1.1583,
+      "step": 9654
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020323441827382213,
+      "loss": 0.0493,
+      "step": 9655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020321667177695512,
+      "loss": 0.4908,
+      "step": 9656
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020319892442790296,
+      "loss": 0.3817,
+      "step": 9657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002031811762269499,
+      "loss": 0.5061,
+      "step": 9658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020316342717438014,
+      "loss": 0.2192,
+      "step": 9659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002031456772704778,
+      "loss": 0.4983,
+      "step": 9660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020312792651552727,
+      "loss": 0.7167,
+      "step": 9661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002031101749098127,
+      "loss": 0.355,
+      "step": 9662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020309242245361835,
+      "loss": 0.9346,
+      "step": 9663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020307466914722862,
+      "loss": 0.0144,
+      "step": 9664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020305691499092766,
+      "loss": 0.0023,
+      "step": 9665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020303915998499985,
+      "loss": 0.7867,
+      "step": 9666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002030214041297295,
+      "loss": 0.8064,
+      "step": 9667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002030036474254009,
+      "loss": 0.1042,
+      "step": 9668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020298588987229847,
+      "loss": 0.4888,
+      "step": 9669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020296813147070654,
+      "loss": 0.8096,
+      "step": 9670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020295037222090945,
+      "loss": 0.4301,
+      "step": 9671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020293261212319159,
+      "loss": 0.6575,
+      "step": 9672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020291485117783744,
+      "loss": 0.6631,
+      "step": 9673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020289708938513136,
+      "loss": 0.3374,
+      "step": 9674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020287932674535776,
+      "loss": 0.0011,
+      "step": 9675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020286156325880113,
+      "loss": 0.3993,
+      "step": 9676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020284379892574587,
+      "loss": 0.3497,
+      "step": 9677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020282603374647648,
+      "loss": 0.4656,
+      "step": 9678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020280826772127745,
+      "loss": 0.5019,
+      "step": 9679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020279050085043329,
+      "loss": 0.3545,
+      "step": 9680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020277273313422844,
+      "loss": 0.6921,
+      "step": 9681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002027549645729475,
+      "loss": 0.6049,
+      "step": 9682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020273719516687494,
+      "loss": 0.0009,
+      "step": 9683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020271942491629537,
+      "loss": 0.1431,
+      "step": 9684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002027016538214934,
+      "loss": 1.4245,
+      "step": 9685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020268388188275343,
+      "loss": 0.1621,
+      "step": 9686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020266610910036023,
+      "loss": 0.0024,
+      "step": 9687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002026483354745983,
+      "loss": 1.0172,
+      "step": 9688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002026305610057523,
+      "loss": 0.1571,
+      "step": 9689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020261278569410683,
+      "loss": 0.6285,
+      "step": 9690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020259500953994657,
+      "loss": 0.887,
+      "step": 9691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020257723254355617,
+      "loss": 0.9507,
+      "step": 9692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020255945470522033,
+      "loss": 0.6352,
+      "step": 9693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020254167602522366,
+      "loss": 0.5706,
+      "step": 9694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020252389650385095,
+      "loss": 2.0595,
+      "step": 9695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020250611614138685,
+      "loss": 0.5788,
+      "step": 9696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020248833493811602,
+      "loss": 0.0095,
+      "step": 9697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020247055289432335,
+      "loss": 0.6956,
+      "step": 9698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020245277001029353,
+      "loss": 1.0176,
+      "step": 9699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002024349862863113,
+      "loss": 0.0016,
+      "step": 9700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020241720172266145,
+      "loss": 0.3427,
+      "step": 9701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020239941631962883,
+      "loss": 0.0076,
+      "step": 9702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002023816300774982,
+      "loss": 2.0218,
+      "step": 9703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020236384299655432,
+      "loss": 0.3653,
+      "step": 9704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020234605507708209,
+      "loss": 0.0009,
+      "step": 9705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020232826631936637,
+      "loss": 0.3963,
+      "step": 9706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020231047672369197,
+      "loss": 0.4234,
+      "step": 9707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020229268629034386,
+      "loss": 0.4555,
+      "step": 9708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020227489501960678,
+      "loss": 1.2217,
+      "step": 9709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002022571029117657,
+      "loss": 2.6856,
+      "step": 9710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020223930996710567,
+      "loss": 0.0021,
+      "step": 9711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020222151618591136,
+      "loss": 0.6054,
+      "step": 9712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002022037215684679,
+      "loss": 0.4114,
+      "step": 9713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020218592611506017,
+      "loss": 0.2984,
+      "step": 9714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020216812982597315,
+      "loss": 0.2079,
+      "step": 9715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020215033270149185,
+      "loss": 0.899,
+      "step": 9716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002021325347419012,
+      "loss": 0.0036,
+      "step": 9717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020211473594748628,
+      "loss": 0.8391,
+      "step": 9718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002020969363185321,
+      "loss": 0.0083,
+      "step": 9719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002020791358553236,
+      "loss": 0.5378,
+      "step": 9720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020206133455814593,
+      "loss": 3.511,
+      "step": 9721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020204353242728415,
+      "loss": 0.019,
+      "step": 9722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020202572946302326,
+      "loss": 0.1182,
+      "step": 9723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020200792566564839,
+      "loss": 0.6108,
+      "step": 9724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020199012103544472,
+      "loss": 0.3852,
+      "step": 9725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002019723155726972,
+      "loss": 0.0922,
+      "step": 9726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020195450927769108,
+      "loss": 0.0737,
+      "step": 9727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020193670215071146,
+      "loss": 0.2866,
+      "step": 9728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020191889419204353,
+      "loss": 0.5029,
+      "step": 9729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002019010854019724,
+      "loss": 0.2497,
+      "step": 9730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002018832757807833,
+      "loss": 0.3023,
+      "step": 9731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020186546532876136,
+      "loss": 0.3691,
+      "step": 9732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020184765404619186,
+      "loss": 0.0136,
+      "step": 9733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020182984193336002,
+      "loss": 0.5327,
+      "step": 9734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020181202899055098,
+      "loss": 0.1469,
+      "step": 9735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020179421521805013,
+      "loss": 0.0157,
+      "step": 9736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020177640061614264,
+      "loss": 0.4468,
+      "step": 9737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020175858518511376,
+      "loss": 0.3757,
+      "step": 9738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020174076892524886,
+      "loss": 0.0134,
+      "step": 9739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020172295183683314,
+      "loss": 0.2626,
+      "step": 9740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000201705133920152,
+      "loss": 0.1186,
+      "step": 9741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020168731517549078,
+      "loss": 0.4205,
+      "step": 9742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020166949560313475,
+      "loss": 0.5157,
+      "step": 9743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020165167520336925,
+      "loss": 0.9728,
+      "step": 9744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020163385397647978,
+      "loss": 0.4624,
+      "step": 9745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020161603192275155,
+      "loss": 2.6878,
+      "step": 9746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020159820904247004,
+      "loss": 0.0366,
+      "step": 9747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002015803853359207,
+      "loss": 0.4819,
+      "step": 9748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020156256080338888,
+      "loss": 0.5856,
+      "step": 9749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020154473544516,
+      "loss": 0.0299,
+      "step": 9750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002015269092615196,
+      "loss": 0.8761,
+      "step": 9751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020150908225275303,
+      "loss": 1.0416,
+      "step": 9752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020149125441914584,
+      "loss": 0.2223,
+      "step": 9753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020147342576098346,
+      "loss": 0.5033,
+      "step": 9754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020145559627855142,
+      "loss": 1.4302,
+      "step": 9755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020143776597213525,
+      "loss": 0.3054,
+      "step": 9756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020141993484202046,
+      "loss": 0.3785,
+      "step": 9757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020140210288849256,
+      "loss": 0.3437,
+      "step": 9758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020138427011183712,
+      "loss": 0.505,
+      "step": 9759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020136643651233978,
+      "loss": 3.0439,
+      "step": 9760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020134860209028595,
+      "loss": 1.3182,
+      "step": 9761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020133076684596138,
+      "loss": 0.3311,
+      "step": 9762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020131293077965164,
+      "loss": 0.4037,
+      "step": 9763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020129509389164225,
+      "loss": 0.0249,
+      "step": 9764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020127725618221894,
+      "loss": 0.154,
+      "step": 9765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020125941765166737,
+      "loss": 0.9674,
+      "step": 9766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002012415783002731,
+      "loss": 0.2952,
+      "step": 9767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020122373812832189,
+      "loss": 0.8531,
+      "step": 9768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020120589713609939,
+      "loss": 0.4695,
+      "step": 9769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002011880553238913,
+      "loss": 1.4276,
+      "step": 9770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020117021269198333,
+      "loss": 0.3581,
+      "step": 9771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002011523692406612,
+      "loss": 2.3089,
+      "step": 9772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002011345249702106,
+      "loss": 0.0295,
+      "step": 9773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020111667988091734,
+      "loss": 0.9375,
+      "step": 9774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002010988339730672,
+      "loss": 0.8648,
+      "step": 9775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002010809872469459,
+      "loss": 2.888,
+      "step": 9776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020106313970283927,
+      "loss": 0.2012,
+      "step": 9777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002010452913410331,
+      "loss": 1.3774,
+      "step": 9778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020102744216181316,
+      "loss": 0.644,
+      "step": 9779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002010095921654653,
+      "loss": 0.0057,
+      "step": 9780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020099174135227544,
+      "loss": 0.2152,
+      "step": 9781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002009738897225293,
+      "loss": 0.6021,
+      "step": 9782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020095603727651292,
+      "loss": 1.1058,
+      "step": 9783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000200938184014512,
+      "loss": 2.4063,
+      "step": 9784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002009203299368125,
+      "loss": 0.0225,
+      "step": 9785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020090247504370035,
+      "loss": 0.2434,
+      "step": 9786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002008846193354615,
+      "loss": 0.0392,
+      "step": 9787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020086676281238182,
+      "loss": 0.0133,
+      "step": 9788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002008489054747472,
+      "loss": 0.356,
+      "step": 9789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020083104732284375,
+      "loss": 0.3503,
+      "step": 9790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020081318835695735,
+      "loss": 0.4438,
+      "step": 9791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000200795328577374,
+      "loss": 0.3043,
+      "step": 9792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020077746798437967,
+      "loss": 0.6878,
+      "step": 9793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020075960657826045,
+      "loss": 0.304,
+      "step": 9794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020074174435930227,
+      "loss": 0.0153,
+      "step": 9795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002007238813277912,
+      "loss": 0.4025,
+      "step": 9796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020070601748401333,
+      "loss": 0.5693,
+      "step": 9797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020068815282825466,
+      "loss": 0.3944,
+      "step": 9798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002006702873608013,
+      "loss": 0.2714,
+      "step": 9799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020065242108193934,
+      "loss": 0.3276,
+      "step": 9800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020063455399195487,
+      "loss": 0.0097,
+      "step": 9801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020061668609113402,
+      "loss": 1.4459,
+      "step": 9802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020059881737976288,
+      "loss": 0.3207,
+      "step": 9803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020058094785812763,
+      "loss": 0.4585,
+      "step": 9804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020056307752651443,
+      "loss": 0.8398,
+      "step": 9805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020054520638520943,
+      "loss": 0.3228,
+      "step": 9806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020052733443449874,
+      "loss": 0.4755,
+      "step": 9807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002005094616746687,
+      "loss": 0.3319,
+      "step": 9808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020049158810600545,
+      "loss": 1.2403,
+      "step": 9809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020047371372879512,
+      "loss": 0.4789,
+      "step": 9810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020045583854332404,
+      "loss": 0.5114,
+      "step": 9811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020043796254987849,
+      "loss": 1.122,
+      "step": 9812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020042008574874457,
+      "loss": 0.7811,
+      "step": 9813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002004022081402087,
+      "loss": 0.4179,
+      "step": 9814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020038432972455708,
+      "loss": 0.6862,
+      "step": 9815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020036645050207605,
+      "loss": 0.2282,
+      "step": 9816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020034857047305192,
+      "loss": 0.0068,
+      "step": 9817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000200330689637771,
+      "loss": 0.0957,
+      "step": 9818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002003128079965196,
+      "loss": 0.0667,
+      "step": 9819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020029492554958408,
+      "loss": 0.6008,
+      "step": 9820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020027704229725086,
+      "loss": 0.3556,
+      "step": 9821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020025915823980618,
+      "loss": 0.7825,
+      "step": 9822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002002412733775366,
+      "loss": 0.6634,
+      "step": 9823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002002233877107284,
+      "loss": 0.5014,
+      "step": 9824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020020550123966795,
+      "loss": 0.2984,
+      "step": 9825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020018761396464182,
+      "loss": 0.3245,
+      "step": 9826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020016972588593636,
+      "loss": 0.5887,
+      "step": 9827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020015183700383804,
+      "loss": 0.1237,
+      "step": 9828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002001339473186333,
+      "loss": 1.3789,
+      "step": 9829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020011605683060865,
+      "loss": 0.0758,
+      "step": 9830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020009816554005056,
+      "loss": 0.4874,
+      "step": 9831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020008027344724555,
+      "loss": 0.1785,
+      "step": 9832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020006238055248008,
+      "loss": 0.1783,
+      "step": 9833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002000444868560407,
+      "loss": 1.5118,
+      "step": 9834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020002659235821406,
+      "loss": 0.3388,
+      "step": 9835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002000086970592866,
+      "loss": 0.2633,
+      "step": 9836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019999080095954482,
+      "loss": 0.0049,
+      "step": 9837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019997290405927547,
+      "loss": 0.3227,
+      "step": 9838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019995500635876506,
+      "loss": 0.2799,
+      "step": 9839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019993710785830014,
+      "loss": 1.1631,
+      "step": 9840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019991920855816743,
+      "loss": 0.4442,
+      "step": 9841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001999013084586535,
+      "loss": 0.2438,
+      "step": 9842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019988340756004502,
+      "loss": 0.6391,
+      "step": 9843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001998655058626286,
+      "loss": 0.4123,
+      "step": 9844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000199847603366691,
+      "loss": 0.1554,
+      "step": 9845
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019982970007251878,
+      "loss": 0.5393,
+      "step": 9846
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019981179598039877,
+      "loss": 0.3167,
+      "step": 9847
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997938910906175,
+      "loss": 0.4491,
+      "step": 9848
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997759854034619,
+      "loss": 0.1897,
+      "step": 9849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997580789192185,
+      "loss": 0.1567,
+      "step": 9850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997401716381742,
+      "loss": 0.6685,
+      "step": 9851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997222635606157,
+      "loss": 0.0213,
+      "step": 9852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019970435468682978,
+      "loss": 0.6315,
+      "step": 9853
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001996864450171032,
+      "loss": 0.2057,
+      "step": 9854
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019966853455172278,
+      "loss": 0.5811,
+      "step": 9855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019965062329097528,
+      "loss": 0.011,
+      "step": 9856
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019963271123514756,
+      "loss": 0.0049,
+      "step": 9857
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019961479838452657,
+      "loss": 0.3936,
+      "step": 9858
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001995968847393989,
+      "loss": 0.3259,
+      "step": 9859
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001995789703000516,
+      "loss": 0.8298,
+      "step": 9860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019956105506677157,
+      "loss": 0.8332,
+      "step": 9861
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019954313903984558,
+      "loss": 0.4184,
+      "step": 9862
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019952522221956055,
+      "loss": 0.3983,
+      "step": 9863
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019950730460620343,
+      "loss": 0.1766,
+      "step": 9864
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019948938620006115,
+      "loss": 0.0044,
+      "step": 9865
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001994714670014206,
+      "loss": 0.6955,
+      "step": 9866
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019945354701056872,
+      "loss": 0.3146,
+      "step": 9867
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019943562622779255,
+      "loss": 0.5757,
+      "step": 9868
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000199417704653379,
+      "loss": 0.7708,
+      "step": 9869
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019939978228761512,
+      "loss": 0.2804,
+      "step": 9870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001993818591307878,
+      "loss": 0.798,
+      "step": 9871
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019936393518318416,
+      "loss": 0.389,
+      "step": 9872
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019934601044509117,
+      "loss": 0.275,
+      "step": 9873
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019932808491679582,
+      "loss": 0.9521,
+      "step": 9874
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019931015859858534,
+      "loss": 0.5906,
+      "step": 9875
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001992922314907466,
+      "loss": 0.2854,
+      "step": 9876
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019927430359356675,
+      "loss": 0.5564,
+      "step": 9877
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001992563749073329,
+      "loss": 0.5128,
+      "step": 9878
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001992384454323321,
+      "loss": 0.0201,
+      "step": 9879
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019922051516885154,
+      "loss": 1.6018,
+      "step": 9880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019920258411717826,
+      "loss": 0.4943,
+      "step": 9881
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019918465227759946,
+      "loss": 0.1844,
+      "step": 9882
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019916671965040223,
+      "loss": 0.7172,
+      "step": 9883
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019914878623587384,
+      "loss": 0.3599,
+      "step": 9884
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019913085203430135,
+      "loss": 1.2205,
+      "step": 9885
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000199112917045972,
+      "loss": 0.5094,
+      "step": 9886
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019909498127117304,
+      "loss": 0.7307,
+      "step": 9887
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001990770447101916,
+      "loss": 0.5243,
+      "step": 9888
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001990591073633149,
+      "loss": 0.001,
+      "step": 9889
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001990411692308303,
+      "loss": 0.0636,
+      "step": 9890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019902323031302494,
+      "loss": 0.3562,
+      "step": 9891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019900529061018613,
+      "loss": 0.4745,
+      "step": 9892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001989873501226011,
+      "loss": 0.3608,
+      "step": 9893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001989694088505572,
+      "loss": 0.3786,
+      "step": 9894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019895146679434173,
+      "loss": 0.984,
+      "step": 9895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019893352395424201,
+      "loss": 0.2767,
+      "step": 9896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019891558033054529,
+      "loss": 0.2519,
+      "step": 9897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000198897635923539,
+      "loss": 0.6781,
+      "step": 9898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019887969073351043,
+      "loss": 0.0094,
+      "step": 9899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000198861744760747,
+      "loss": 0.489,
+      "step": 9900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000198843798005536,
+      "loss": 1.4873,
+      "step": 9901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019882585046816494,
+      "loss": 0.3671,
+      "step": 9902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019880790214892111,
+      "loss": 0.8951,
+      "step": 9903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019878995304809202,
+      "loss": 0.2022,
+      "step": 9904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019877200316596505,
+      "loss": 0.6344,
+      "step": 9905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001987540525028276,
+      "loss": 0.4242,
+      "step": 9906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019873610105896724,
+      "loss": 0.3882,
+      "step": 9907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019871814883467128,
+      "loss": 0.7044,
+      "step": 9908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019870019583022734,
+      "loss": 0.0071,
+      "step": 9909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019868224204592288,
+      "loss": 0.2722,
+      "step": 9910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019866428748204534,
+      "loss": 0.3313,
+      "step": 9911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019864633213888223,
+      "loss": 0.4247,
+      "step": 9912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019862837601672116,
+      "loss": 0.0214,
+      "step": 9913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019861041911584966,
+      "loss": 0.8795,
+      "step": 9914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001985924614365552,
+      "loss": 0.011,
+      "step": 9915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019857450297912542,
+      "loss": 0.3256,
+      "step": 9916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001985565437438479,
+      "loss": 0.2161,
+      "step": 9917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019853858373101015,
+      "loss": 0.3745,
+      "step": 9918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019852062294089986,
+      "loss": 2.4294,
+      "step": 9919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019850266137380458,
+      "loss": 0.0314,
+      "step": 9920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000198484699030012,
+      "loss": 0.3743,
+      "step": 9921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019846673590980977,
+      "loss": 0.2962,
+      "step": 9922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019844877201348541,
+      "loss": 0.4672,
+      "step": 9923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019843080734132678,
+      "loss": 0.5733,
+      "step": 9924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019841284189362138,
+      "loss": 0.8232,
+      "step": 9925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019839487567065698,
+      "loss": 1.0552,
+      "step": 9926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001983769086727213,
+      "loss": 0.656,
+      "step": 9927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019835894090010204,
+      "loss": 0.3403,
+      "step": 9928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001983409723530869,
+      "loss": 0.9038,
+      "step": 9929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019832300303196363,
+      "loss": 0.2913,
+      "step": 9930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019830503293702,
+      "loss": 0.6101,
+      "step": 9931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019828706206854375,
+      "loss": 0.194,
+      "step": 9932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001982690904268227,
+      "loss": 0.458,
+      "step": 9933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019825111801214458,
+      "loss": 0.2225,
+      "step": 9934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019823314482479725,
+      "loss": 0.8595,
+      "step": 9935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019821517086506847,
+      "loss": 0.2806,
+      "step": 9936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001981971961332461,
+      "loss": 0.3034,
+      "step": 9937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019817922062961797,
+      "loss": 0.514,
+      "step": 9938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001981612443544719,
+      "loss": 0.0009,
+      "step": 9939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019814326730809581,
+      "loss": 0.3787,
+      "step": 9940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019812528949077752,
+      "loss": 0.001,
+      "step": 9941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019810731090280496,
+      "loss": 0.6375,
+      "step": 9942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019808933154446601,
+      "loss": 0.0082,
+      "step": 9943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019807135141604863,
+      "loss": 0.0042,
+      "step": 9944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019805337051784067,
+      "loss": 0.6967,
+      "step": 9945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019803538885013007,
+      "loss": 0.0058,
+      "step": 9946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019801740641320483,
+      "loss": 0.2935,
+      "step": 9947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019799942320735292,
+      "loss": 0.0035,
+      "step": 9948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019798143923286228,
+      "loss": 0.0479,
+      "step": 9949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019796345449002087,
+      "loss": 0.4144,
+      "step": 9950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019794546897911673,
+      "loss": 0.2431,
+      "step": 9951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019792748270043787,
+      "loss": 0.303,
+      "step": 9952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019790949565427228,
+      "loss": 0.3212,
+      "step": 9953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019789150784090803,
+      "loss": 0.3501,
+      "step": 9954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019787351926063316,
+      "loss": 0.1493,
+      "step": 9955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019785552991373573,
+      "loss": 1.1586,
+      "step": 9956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001978375398005038,
+      "loss": 0.1605,
+      "step": 9957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001978195489212255,
+      "loss": 0.7064,
+      "step": 9958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019780155727618884,
+      "loss": 1.4122,
+      "step": 9959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019778356486568202,
+      "loss": 3.2208,
+      "step": 9960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019776557168999306,
+      "loss": 0.3367,
+      "step": 9961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019774757774941026,
+      "loss": 1.3165,
+      "step": 9962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019772958304422158,
+      "loss": 0.0016,
+      "step": 9963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019771158757471527,
+      "loss": 0.2722,
+      "step": 9964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001976935913411795,
+      "loss": 1.0722,
+      "step": 9965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019767559434390246,
+      "loss": 0.0631,
+      "step": 9966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001976575965831723,
+      "loss": 1.0789,
+      "step": 9967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019763959805927728,
+      "loss": 0.8122,
+      "step": 9968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019762159877250555,
+      "loss": 0.3345,
+      "step": 9969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019760359872314541,
+      "loss": 0.123,
+      "step": 9970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019758559791148512,
+      "loss": 0.5866,
+      "step": 9971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019756759633781284,
+      "loss": 0.3425,
+      "step": 9972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019754959400241693,
+      "loss": 0.3251,
+      "step": 9973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019753159090558561,
+      "loss": 0.0608,
+      "step": 9974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001975135870476072,
+      "loss": 0.3775,
+      "step": 9975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019749558242877004,
+      "loss": 0.0114,
+      "step": 9976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019747757704936236,
+      "loss": 2.3771,
+      "step": 9977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019745957090967255,
+      "loss": 0.7653,
+      "step": 9978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019744156400998894,
+      "loss": 0.0092,
+      "step": 9979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019742355635059988,
+      "loss": 0.035,
+      "step": 9980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019740554793179374,
+      "loss": 0.0058,
+      "step": 9981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019738753875385892,
+      "loss": 0.3105,
+      "step": 9982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019736952881708374,
+      "loss": 1.1187,
+      "step": 9983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019735151812175667,
+      "loss": 0.808,
+      "step": 9984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001973335066681661,
+      "loss": 0.0036,
+      "step": 9985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019731549445660046,
+      "loss": 0.0079,
+      "step": 9986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001972974814873482,
+      "loss": 0.5507,
+      "step": 9987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019727946776069774,
+      "loss": 0.0109,
+      "step": 9988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019726145327693756,
+      "loss": 0.451,
+      "step": 9989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019724343803635615,
+      "loss": 0.0033,
+      "step": 9990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019722542203924197,
+      "loss": 0.0049,
+      "step": 9991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019720740528588355,
+      "loss": 1.9435,
+      "step": 9992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019718938777656933,
+      "loss": 0.8925,
+      "step": 9993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019717136951158794,
+      "loss": 0.3121,
+      "step": 9994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019715335049122785,
+      "loss": 0.0445,
+      "step": 9995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019713533071577757,
+      "loss": 1.906,
+      "step": 9996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001971173101855258,
+      "loss": 0.3503,
+      "step": 9997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019709928890076096,
+      "loss": 0.524,
+      "step": 9998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001970812668617717,
+      "loss": 0.4916,
+      "step": 9999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019706324406884666,
+      "loss": 0.278,
+      "step": 10000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019704522052227435,
+      "loss": 0.3822,
+      "step": 10001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019702719622234342,
+      "loss": 0.0188,
+      "step": 10002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019700917116934256,
+      "loss": 0.3078,
+      "step": 10003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019699114536356034,
+      "loss": 0.3775,
+      "step": 10004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019697311880528546,
+      "loss": 0.9693,
+      "step": 10005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019695509149480662,
+      "loss": 0.0072,
+      "step": 10006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001969370634324124,
+      "loss": 1.966,
+      "step": 10007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001969190346183916,
+      "loss": 0.0074,
+      "step": 10008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019690100505303287,
+      "loss": 0.4529,
+      "step": 10009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001968829747366249,
+      "loss": 0.3094,
+      "step": 10010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019686494366945648,
+      "loss": 0.2204,
+      "step": 10011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001968469118518163,
+      "loss": 0.5246,
+      "step": 10012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019682887928399316,
+      "loss": 0.2834,
+      "step": 10013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001968108459662758,
+      "loss": 0.8908,
+      "step": 10014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019679281189895298,
+      "loss": 0.2918,
+      "step": 10015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001967747770823135,
+      "loss": 0.3483,
+      "step": 10016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019675674151664618,
+      "loss": 0.0171,
+      "step": 10017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019673870520223974,
+      "loss": 0.133,
+      "step": 10018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019672066813938318,
+      "loss": 0.6055,
+      "step": 10019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019670263032836523,
+      "loss": 0.246,
+      "step": 10020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019668459176947468,
+      "loss": 0.2972,
+      "step": 10021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001966665524630005,
+      "loss": 0.464,
+      "step": 10022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019664851240923153,
+      "loss": 0.8457,
+      "step": 10023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019663047160845663,
+      "loss": 0.5884,
+      "step": 10024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019661243006096469,
+      "loss": 0.432,
+      "step": 10025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019659438776704468,
+      "loss": 0.3502,
+      "step": 10026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019657634472698544,
+      "loss": 0.5002,
+      "step": 10027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019655830094107596,
+      "loss": 0.0144,
+      "step": 10028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019654025640960512,
+      "loss": 1.2998,
+      "step": 10029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000196522211132862,
+      "loss": 0.7603,
+      "step": 10030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019650416511113544,
+      "loss": 0.4468,
+      "step": 10031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019648611834471448,
+      "loss": 0.3678,
+      "step": 10032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019646807083388806,
+      "loss": 4.8954,
+      "step": 10033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001964500225789453,
+      "loss": 0.4632,
+      "step": 10034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019643197358017511,
+      "loss": 0.1745,
+      "step": 10035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019641392383786655,
+      "loss": 0.8584,
+      "step": 10036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019639587335230864,
+      "loss": 0.0031,
+      "step": 10037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019637782212379045,
+      "loss": 0.2076,
+      "step": 10038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019635977015260107,
+      "loss": 0.4194,
+      "step": 10039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001963417174390295,
+      "loss": 0.0087,
+      "step": 10040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001963236639833649,
+      "loss": 0.4924,
+      "step": 10041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019630560978589636,
+      "loss": 0.0064,
+      "step": 10042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019628755484691298,
+      "loss": 0.3708,
+      "step": 10043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019626949916670382,
+      "loss": 0.1884,
+      "step": 10044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001962514427455581,
+      "loss": 2.7561,
+      "step": 10045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019623338558376495,
+      "loss": 0.8762,
+      "step": 10046
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019621532768161352,
+      "loss": 0.8672,
+      "step": 10047
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019619726903939293,
+      "loss": 0.3106,
+      "step": 10048
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001961792096573925,
+      "loss": 0.0052,
+      "step": 10049
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019616114953590127,
+      "loss": 3.6192,
+      "step": 10050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001961430886752085,
+      "loss": 1.1958,
+      "step": 10051
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019612502707560343,
+      "loss": 0.5643,
+      "step": 10052
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019610696473737528,
+      "loss": 0.4997,
+      "step": 10053
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001960889016608133,
+      "loss": 0.5004,
+      "step": 10054
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019607083784620673,
+      "loss": 0.3352,
+      "step": 10055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001960527732938448,
+      "loss": 0.7331,
+      "step": 10056
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001960347080040168,
+      "loss": 0.1627,
+      "step": 10057
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019601664197701213,
+      "loss": 0.0241,
+      "step": 10058
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019599857521311994,
+      "loss": 1.0207,
+      "step": 10059
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019598050771262962,
+      "loss": 0.5612,
+      "step": 10060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019596243947583045,
+      "loss": 0.2697,
+      "step": 10061
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019594437050301182,
+      "loss": 0.529,
+      "step": 10062
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019592630079446306,
+      "loss": 0.3047,
+      "step": 10063
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019590823035047346,
+      "loss": 0.0111,
+      "step": 10064
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001958901591713325,
+      "loss": 0.6289,
+      "step": 10065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019587208725732955,
+      "loss": 0.7084,
+      "step": 10066
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001958540146087539,
+      "loss": 0.0039,
+      "step": 10067
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019583594122589502,
+      "loss": 0.2462,
+      "step": 10068
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019581786710904238,
+      "loss": 0.8627,
+      "step": 10069
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019579979225848527,
+      "loss": 3.3193,
+      "step": 10070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019578171667451332,
+      "loss": 1.7227,
+      "step": 10071
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019576364035741586,
+      "loss": 0.5221,
+      "step": 10072
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019574556330748236,
+      "loss": 1.9904,
+      "step": 10073
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019572748552500231,
+      "loss": 0.5436,
+      "step": 10074
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019570940701026527,
+      "loss": 0.2648,
+      "step": 10075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019569132776356066,
+      "loss": 0.0675,
+      "step": 10076
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019567324778517795,
+      "loss": 0.0165,
+      "step": 10077
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019565516707540677,
+      "loss": 0.7927,
+      "step": 10078
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001956370856345366,
+      "loss": 0.5897,
+      "step": 10079
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019561900346285694,
+      "loss": 0.0273,
+      "step": 10080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019560092056065753,
+      "loss": 0.6859,
+      "step": 10081
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001955828369282277,
+      "loss": 0.9417,
+      "step": 10082
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019556475256585717,
+      "loss": 0.8288,
+      "step": 10083
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019554666747383554,
+      "loss": 0.1923,
+      "step": 10084
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019552858165245237,
+      "loss": 0.0049,
+      "step": 10085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001955104951019973,
+      "loss": 0.5299,
+      "step": 10086
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019549240782275996,
+      "loss": 0.4983,
+      "step": 10087
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019547431981502996,
+      "loss": 1.0788,
+      "step": 10088
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000195456231079097,
+      "loss": 0.013,
+      "step": 10089
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019543814161525068,
+      "loss": 0.0062,
+      "step": 10090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019542005142378076,
+      "loss": 0.3012,
+      "step": 10091
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019540196050497687,
+      "loss": 0.2622,
+      "step": 10092
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019538386885912872,
+      "loss": 0.6219,
+      "step": 10093
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019536577648652598,
+      "loss": 0.1924,
+      "step": 10094
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019534768338745848,
+      "loss": 0.7321,
+      "step": 10095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001953295895622159,
+      "loss": 0.3262,
+      "step": 10096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019531149501108788,
+      "loss": 0.6851,
+      "step": 10097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019529339973436434,
+      "loss": 2.3533,
+      "step": 10098
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019527530373233498,
+      "loss": 0.0104,
+      "step": 10099
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019525720700528954,
+      "loss": 0.3167,
+      "step": 10100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001952391095535179,
+      "loss": 0.3894,
+      "step": 10101
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019522101137730977,
+      "loss": 0.144,
+      "step": 10102
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000195202912476955,
+      "loss": 0.757,
+      "step": 10103
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019518481285274347,
+      "loss": 0.5734,
+      "step": 10104
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019516671250496498,
+      "loss": 0.7229,
+      "step": 10105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019514861143390933,
+      "loss": 0.4366,
+      "step": 10106
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019513050963986648,
+      "loss": 0.7839,
+      "step": 10107
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001951124071231262,
+      "loss": 0.6552,
+      "step": 10108
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019509430388397849,
+      "loss": 0.4587,
+      "step": 10109
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950761999227131,
+      "loss": 0.7134,
+      "step": 10110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019505809523962005,
+      "loss": 0.0557,
+      "step": 10111
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019503998983498924,
+      "loss": 0.4992,
+      "step": 10112
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019502188370911058,
+      "loss": 0.0132,
+      "step": 10113
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019500377686227402,
+      "loss": 0.4228,
+      "step": 10114
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019498566929476954,
+      "loss": 0.3147,
+      "step": 10115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019496756100688701,
+      "loss": 0.4338,
+      "step": 10116
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019494945199891655,
+      "loss": 0.0085,
+      "step": 10117
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019493134227114805,
+      "loss": 0.6633,
+      "step": 10118
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019491323182387152,
+      "loss": 0.3447,
+      "step": 10119
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000194895120657377,
+      "loss": 0.2175,
+      "step": 10120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019487700877195456,
+      "loss": 0.6588,
+      "step": 10121
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019485889616789408,
+      "loss": 0.004,
+      "step": 10122
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001948407828454857,
+      "loss": 0.207,
+      "step": 10123
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019482266880501954,
+      "loss": 4.3148,
+      "step": 10124
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001948045540467856,
+      "loss": 0.0492,
+      "step": 10125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019478643857107398,
+      "loss": 0.2009,
+      "step": 10126
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019476832237817472,
+      "loss": 1.2412,
+      "step": 10127
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000194750205468378,
+      "loss": 0.3836,
+      "step": 10128
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019473208784197385,
+      "loss": 0.8216,
+      "step": 10129
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019471396949925253,
+      "loss": 0.0453,
+      "step": 10130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000194695850440504,
+      "loss": 0.1713,
+      "step": 10131
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019467773066601855,
+      "loss": 0.0349,
+      "step": 10132
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001946596101760863,
+      "loss": 1.5185,
+      "step": 10133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019464148897099744,
+      "loss": 0.4264,
+      "step": 10134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019462336705104212,
+      "loss": 0.5312,
+      "step": 10135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019460524441651053,
+      "loss": 0.2368,
+      "step": 10136
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001945871210676929,
+      "loss": 0.302,
+      "step": 10137
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019456899700487942,
+      "loss": 0.3762,
+      "step": 10138
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019455087222836035,
+      "loss": 1.0077,
+      "step": 10139
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019453274673842592,
+      "loss": 0.3641,
+      "step": 10140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001945146205353664,
+      "loss": 0.5722,
+      "step": 10141
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019449649361947201,
+      "loss": 0.554,
+      "step": 10142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019447836599103312,
+      "loss": 0.8956,
+      "step": 10143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019446023765033992,
+      "loss": 0.2633,
+      "step": 10144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019444210859768273,
+      "loss": 0.2442,
+      "step": 10145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019442397883335185,
+      "loss": 0.5417,
+      "step": 10146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019440584835763767,
+      "loss": 0.6232,
+      "step": 10147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019438771717083042,
+      "loss": 0.7403,
+      "step": 10148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001943695852732205,
+      "loss": 0.5726,
+      "step": 10149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001943514526650983,
+      "loss": 0.2371,
+      "step": 10150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019433331934675413,
+      "loss": 0.5468,
+      "step": 10151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019431518531847836,
+      "loss": 0.2793,
+      "step": 10152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019429705058056145,
+      "loss": 0.626,
+      "step": 10153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019427891513329373,
+      "loss": 0.4377,
+      "step": 10154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019426077897696564,
+      "loss": 0.3889,
+      "step": 10155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019424264211186763,
+      "loss": 0.4417,
+      "step": 10156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019422450453829006,
+      "loss": 0.4295,
+      "step": 10157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019420636625652342,
+      "loss": 0.748,
+      "step": 10158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019418822726685822,
+      "loss": 6.0747,
+      "step": 10159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019417008756958482,
+      "loss": 0.0042,
+      "step": 10160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019415194716499378,
+      "loss": 0.5626,
+      "step": 10161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019413380605337558,
+      "loss": 0.4108,
+      "step": 10162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001941156642350207,
+      "loss": 0.6667,
+      "step": 10163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019409752171021968,
+      "loss": 0.5798,
+      "step": 10164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000194079378479263,
+      "loss": 0.3224,
+      "step": 10165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019406123454244128,
+      "loss": 0.7788,
+      "step": 10166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019404308990004495,
+      "loss": 0.6013,
+      "step": 10167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019402494455236468,
+      "loss": 0.3051,
+      "step": 10168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019400679849969097,
+      "loss": 0.0126,
+      "step": 10169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001939886517423145,
+      "loss": 0.1359,
+      "step": 10170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019397050428052572,
+      "loss": 0.2884,
+      "step": 10171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019395235611461528,
+      "loss": 0.003,
+      "step": 10172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001939342072448739,
+      "loss": 2.4266,
+      "step": 10173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019391605767159206,
+      "loss": 0.3629,
+      "step": 10174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001938979073950605,
+      "loss": 1.0506,
+      "step": 10175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001938797564155698,
+      "loss": 0.2989,
+      "step": 10176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019386160473341068,
+      "loss": 3.1247,
+      "step": 10177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001938434523488738,
+      "loss": 0.5048,
+      "step": 10178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019382529926224984,
+      "loss": 1.1619,
+      "step": 10179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019380714547382943,
+      "loss": 0.345,
+      "step": 10180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019378899098390336,
+      "loss": 0.4154,
+      "step": 10181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019377083579276232,
+      "loss": 0.0096,
+      "step": 10182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019375267990069705,
+      "loss": 0.5007,
+      "step": 10183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019373452330799818,
+      "loss": 0.5087,
+      "step": 10184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019371636601495668,
+      "loss": 1.5758,
+      "step": 10185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019369820802186315,
+      "loss": 0.5388,
+      "step": 10186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019368004932900838,
+      "loss": 0.5404,
+      "step": 10187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019366188993668319,
+      "loss": 0.2824,
+      "step": 10188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019364372984517834,
+      "loss": 0.2064,
+      "step": 10189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001936255690547847,
+      "loss": 1.049,
+      "step": 10190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000193607407565793,
+      "loss": 0.0019,
+      "step": 10191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019358924537849414,
+      "loss": 0.309,
+      "step": 10192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019357108249317892,
+      "loss": 0.6444,
+      "step": 10193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019355291891013824,
+      "loss": 0.2538,
+      "step": 10194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019353475462966283,
+      "loss": 0.0189,
+      "step": 10195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001935165896520438,
+      "loss": 0.0024,
+      "step": 10196
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019349842397757178,
+      "loss": 2.6485,
+      "step": 10197
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019348025760653782,
+      "loss": 0.0036,
+      "step": 10198
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001934620905392328,
+      "loss": 0.6729,
+      "step": 10199
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019344392277594762,
+      "loss": 0.4236,
+      "step": 10200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019342575431697325,
+      "loss": 0.2683,
+      "step": 10201
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019340758516260054,
+      "loss": 0.2444,
+      "step": 10202
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001933894153131205,
+      "loss": 0.2788,
+      "step": 10203
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019337124476882415,
+      "loss": 0.1543,
+      "step": 10204
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001933530735300024,
+      "loss": 0.1603,
+      "step": 10205
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001933349015969462,
+      "loss": 1.448,
+      "step": 10206
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001933167289699466,
+      "loss": 0.7955,
+      "step": 10207
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001932985556492946,
+      "loss": 0.8667,
+      "step": 10208
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019328038163528123,
+      "loss": 0.2928,
+      "step": 10209
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019326220692819743,
+      "loss": 1.0118,
+      "step": 10210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019324403152833438,
+      "loss": 0.8623,
+      "step": 10211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019322585543598306,
+      "loss": 0.742,
+      "step": 10212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019320767865143454,
+      "loss": 0.205,
+      "step": 10213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019318950117497988,
+      "loss": 0.2952,
+      "step": 10214
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019317132300691018,
+      "loss": 0.3517,
+      "step": 10215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019315314414751654,
+      "loss": 2.3771,
+      "step": 10216
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019313496459709005,
+      "loss": 0.7316,
+      "step": 10217
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001931167843559218,
+      "loss": 0.7733,
+      "step": 10218
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019309860342430307,
+      "loss": 0.197,
+      "step": 10219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001930804218025248,
+      "loss": 0.3705,
+      "step": 10220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019306223949087822,
+      "loss": 0.7435,
+      "step": 10221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019304405648965455,
+      "loss": 0.2448,
+      "step": 10222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001930258727991449,
+      "loss": 0.5696,
+      "step": 10223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019300768841964044,
+      "loss": 0.0427,
+      "step": 10224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019298950335143244,
+      "loss": 0.2739,
+      "step": 10225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019297131759481203,
+      "loss": 0.0053,
+      "step": 10226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019295313115007048,
+      "loss": 0.2791,
+      "step": 10227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000192934944017499,
+      "loss": 0.345,
+      "step": 10228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001929167561973888,
+      "loss": 0.3998,
+      "step": 10229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019289856769003115,
+      "loss": 0.1719,
+      "step": 10230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019288037849571742,
+      "loss": 0.2546,
+      "step": 10231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001928621886147387,
+      "loss": 0.0103,
+      "step": 10232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019284399804738632,
+      "loss": 0.0592,
+      "step": 10233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019282580679395167,
+      "loss": 0.0404,
+      "step": 10234
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019280761485472602,
+      "loss": 1.9289,
+      "step": 10235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019278942223000057,
+      "loss": 1.2985,
+      "step": 10236
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019277122892006685,
+      "loss": 0.4888,
+      "step": 10237
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000192753034925216,
+      "loss": 0.4216,
+      "step": 10238
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001927348402457395,
+      "loss": 0.738,
+      "step": 10239
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019271664488192868,
+      "loss": 0.4823,
+      "step": 10240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019269844883407488,
+      "loss": 0.3533,
+      "step": 10241
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019268025210246952,
+      "loss": 0.6814,
+      "step": 10242
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019266205468740396,
+      "loss": 0.7545,
+      "step": 10243
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019264385658916962,
+      "loss": 0.6435,
+      "step": 10244
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019262565780805795,
+      "loss": 0.6083,
+      "step": 10245
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019260745834436034,
+      "loss": 0.241,
+      "step": 10246
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019258925819836816,
+      "loss": 0.3632,
+      "step": 10247
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000192571057370373,
+      "loss": 1.6632,
+      "step": 10248
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019255285586066621,
+      "loss": 0.5472,
+      "step": 10249
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001925346536695393,
+      "loss": 0.2271,
+      "step": 10250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019251645079728375,
+      "loss": 0.3805,
+      "step": 10251
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019249824724419103,
+      "loss": 0.4974,
+      "step": 10252
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001924800430105527,
+      "loss": 0.1372,
+      "step": 10253
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019246183809666024,
+      "loss": 0.3907,
+      "step": 10254
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001924436325028051,
+      "loss": 1.3297,
+      "step": 10255
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019242542622927895,
+      "loss": 1.3975,
+      "step": 10256
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019240721927637328,
+      "loss": 0.2769,
+      "step": 10257
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001923890116443796,
+      "loss": 0.1732,
+      "step": 10258
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019237080333358949,
+      "loss": 0.0063,
+      "step": 10259
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019235259434429464,
+      "loss": 0.3486,
+      "step": 10260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001923343846767865,
+      "loss": 1.4199,
+      "step": 10261
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019231617433135672,
+      "loss": 0.4491,
+      "step": 10262
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001922979633082969,
+      "loss": 0.8063,
+      "step": 10263
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001922797516078987,
+      "loss": 0.7906,
+      "step": 10264
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019226153923045377,
+      "loss": 0.245,
+      "step": 10265
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019224332617625366,
+      "loss": 3.5488,
+      "step": 10266
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019222511244559008,
+      "loss": 0.3121,
+      "step": 10267
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019220689803875474,
+      "loss": 0.0519,
+      "step": 10268
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019218868295603925,
+      "loss": 0.0128,
+      "step": 10269
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019217046719773528,
+      "loss": 0.4507,
+      "step": 10270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001921522507641346,
+      "loss": 0.6264,
+      "step": 10271
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019213403365552889,
+      "loss": 0.4215,
+      "step": 10272
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019211581587220988,
+      "loss": 0.3652,
+      "step": 10273
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019209759741446925,
+      "loss": 0.4353,
+      "step": 10274
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001920793782825988,
+      "loss": 0.7433,
+      "step": 10275
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019206115847689024,
+      "loss": 0.8134,
+      "step": 10276
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019204293799763536,
+      "loss": 0.0059,
+      "step": 10277
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019202471684512593,
+      "loss": 0.6835,
+      "step": 10278
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019200649501965373,
+      "loss": 0.5911,
+      "step": 10279
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001919882725215106,
+      "loss": 0.6855,
+      "step": 10280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019197004935098824,
+      "loss": 0.6616,
+      "step": 10281
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019195182550837853,
+      "loss": 0.1954,
+      "step": 10282
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019193360099397334,
+      "loss": 0.1881,
+      "step": 10283
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001919153758080644,
+      "loss": 0.4153,
+      "step": 10284
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019189714995094365,
+      "loss": 0.3832,
+      "step": 10285
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019187892342290296,
+      "loss": 1.3482,
+      "step": 10286
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001918606962242341,
+      "loss": 0.3983,
+      "step": 10287
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019184246835522904,
+      "loss": 0.4395,
+      "step": 10288
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019182423981617964,
+      "loss": 1.0403,
+      "step": 10289
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019180601060737786,
+      "loss": 0.597,
+      "step": 10290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001917877807291155,
+      "loss": 0.5666,
+      "step": 10291
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019176955018168458,
+      "loss": 0.3589,
+      "step": 10292
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019175131896537697,
+      "loss": 0.3908,
+      "step": 10293
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019173308708048467,
+      "loss": 1.0634,
+      "step": 10294
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019171485452729964,
+      "loss": 0.3354,
+      "step": 10295
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019169662130611378,
+      "loss": 3.1251,
+      "step": 10296
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001916783874172191,
+      "loss": 0.6232,
+      "step": 10297
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019166015286090763,
+      "loss": 1.3241,
+      "step": 10298
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019164191763747132,
+      "loss": 0.0079,
+      "step": 10299
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019162368174720218,
+      "loss": 1.5039,
+      "step": 10300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019160544519039225,
+      "loss": 1.5077,
+      "step": 10301
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019158720796733357,
+      "loss": 0.0073,
+      "step": 10302
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019156897007831816,
+      "loss": 0.9971,
+      "step": 10303
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019155073152363807,
+      "loss": 0.0079,
+      "step": 10304
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019153249230358538,
+      "loss": 0.325,
+      "step": 10305
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001915142524184522,
+      "loss": 0.5952,
+      "step": 10306
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019149601186853053,
+      "loss": 0.4296,
+      "step": 10307
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019147777065411248,
+      "loss": 0.8874,
+      "step": 10308
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019145952877549027,
+      "loss": 0.4909,
+      "step": 10309
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019144128623295586,
+      "loss": 0.1195,
+      "step": 10310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019142304302680147,
+      "loss": 0.0073,
+      "step": 10311
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019140479915731918,
+      "loss": 1.9931,
+      "step": 10312
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019138655462480118,
+      "loss": 2.1944,
+      "step": 10313
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019136830942953964,
+      "loss": 0.7357,
+      "step": 10314
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001913500635718267,
+      "loss": 0.743,
+      "step": 10315
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019133181705195456,
+      "loss": 1.3232,
+      "step": 10316
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001913135698702154,
+      "loss": 0.3631,
+      "step": 10317
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019129532202690142,
+      "loss": 0.4376,
+      "step": 10318
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019127707352230479,
+      "loss": 0.007,
+      "step": 10319
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019125882435671785,
+      "loss": 1.2484,
+      "step": 10320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019124057453043274,
+      "loss": 0.5791,
+      "step": 10321
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001912223240437417,
+      "loss": 0.5616,
+      "step": 10322
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000191204072896937,
+      "loss": 1.0256,
+      "step": 10323
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019118582109031092,
+      "loss": 0.8885,
+      "step": 10324
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019116756862415572,
+      "loss": 0.0123,
+      "step": 10325
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001911493154987637,
+      "loss": 0.896,
+      "step": 10326
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019113106171442717,
+      "loss": 0.7306,
+      "step": 10327
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019111280727143837,
+      "loss": 0.1324,
+      "step": 10328
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019109455217008973,
+      "loss": 0.184,
+      "step": 10329
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019107629641067347,
+      "loss": 0.4833,
+      "step": 10330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019105803999348194,
+      "loss": 0.0127,
+      "step": 10331
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019103978291880758,
+      "loss": 0.3527,
+      "step": 10332
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019102152518694264,
+      "loss": 0.292,
+      "step": 10333
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019100326679817952,
+      "loss": 0.1172,
+      "step": 10334
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001909850077528107,
+      "loss": 1.0891,
+      "step": 10335
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001909667480511284,
+      "loss": 0.7626,
+      "step": 10336
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001909484876934252,
+      "loss": 0.0194,
+      "step": 10337
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019093022667999337,
+      "loss": 1.0266,
+      "step": 10338
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019091196501112541,
+      "loss": 0.6142,
+      "step": 10339
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001908937026871137,
+      "loss": 0.2335,
+      "step": 10340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019087543970825079,
+      "loss": 0.7383,
+      "step": 10341
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019085717607482898,
+      "loss": 0.5382,
+      "step": 10342
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019083891178714083,
+      "loss": 0.3411,
+      "step": 10343
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019082064684547884,
+      "loss": 0.3067,
+      "step": 10344
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001908023812501354,
+      "loss": 0.3637,
+      "step": 10345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019078411500140304,
+      "loss": 0.1537,
+      "step": 10346
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019076584809957437,
+      "loss": 0.7199,
+      "step": 10347
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019074758054494175,
+      "loss": 0.633,
+      "step": 10348
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019072931233779785,
+      "loss": 0.0806,
+      "step": 10349
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019071104347843508,
+      "loss": 0.372,
+      "step": 10350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019069277396714608,
+      "loss": 2.4063,
+      "step": 10351
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019067450380422333,
+      "loss": 0.0326,
+      "step": 10352
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001906562329899595,
+      "loss": 0.4156,
+      "step": 10353
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019063796152464708,
+      "loss": 1.8339,
+      "step": 10354
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019061968940857875,
+      "loss": 0.3084,
+      "step": 10355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019060141664204702,
+      "loss": 0.7064,
+      "step": 10356
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019058314322534448,
+      "loss": 1.1097,
+      "step": 10357
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001905648691587639,
+      "loss": 0.5593,
+      "step": 10358
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019054659444259775,
+      "loss": 0.501,
+      "step": 10359
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019052831907713882,
+      "loss": 0.387,
+      "step": 10360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001905100430626796,
+      "loss": 0.8787,
+      "step": 10361
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019049176639951288,
+      "loss": 0.0112,
+      "step": 10362
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001904734890879313,
+      "loss": 0.0081,
+      "step": 10363
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019045521112822752,
+      "loss": 0.3041,
+      "step": 10364
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019043693252069423,
+      "loss": 0.6035,
+      "step": 10365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019041865326562416,
+      "loss": 0.9494,
+      "step": 10366
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019040037336331007,
+      "loss": 0.2167,
+      "step": 10367
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019038209281404457,
+      "loss": 0.0099,
+      "step": 10368
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019036381161812046,
+      "loss": 0.7551,
+      "step": 10369
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019034552977583054,
+      "loss": 0.1885,
+      "step": 10370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019032724728746748,
+      "loss": 0.014,
+      "step": 10371
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019030896415332408,
+      "loss": 0.2345,
+      "step": 10372
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001902906803736931,
+      "loss": 0.3564,
+      "step": 10373
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019027239594886732,
+      "loss": 0.3739,
+      "step": 10374
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001902541108791396,
+      "loss": 0.183,
+      "step": 10375
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001902358251648027,
+      "loss": 0.3509,
+      "step": 10376
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019021753880614942,
+      "loss": 0.0089,
+      "step": 10377
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019019925180347266,
+      "loss": 0.4999,
+      "step": 10378
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019018096415706522,
+      "loss": 0.0053,
+      "step": 10379
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019016267586721987,
+      "loss": 0.0019,
+      "step": 10380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001901443869342296,
+      "loss": 0.2746,
+      "step": 10381
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001901260973583872,
+      "loss": 0.1578,
+      "step": 10382
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019010780713998556,
+      "loss": 0.1621,
+      "step": 10383
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001900895162793176,
+      "loss": 1.0351,
+      "step": 10384
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001900712247766762,
+      "loss": 0.0065,
+      "step": 10385
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019005293263235428,
+      "loss": 0.8029,
+      "step": 10386
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019003463984664472,
+      "loss": 0.1659,
+      "step": 10387
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019001634641984056,
+      "loss": 0.2935,
+      "step": 10388
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018999805235223461,
+      "loss": 0.2904,
+      "step": 10389
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018997975764411992,
+      "loss": 0.4704,
+      "step": 10390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018996146229578937,
+      "loss": 0.0054,
+      "step": 10391
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018994316630753597,
+      "loss": 0.1842,
+      "step": 10392
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018992486967965276,
+      "loss": 0.649,
+      "step": 10393
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018990657241243263,
+      "loss": 0.8866,
+      "step": 10394
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018988827450616865,
+      "loss": 0.2386,
+      "step": 10395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018986997596115383,
+      "loss": 0.8035,
+      "step": 10396
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001898516767776812,
+      "loss": 0.4389,
+      "step": 10397
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018983337695604373,
+      "loss": 0.7765,
+      "step": 10398
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018981507649653453,
+      "loss": 1.5515,
+      "step": 10399
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018979677539944663,
+      "loss": 0.4442,
+      "step": 10400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018977847366507308,
+      "loss": 0.6948,
+      "step": 10401
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018976017129370703,
+      "loss": 0.2659,
+      "step": 10402
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018974186828564146,
+      "loss": 0.6062,
+      "step": 10403
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001897235646411695,
+      "loss": 0.3372,
+      "step": 10404
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018970526036058433,
+      "loss": 0.6496,
+      "step": 10405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018968695544417896,
+      "loss": 0.5535,
+      "step": 10406
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018966864989224659,
+      "loss": 0.0017,
+      "step": 10407
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018965034370508032,
+      "loss": 0.0014,
+      "step": 10408
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018963203688297328,
+      "loss": 1.3208,
+      "step": 10409
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018961372942621863,
+      "loss": 0.9907,
+      "step": 10410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001895954213351096,
+      "loss": 0.5037,
+      "step": 10411
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001895771126099393,
+      "loss": 0.3454,
+      "step": 10412
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018955880325100091,
+      "loss": 0.554,
+      "step": 10413
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001895404932585877,
+      "loss": 2.7715,
+      "step": 10414
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018952218263299284,
+      "loss": 0.3533,
+      "step": 10415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018950387137450954,
+      "loss": 0.3383,
+      "step": 10416
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018948555948343097,
+      "loss": 0.0094,
+      "step": 10417
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018946724696005047,
+      "loss": 0.0231,
+      "step": 10418
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018944893380466125,
+      "loss": 0.7094,
+      "step": 10419
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018943062001755654,
+      "loss": 0.8826,
+      "step": 10420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001894123055990296,
+      "loss": 0.3897,
+      "step": 10421
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018939399054937377,
+      "loss": 0.5977,
+      "step": 10422
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018937567486888233,
+      "loss": 0.8332,
+      "step": 10423
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001893573585578485,
+      "loss": 0.6374,
+      "step": 10424
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018933904161656567,
+      "loss": 0.0354,
+      "step": 10425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018932072404532715,
+      "loss": 0.2379,
+      "step": 10426
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001893024058444262,
+      "loss": 0.9867,
+      "step": 10427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018928408701415623,
+      "loss": 0.1978,
+      "step": 10428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018926576755481052,
+      "loss": 0.3141,
+      "step": 10429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018924744746668252,
+      "loss": 0.2707,
+      "step": 10430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018922912675006556,
+      "loss": 0.2694,
+      "step": 10431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018921080540525295,
+      "loss": 1.0119,
+      "step": 10432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018919248343253822,
+      "loss": 0.6146,
+      "step": 10433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018917416083221464,
+      "loss": 0.7328,
+      "step": 10434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018915583760457567,
+      "loss": 0.0046,
+      "step": 10435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018913751374991474,
+      "loss": 0.0098,
+      "step": 10436
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001891191892685252,
+      "loss": 2.6222,
+      "step": 10437
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018910086416070061,
+      "loss": 0.836,
+      "step": 10438
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001890825384267344,
+      "loss": 0.0021,
+      "step": 10439
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018906421206691992,
+      "loss": 0.1749,
+      "step": 10440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018904588508155074,
+      "loss": 0.2164,
+      "step": 10441
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018902755747092032,
+      "loss": 1.3978,
+      "step": 10442
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018900922923532212,
+      "loss": 0.4988,
+      "step": 10443
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018899090037504963,
+      "loss": 0.2636,
+      "step": 10444
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018897257089039646,
+      "loss": 0.7424,
+      "step": 10445
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018895424078165603,
+      "loss": 0.2238,
+      "step": 10446
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018893591004912186,
+      "loss": 0.2463,
+      "step": 10447
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001889175786930876,
+      "loss": 0.227,
+      "step": 10448
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018889924671384664,
+      "loss": 0.0166,
+      "step": 10449
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018888091411169267,
+      "loss": 0.337,
+      "step": 10450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018886258088691922,
+      "loss": 0.1885,
+      "step": 10451
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018884424703981983,
+      "loss": 0.58,
+      "step": 10452
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018882591257068812,
+      "loss": 4.2964,
+      "step": 10453
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018880757747981775,
+      "loss": 0.9495,
+      "step": 10454
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018878924176750217,
+      "loss": 0.4221,
+      "step": 10455
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018877090543403514,
+      "loss": 0.0375,
+      "step": 10456
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018875256847971028,
+      "loss": 0.9246,
+      "step": 10457
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001887342309048212,
+      "loss": 0.2325,
+      "step": 10458
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001887158927096615,
+      "loss": 0.8364,
+      "step": 10459
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001886975538945249,
+      "loss": 0.371,
+      "step": 10460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018867921445970504,
+      "loss": 0.8108,
+      "step": 10461
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018866087440549563,
+      "loss": 0.0168,
+      "step": 10462
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018864253373219032,
+      "loss": 0.2841,
+      "step": 10463
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018862419244008286,
+      "loss": 0.9958,
+      "step": 10464
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018860585052946693,
+      "loss": 0.4325,
+      "step": 10465
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018858750800063617,
+      "loss": 0.0088,
+      "step": 10466
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001885691648538844,
+      "loss": 0.0869,
+      "step": 10467
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018855082108950542,
+      "loss": 0.7236,
+      "step": 10468
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018853247670779283,
+      "loss": 0.0079,
+      "step": 10469
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018851413170904043,
+      "loss": 0.2729,
+      "step": 10470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001884957860935421,
+      "loss": 0.19,
+      "step": 10471
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018847743986159148,
+      "loss": 0.3514,
+      "step": 10472
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001884590930134824,
+      "loss": 0.0042,
+      "step": 10473
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018844074554950864,
+      "loss": 0.0025,
+      "step": 10474
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018842239746996408,
+      "loss": 0.8413,
+      "step": 10475
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018840404877514243,
+      "loss": 0.6925,
+      "step": 10476
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018838569946533768,
+      "loss": 0.3152,
+      "step": 10477
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018836734954084345,
+      "loss": 0.539,
+      "step": 10478
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018834899900195375,
+      "loss": 0.9462,
+      "step": 10479
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001883306478489624,
+      "loss": 0.3476,
+      "step": 10480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001883122960821632,
+      "loss": 0.6839,
+      "step": 10481
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001882939437018501,
+      "loss": 0.4835,
+      "step": 10482
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000188275590708317,
+      "loss": 0.0018,
+      "step": 10483
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018825723710185773,
+      "loss": 0.543,
+      "step": 10484
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001882388828827662,
+      "loss": 0.5091,
+      "step": 10485
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001882205280513364,
+      "loss": 1.4287,
+      "step": 10486
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018820217260786216,
+      "loss": 0.3769,
+      "step": 10487
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018818381655263746,
+      "loss": 0.6064,
+      "step": 10488
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018816545988595626,
+      "loss": 0.485,
+      "step": 10489
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001881471026081125,
+      "loss": 0.514,
+      "step": 10490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001881287447194002,
+      "loss": 0.8659,
+      "step": 10491
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018811038622011317,
+      "loss": 0.1555,
+      "step": 10492
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018809202711054555,
+      "loss": 0.7476,
+      "step": 10493
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018807366739099127,
+      "loss": 0.4446,
+      "step": 10494
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001880553070617444,
+      "loss": 0.1989,
+      "step": 10495
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001880369461230988,
+      "loss": 0.2991,
+      "step": 10496
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001880185845753487,
+      "loss": 0.0173,
+      "step": 10497
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018800022241878794,
+      "loss": 1.0073,
+      "step": 10498
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001879818596537107,
+      "loss": 0.2634,
+      "step": 10499
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018796349628041095,
+      "loss": 0.0059,
+      "step": 10500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001879451322991828,
+      "loss": 0.0056,
+      "step": 10501
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001879267677103203,
+      "loss": 0.5516,
+      "step": 10502
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018790840251411755,
+      "loss": 0.5093,
+      "step": 10503
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018789003671086863,
+      "loss": 0.5969,
+      "step": 10504
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001878716703008676,
+      "loss": 0.4271,
+      "step": 10505
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018785330328440866,
+      "loss": 0.8851,
+      "step": 10506
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018783493566178582,
+      "loss": 0.3768,
+      "step": 10507
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001878165674332933,
+      "loss": 0.8389,
+      "step": 10508
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001877981985992252,
+      "loss": 0.371,
+      "step": 10509
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018777982915987567,
+      "loss": 0.3012,
+      "step": 10510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018776145911553887,
+      "loss": 0.2893,
+      "step": 10511
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187743088466509,
+      "loss": 0.0883,
+      "step": 10512
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018772471721308022,
+      "loss": 0.0033,
+      "step": 10513
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018770634535554667,
+      "loss": 0.5909,
+      "step": 10514
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001876879728942026,
+      "loss": 0.6488,
+      "step": 10515
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001876695998293422,
+      "loss": 0.4109,
+      "step": 10516
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018765122616125977,
+      "loss": 0.7351,
+      "step": 10517
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001876328518902494,
+      "loss": 0.189,
+      "step": 10518
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018761447701660534,
+      "loss": 0.1659,
+      "step": 10519
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018759610154062196,
+      "loss": 0.9186,
+      "step": 10520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001875777254625934,
+      "loss": 0.8754,
+      "step": 10521
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018755934878281398,
+      "loss": 0.0107,
+      "step": 10522
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001875409715015779,
+      "loss": 0.9599,
+      "step": 10523
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018752259361917958,
+      "loss": 0.0119,
+      "step": 10524
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001875042151359132,
+      "loss": 0.2527,
+      "step": 10525
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018748583605207312,
+      "loss": 0.0052,
+      "step": 10526
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001874674563679536,
+      "loss": 0.7072,
+      "step": 10527
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187449076083849,
+      "loss": 0.4635,
+      "step": 10528
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018743069520005368,
+      "loss": 0.0425,
+      "step": 10529
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001874123137168619,
+      "loss": 0.4206,
+      "step": 10530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001873939316345681,
+      "loss": 0.2341,
+      "step": 10531
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018737554895346666,
+      "loss": 0.0009,
+      "step": 10532
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001873571656738518,
+      "loss": 0.3379,
+      "step": 10533
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018733878179601803,
+      "loss": 0.367,
+      "step": 10534
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018732039732025977,
+      "loss": 0.0006,
+      "step": 10535
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018730201224687127,
+      "loss": 0.0113,
+      "step": 10536
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018728362657614706,
+      "loss": 0.4975,
+      "step": 10537
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018726524030838153,
+      "loss": 0.4556,
+      "step": 10538
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018724685344386909,
+      "loss": 0.3812,
+      "step": 10539
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001872284659829042,
+      "loss": 0.4407,
+      "step": 10540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018721007792578135,
+      "loss": 0.7213,
+      "step": 10541
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018719168927279484,
+      "loss": 0.0028,
+      "step": 10542
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018717330002423932,
+      "loss": 0.1945,
+      "step": 10543
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018715491018040918,
+      "loss": 0.2592,
+      "step": 10544
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018713651974159884,
+      "loss": 0.0027,
+      "step": 10545
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187118128708103,
+      "loss": 0.1516,
+      "step": 10546
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018709973708021596,
+      "loss": 2.8916,
+      "step": 10547
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001870813448582323,
+      "loss": 0.0031,
+      "step": 10548
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018706295204244658,
+      "loss": 0.2003,
+      "step": 10549
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001870445586331533,
+      "loss": 0.9783,
+      "step": 10550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018702616463064703,
+      "loss": 0.183,
+      "step": 10551
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001870077700352223,
+      "loss": 0.1779,
+      "step": 10552
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001869893748471736,
+      "loss": 0.4526,
+      "step": 10553
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018697097906679563,
+      "loss": 0.4096,
+      "step": 10554
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018695258269438297,
+      "loss": 0.459,
+      "step": 10555
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018693418573023011,
+      "loss": 0.4147,
+      "step": 10556
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018691578817463163,
+      "loss": 0.2902,
+      "step": 10557
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018689739002788227,
+      "loss": 0.6561,
+      "step": 10558
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001868789912902766,
+      "loss": 0.7207,
+      "step": 10559
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018686059196210922,
+      "loss": 0.0056,
+      "step": 10560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018684219204367477,
+      "loss": 0.9336,
+      "step": 10561
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018682379153526788,
+      "loss": 1.9099,
+      "step": 10562
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018680539043718327,
+      "loss": 0.007,
+      "step": 10563
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018678698874971555,
+      "loss": 0.2591,
+      "step": 10564
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018676858647315942,
+      "loss": 0.1164,
+      "step": 10565
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018675018360780958,
+      "loss": 3.8633,
+      "step": 10566
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001867317801539607,
+      "loss": 0.2245,
+      "step": 10567
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018671337611190743,
+      "loss": 0.4199,
+      "step": 10568
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001866949714819446,
+      "loss": 3.5283,
+      "step": 10569
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018667656626436687,
+      "loss": 0.278,
+      "step": 10570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018665816045946898,
+      "loss": 0.3626,
+      "step": 10571
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018663975406754566,
+      "loss": 0.2542,
+      "step": 10572
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018662134708889166,
+      "loss": 0.4173,
+      "step": 10573
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001866029395238017,
+      "loss": 2.4063,
+      "step": 10574
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001865845313725707,
+      "loss": 1.7482,
+      "step": 10575
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018656612263549322,
+      "loss": 0.4617,
+      "step": 10576
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018654771331286425,
+      "loss": 2.1735,
+      "step": 10577
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001865293034049785,
+      "loss": 0.5813,
+      "step": 10578
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018651089291213068,
+      "loss": 0.1385,
+      "step": 10579
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001864924818346158,
+      "loss": 0.5697,
+      "step": 10580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018647407017272858,
+      "loss": 0.2255,
+      "step": 10581
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018645565792676388,
+      "loss": 1.125,
+      "step": 10582
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018643724509701643,
+      "loss": 0.3084,
+      "step": 10583
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001864188316837813,
+      "loss": 0.3769,
+      "step": 10584
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018640041768735321,
+      "loss": 0.4686,
+      "step": 10585
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000186382003108027,
+      "loss": 0.3707,
+      "step": 10586
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001863635879460977,
+      "loss": 0.2301,
+      "step": 10587
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018634517220186004,
+      "loss": 0.2198,
+      "step": 10588
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018632675587560907,
+      "loss": 0.0026,
+      "step": 10589
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001863083389676396,
+      "loss": 0.4626,
+      "step": 10590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018628992147824648,
+      "loss": 2.1267,
+      "step": 10591
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018627150340772483,
+      "loss": 0.0051,
+      "step": 10592
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018625308475636945,
+      "loss": 0.3531,
+      "step": 10593
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018623466552447534,
+      "loss": 0.3249,
+      "step": 10594
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018621624571233745,
+      "loss": 0.051,
+      "step": 10595
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018619782532025076,
+      "loss": 0.3615,
+      "step": 10596
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018617940434851015,
+      "loss": 0.6083,
+      "step": 10597
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018616098279741072,
+      "loss": 0.0018,
+      "step": 10598
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018614256066724742,
+      "loss": 3.1205,
+      "step": 10599
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018612413795831526,
+      "loss": 0.7228,
+      "step": 10600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018610571467090925,
+      "loss": 0.0151,
+      "step": 10601
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018608729080532438,
+      "loss": 0.3236,
+      "step": 10602
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018606886636185572,
+      "loss": 0.0447,
+      "step": 10603
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018605044134079832,
+      "loss": 0.5644,
+      "step": 10604
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018603201574244718,
+      "loss": 0.0056,
+      "step": 10605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018601358956709735,
+      "loss": 0.1614,
+      "step": 10606
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000185995162815044,
+      "loss": 0.16,
+      "step": 10607
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001859767354865821,
+      "loss": 1.4092,
+      "step": 10608
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018595830758200678,
+      "loss": 0.1759,
+      "step": 10609
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018593987910161313,
+      "loss": 0.2785,
+      "step": 10610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018592145004569624,
+      "loss": 0.254,
+      "step": 10611
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018590302041455127,
+      "loss": 0.5629,
+      "step": 10612
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018588459020847328,
+      "loss": 0.0045,
+      "step": 10613
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018586615942775745,
+      "loss": 3.0879,
+      "step": 10614
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018584772807269892,
+      "loss": 0.7071,
+      "step": 10615
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018582929614359282,
+      "loss": 0.0092,
+      "step": 10616
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018581086364073427,
+      "loss": 0.4,
+      "step": 10617
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018579243056441854,
+      "loss": 0.0277,
+      "step": 10618
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018577399691494073,
+      "loss": 0.5485,
+      "step": 10619
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018575556269259606,
+      "loss": 1.4239,
+      "step": 10620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018573712789767967,
+      "loss": 0.6472,
+      "step": 10621
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018571869253048687,
+      "loss": 0.0261,
+      "step": 10622
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018570025659131277,
+      "loss": 0.6688,
+      "step": 10623
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018568182008045267,
+      "loss": 1.4305,
+      "step": 10624
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018566338299820176,
+      "loss": 0.3423,
+      "step": 10625
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018564494534485527,
+      "loss": 0.3138,
+      "step": 10626
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018562650712070853,
+      "loss": 0.0035,
+      "step": 10627
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001856080683260567,
+      "loss": 0.3091,
+      "step": 10628
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018558962896119508,
+      "loss": 0.8626,
+      "step": 10629
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018557118902641905,
+      "loss": 0.8495,
+      "step": 10630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018555274852202373,
+      "loss": 0.0089,
+      "step": 10631
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001855343074483045,
+      "loss": 0.8057,
+      "step": 10632
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018551586580555673,
+      "loss": 0.469,
+      "step": 10633
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018549742359407563,
+      "loss": 0.3355,
+      "step": 10634
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018547898081415658,
+      "loss": 0.6155,
+      "step": 10635
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018546053746609488,
+      "loss": 0.5495,
+      "step": 10636
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001854420935501859,
+      "loss": 0.0028,
+      "step": 10637
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018542364906672497,
+      "loss": 0.0925,
+      "step": 10638
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018540520401600754,
+      "loss": 1.7994,
+      "step": 10639
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001853867583983288,
+      "loss": 0.282,
+      "step": 10640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018536831221398428,
+      "loss": 1.3025,
+      "step": 10641
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018534986546326932,
+      "loss": 0.7554,
+      "step": 10642
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018533141814647928,
+      "loss": 0.8376,
+      "step": 10643
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018531297026390967,
+      "loss": 0.7186,
+      "step": 10644
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001852945218158558,
+      "loss": 0.2769,
+      "step": 10645
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018527607280261312,
+      "loss": 0.4045,
+      "step": 10646
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001852576232244771,
+      "loss": 0.2944,
+      "step": 10647
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018523917308174313,
+      "loss": 0.3982,
+      "step": 10648
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018522072237470673,
+      "loss": 0.7401,
+      "step": 10649
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018520227110366326,
+      "loss": 0.2874,
+      "step": 10650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001851838192689083,
+      "loss": 0.2049,
+      "step": 10651
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018516536687073726,
+      "loss": 0.622,
+      "step": 10652
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001851469139094457,
+      "loss": 0.3162,
+      "step": 10653
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018512846038532901,
+      "loss": 0.2163,
+      "step": 10654
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018511000629868273,
+      "loss": 0.1406,
+      "step": 10655
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018509155164980244,
+      "loss": 0.5123,
+      "step": 10656
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018507309643898357,
+      "loss": 0.6367,
+      "step": 10657
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018505464066652169,
+      "loss": 0.4199,
+      "step": 10658
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001850361843327124,
+      "loss": 0.5111,
+      "step": 10659
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001850177274378512,
+      "loss": 0.0042,
+      "step": 10660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001849992699822336,
+      "loss": 0.3312,
+      "step": 10661
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018498081196615524,
+      "loss": 0.4451,
+      "step": 10662
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018496235338991168,
+      "loss": 0.3256,
+      "step": 10663
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001849438942537985,
+      "loss": 0.8959,
+      "step": 10664
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018492543455811132,
+      "loss": 1.1673,
+      "step": 10665
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018490697430314568,
+      "loss": 0.4109,
+      "step": 10666
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018488851348919723,
+      "loss": 0.0156,
+      "step": 10667
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018487005211656165,
+      "loss": 0.327,
+      "step": 10668
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018485159018553449,
+      "loss": 0.7891,
+      "step": 10669
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001848331276964114,
+      "loss": 0.9547,
+      "step": 10670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018481466464948809,
+      "loss": 0.692,
+      "step": 10671
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018479620104506011,
+      "loss": 0.4258,
+      "step": 10672
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018477773688342324,
+      "loss": 0.2797,
+      "step": 10673
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018475927216487313,
+      "loss": 0.0059,
+      "step": 10674
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018474080688970538,
+      "loss": 0.1789,
+      "step": 10675
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001847223410582158,
+      "loss": 0.8278,
+      "step": 10676
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018470387467069997,
+      "loss": 1.5695,
+      "step": 10677
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018468540772745374,
+      "loss": 0.5646,
+      "step": 10678
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018466694022877274,
+      "loss": 0.3645,
+      "step": 10679
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001846484721749527,
+      "loss": 0.0168,
+      "step": 10680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001846300035662894,
+      "loss": 0.2965,
+      "step": 10681
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018461153440307856,
+      "loss": 0.1809,
+      "step": 10682
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018459306468561595,
+      "loss": 0.3388,
+      "step": 10683
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018457459441419732,
+      "loss": 0.3527,
+      "step": 10684
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018455612358911846,
+      "loss": 0.0025,
+      "step": 10685
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018453765221067511,
+      "loss": 0.6475,
+      "step": 10686
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001845191802791631,
+      "loss": 0.4168,
+      "step": 10687
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018450070779487828,
+      "loss": 0.052,
+      "step": 10688
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018448223475811635,
+      "loss": 0.0037,
+      "step": 10689
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018446376116917315,
+      "loss": 1.8087,
+      "step": 10690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001844452870283446,
+      "loss": 0.0297,
+      "step": 10691
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018442681233592646,
+      "loss": 0.4018,
+      "step": 10692
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018440833709221455,
+      "loss": 0.4724,
+      "step": 10693
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018438986129750483,
+      "loss": 1.0252,
+      "step": 10694
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018437138495209303,
+      "loss": 0.0044,
+      "step": 10695
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018435290805627512,
+      "loss": 0.2637,
+      "step": 10696
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018433443061034694,
+      "loss": 0.1883,
+      "step": 10697
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018431595261460436,
+      "loss": 0.2547,
+      "step": 10698
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018429747406934334,
+      "loss": 0.0057,
+      "step": 10699
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001842789949748597,
+      "loss": 0.442,
+      "step": 10700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018426051533144942,
+      "loss": 0.455,
+      "step": 10701
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018424203513940843,
+      "loss": 0.2865,
+      "step": 10702
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001842235543990326,
+      "loss": 0.4312,
+      "step": 10703
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001842050731106179,
+      "loss": 0.985,
+      "step": 10704
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018418659127446036,
+      "loss": 0.0024,
+      "step": 10705
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001841681088908558,
+      "loss": 0.4865,
+      "step": 10706
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018414962596010023,
+      "loss": 0.5605,
+      "step": 10707
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018413114248248973,
+      "loss": 1.2147,
+      "step": 10708
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018411265845832018,
+      "loss": 0.9312,
+      "step": 10709
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001840941738878876,
+      "loss": 0.5362,
+      "step": 10710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018407568877148797,
+      "loss": 0.653,
+      "step": 10711
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018405720310941736,
+      "loss": 0.5237,
+      "step": 10712
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001840387169019717,
+      "loss": 0.6041,
+      "step": 10713
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018402023014944718,
+      "loss": 0.0007,
+      "step": 10714
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001840017428521396,
+      "loss": 0.0479,
+      "step": 10715
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001839832550103452,
+      "loss": 0.9631,
+      "step": 10716
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018396476662436,
+      "loss": 0.3657,
+      "step": 10717
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018394627769448003,
+      "loss": 0.216,
+      "step": 10718
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018392778822100132,
+      "loss": 0.4847,
+      "step": 10719
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018390929820422003,
+      "loss": 0.4603,
+      "step": 10720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001838908076444322,
+      "loss": 1.005,
+      "step": 10721
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018387231654193399,
+      "loss": 0.4789,
+      "step": 10722
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018385382489702142,
+      "loss": 0.2194,
+      "step": 10723
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018383533270999067,
+      "loss": 0.1151,
+      "step": 10724
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018381683998113787,
+      "loss": 0.0048,
+      "step": 10725
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018379834671075908,
+      "loss": 0.0098,
+      "step": 10726
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018377985289915052,
+      "loss": 0.8799,
+      "step": 10727
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018376135854660837,
+      "loss": 0.3914,
+      "step": 10728
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018374286365342864,
+      "loss": 0.0288,
+      "step": 10729
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018372436821990763,
+      "loss": 1.131,
+      "step": 10730
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001837058722463415,
+      "loss": 0.1187,
+      "step": 10731
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018368737573302638,
+      "loss": 0.3371,
+      "step": 10732
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018366887868025847,
+      "loss": 0.004,
+      "step": 10733
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018365038108833403,
+      "loss": 0.2573,
+      "step": 10734
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018363188295754922,
+      "loss": 0.6128,
+      "step": 10735
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001836133842882003,
+      "loss": 0.3706,
+      "step": 10736
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001835948850805835,
+      "loss": 0.186,
+      "step": 10737
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018357638533499497,
+      "loss": 0.0202,
+      "step": 10738
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018355788505173106,
+      "loss": 0.55,
+      "step": 10739
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000183539384231088,
+      "loss": 0.0115,
+      "step": 10740
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018352088287336196,
+      "loss": 0.2517,
+      "step": 10741
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018350238097884937,
+      "loss": 0.8364,
+      "step": 10742
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018348387854784644,
+      "loss": 3.3419,
+      "step": 10743
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001834653755806494,
+      "loss": 0.002,
+      "step": 10744
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001834468720775546,
+      "loss": 0.1922,
+      "step": 10745
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018342836803885833,
+      "loss": 0.5209,
+      "step": 10746
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018340986346485695,
+      "loss": 0.258,
+      "step": 10747
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001833913583558467,
+      "loss": 0.2111,
+      "step": 10748
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000183372852712124,
+      "loss": 1.0787,
+      "step": 10749
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018335434653398515,
+      "loss": 0.5032,
+      "step": 10750
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018333583982172648,
+      "loss": 0.0077,
+      "step": 10751
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018331733257564439,
+      "loss": 0.0215,
+      "step": 10752
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018329882479603516,
+      "loss": 0.4099,
+      "step": 10753
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001832803164831953,
+      "loss": 2.4765,
+      "step": 10754
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001832618076374211,
+      "loss": 0.4213,
+      "step": 10755
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001832432982590089,
+      "loss": 0.0122,
+      "step": 10756
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018322478834825525,
+      "loss": 0.0076,
+      "step": 10757
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018320627790545648,
+      "loss": 0.4293,
+      "step": 10758
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018318776693090896,
+      "loss": 0.6815,
+      "step": 10759
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018316925542490918,
+      "loss": 0.0064,
+      "step": 10760
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018315074338775354,
+      "loss": 0.8434,
+      "step": 10761
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018313223081973853,
+      "loss": 0.4717,
+      "step": 10762
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018311371772116055,
+      "loss": 0.3856,
+      "step": 10763
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018309520409231605,
+      "loss": 0.4912,
+      "step": 10764
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018307668993350153,
+      "loss": 0.4125,
+      "step": 10765
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018305817524501354,
+      "loss": 0.8076,
+      "step": 10766
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018303966002714844,
+      "loss": 0.0021,
+      "step": 10767
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001830211442802027,
+      "loss": 0.0018,
+      "step": 10768
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018300262800447298,
+      "loss": 0.5674,
+      "step": 10769
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018298411120025568,
+      "loss": 0.4401,
+      "step": 10770
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018296559386784736,
+      "loss": 1.1087,
+      "step": 10771
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018294707600754445,
+      "loss": 0.9873,
+      "step": 10772
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018292855761964364,
+      "loss": 0.3999,
+      "step": 10773
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018291003870444137,
+      "loss": 0.524,
+      "step": 10774
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018289151926223425,
+      "loss": 0.0106,
+      "step": 10775
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018287299929331877,
+      "loss": 0.6149,
+      "step": 10776
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018285447879799158,
+      "loss": 0.9193,
+      "step": 10777
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001828359577765492,
+      "loss": 0.3491,
+      "step": 10778
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001828174362292882,
+      "loss": 0.6025,
+      "step": 10779
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827989141565053,
+      "loss": 0.009,
+      "step": 10780
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018278039155849695,
+      "loss": 0.0022,
+      "step": 10781
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018276186843555985,
+      "loss": 0.7667,
+      "step": 10782
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018274334478799058,
+      "loss": 0.55,
+      "step": 10783
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827248206160858,
+      "loss": 0.4447,
+      "step": 10784
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827062959201421,
+      "loss": 2.4298,
+      "step": 10785
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001826877707004562,
+      "loss": 0.1942,
+      "step": 10786
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018266924495732466,
+      "loss": 0.1696,
+      "step": 10787
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001826507186910442,
+      "loss": 0.0151,
+      "step": 10788
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018263219190191153,
+      "loss": 0.0108,
+      "step": 10789
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018261366459022322,
+      "loss": 2.5275,
+      "step": 10790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018259513675627606,
+      "loss": 0.6812,
+      "step": 10791
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018257660840036673,
+      "loss": 0.8934,
+      "step": 10792
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018255807952279184,
+      "loss": 0.672,
+      "step": 10793
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018253955012384818,
+      "loss": 0.5117,
+      "step": 10794
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018252102020383247,
+      "loss": 0.5477,
+      "step": 10795
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018250248976304148,
+      "loss": 2.0269,
+      "step": 10796
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018248395880177183,
+      "loss": 0.5329,
+      "step": 10797
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018246542732032033,
+      "loss": 1.3726,
+      "step": 10798
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018244689531898375,
+      "loss": 1.0518,
+      "step": 10799
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018242836279805887,
+      "loss": 0.3746,
+      "step": 10800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018240982975784245,
+      "loss": 1.0948,
+      "step": 10801
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018239129619863114,
+      "loss": 0.0043,
+      "step": 10802
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001823727621207219,
+      "loss": 0.6276,
+      "step": 10803
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001823542275244115,
+      "loss": 0.8442,
+      "step": 10804
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001823356924099966,
+      "loss": 0.0102,
+      "step": 10805
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001823171567777742,
+      "loss": 0.4997,
+      "step": 10806
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000182298620628041,
+      "loss": 1.0753,
+      "step": 10807
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018228008396109388,
+      "loss": 0.1662,
+      "step": 10808
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018226154677722964,
+      "loss": 0.411,
+      "step": 10809
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018224300907674514,
+      "loss": 0.5677,
+      "step": 10810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018222447085993726,
+      "loss": 3.1262,
+      "step": 10811
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018220593212710286,
+      "loss": 1.5185,
+      "step": 10812
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018218739287853875,
+      "loss": 0.1624,
+      "step": 10813
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018216885311454187,
+      "loss": 0.3098,
+      "step": 10814
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018215031283540912,
+      "loss": 0.0166,
+      "step": 10815
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018213177204143733,
+      "loss": 0.0747,
+      "step": 10816
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001821132307329234,
+      "loss": 1.6896,
+      "step": 10817
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018209468891016432,
+      "loss": 1.4595,
+      "step": 10818
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018207614657345697,
+      "loss": 0.1977,
+      "step": 10819
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018205760372309824,
+      "loss": 0.619,
+      "step": 10820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001820390603593851,
+      "loss": 0.8511,
+      "step": 10821
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018202051648261448,
+      "loss": 0.2785,
+      "step": 10822
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001820019720930834,
+      "loss": 0.0912,
+      "step": 10823
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018198342719108872,
+      "loss": 0.2873,
+      "step": 10824
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018196488177692747,
+      "loss": 1.0418,
+      "step": 10825
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001819463358508966,
+      "loss": 0.4548,
+      "step": 10826
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018192778941329314,
+      "loss": 0.0113,
+      "step": 10827
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000181909242464414,
+      "loss": 0.1559,
+      "step": 10828
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018189069500455627,
+      "loss": 0.5449,
+      "step": 10829
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001818721470340169,
+      "loss": 0.4356,
+      "step": 10830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018185359855309296,
+      "loss": 0.1311,
+      "step": 10831
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001818350495620814,
+      "loss": 0.2774,
+      "step": 10832
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001818165000612793,
+      "loss": 0.0089,
+      "step": 10833
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018179795005098375,
+      "loss": 0.3987,
+      "step": 10834
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001817793995314917,
+      "loss": 0.301,
+      "step": 10835
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018176084850310027,
+      "loss": 0.3872,
+      "step": 10836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018174229696610653,
+      "loss": 0.4284,
+      "step": 10837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018172374492080758,
+      "loss": 0.0145,
+      "step": 10838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001817051923675004,
+      "loss": 1.3412,
+      "step": 10839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018168663930648215,
+      "loss": 0.0458,
+      "step": 10840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018166808573804997,
+      "loss": 0.2372,
+      "step": 10841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018164953166250093,
+      "loss": 0.3534,
+      "step": 10842
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018163097708013206,
+      "loss": 0.2705,
+      "step": 10843
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018161242199124067,
+      "loss": 0.635,
+      "step": 10844
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001815938663961237,
+      "loss": 0.6122,
+      "step": 10845
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018157531029507843,
+      "loss": 0.0147,
+      "step": 10846
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018155675368840193,
+      "loss": 0.4746,
+      "step": 10847
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018153819657639136,
+      "loss": 0.628,
+      "step": 10848
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018151963895934393,
+      "loss": 0.7033,
+      "step": 10849
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018150108083755685,
+      "loss": 0.3552,
+      "step": 10850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001814825222113271,
+      "loss": 0.344,
+      "step": 10851
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018146396308095204,
+      "loss": 0.9538,
+      "step": 10852
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018144540344672895,
+      "loss": 0.4069,
+      "step": 10853
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001814268433089548,
+      "loss": 0.0104,
+      "step": 10854
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018140828266792697,
+      "loss": 0.6231,
+      "step": 10855
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001813897215239426,
+      "loss": 0.017,
+      "step": 10856
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018137115987729898,
+      "loss": 0.4698,
+      "step": 10857
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018135259772829326,
+      "loss": 0.2732,
+      "step": 10858
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018133403507722282,
+      "loss": 0.5847,
+      "step": 10859
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018131547192438478,
+      "loss": 1.0674,
+      "step": 10860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018129690827007645,
+      "loss": 0.3661,
+      "step": 10861
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018127834411459513,
+      "loss": 0.8159,
+      "step": 10862
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018125977945823805,
+      "loss": 0.3931,
+      "step": 10863
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001812412143013026,
+      "loss": 0.2329,
+      "step": 10864
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001812226486440859,
+      "loss": 0.3406,
+      "step": 10865
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018120408248688532,
+      "loss": 1.0667,
+      "step": 10866
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018118551582999826,
+      "loss": 0.3828,
+      "step": 10867
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018116694867372193,
+      "loss": 0.2027,
+      "step": 10868
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001811483810183537,
+      "loss": 0.6642,
+      "step": 10869
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001811298128641909,
+      "loss": 0.312,
+      "step": 10870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018111124421153085,
+      "loss": 0.4434,
+      "step": 10871
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018109267506067093,
+      "loss": 0.089,
+      "step": 10872
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018107410541190846,
+      "loss": 0.5127,
+      "step": 10873
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018105553526554086,
+      "loss": 0.5112,
+      "step": 10874
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018103696462186546,
+      "loss": 0.3306,
+      "step": 10875
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018101839348117965,
+      "loss": 0.4577,
+      "step": 10876
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001809998218437808,
+      "loss": 0.2261,
+      "step": 10877
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018098124970996634,
+      "loss": 0.0024,
+      "step": 10878
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001809626770800337,
+      "loss": 2.7641,
+      "step": 10879
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001809441039542802,
+      "loss": 0.2122,
+      "step": 10880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018092553033300336,
+      "loss": 0.3658,
+      "step": 10881
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018090695621650053,
+      "loss": 0.3658,
+      "step": 10882
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018088838160506922,
+      "loss": 0.5187,
+      "step": 10883
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001808698064990068,
+      "loss": 1.6924,
+      "step": 10884
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018085123089861074,
+      "loss": 0.1985,
+      "step": 10885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018083265480417854,
+      "loss": 0.705,
+      "step": 10886
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018081407821600768,
+      "loss": 0.0063,
+      "step": 10887
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018079550113439557,
+      "loss": 0.1478,
+      "step": 10888
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018077692355963974,
+      "loss": 0.8262,
+      "step": 10889
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001807583454920377,
+      "loss": 0.9108,
+      "step": 10890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018073976693188686,
+      "loss": 0.2542,
+      "step": 10891
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018072118787948476,
+      "loss": 0.1891,
+      "step": 10892
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018070260833512902,
+      "loss": 0.6029,
+      "step": 10893
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018068402829911706,
+      "loss": 0.1446,
+      "step": 10894
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018066544777174643,
+      "loss": 0.0862,
+      "step": 10895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018064686675331467,
+      "loss": 0.4348,
+      "step": 10896
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018062828524411935,
+      "loss": 0.0043,
+      "step": 10897
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000180609703244458,
+      "loss": 0.5377,
+      "step": 10898
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001805911207546282,
+      "loss": 1.905,
+      "step": 10899
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018057253777492747,
+      "loss": 0.2561,
+      "step": 10900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018055395430565348,
+      "loss": 2.3246,
+      "step": 10901
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018053537034710377,
+      "loss": 0.0475,
+      "step": 10902
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001805167858995759,
+      "loss": 0.986,
+      "step": 10903
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001804982009633675,
+      "loss": 0.0083,
+      "step": 10904
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018047961553877624,
+      "loss": 0.2544,
+      "step": 10905
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001804610296260996,
+      "loss": 0.0071,
+      "step": 10906
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018044244322563534,
+      "loss": 0.0027,
+      "step": 10907
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018042385633768103,
+      "loss": 0.5346,
+      "step": 10908
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001804052689625343,
+      "loss": 1.2685,
+      "step": 10909
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018038668110049284,
+      "loss": 0.2247,
+      "step": 10910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001803680927518543,
+      "loss": 0.2753,
+      "step": 10911
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018034950391691628,
+      "loss": 0.6494,
+      "step": 10912
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018033091459597654,
+      "loss": 0.1754,
+      "step": 10913
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018031232478933276,
+      "loss": 0.3959,
+      "step": 10914
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018029373449728248,
+      "loss": 0.3919,
+      "step": 10915
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018027514372012362,
+      "loss": 0.0019,
+      "step": 10916
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018025655245815372,
+      "loss": 0.2555,
+      "step": 10917
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018023796071167054,
+      "loss": 0.3928,
+      "step": 10918
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018021936848097183,
+      "loss": 0.2813,
+      "step": 10919
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018020077576635525,
+      "loss": 0.4815,
+      "step": 10920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018018218256811856,
+      "loss": 2.5258,
+      "step": 10921
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018016358888655953,
+      "loss": 0.1242,
+      "step": 10922
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001801449947219759,
+      "loss": 0.2859,
+      "step": 10923
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018012640007466544,
+      "loss": 0.3337,
+      "step": 10924
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001801078049449259,
+      "loss": 0.0026,
+      "step": 10925
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000180089209333055,
+      "loss": 0.2137,
+      "step": 10926
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018007061323935058,
+      "loss": 0.6979,
+      "step": 10927
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018005201666411048,
+      "loss": 0.6927,
+      "step": 10928
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018003341960763237,
+      "loss": 1.5199,
+      "step": 10929
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001800148220702141,
+      "loss": 0.8671,
+      "step": 10930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017999622405215357,
+      "loss": 1.7826,
+      "step": 10931
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001799776255537485,
+      "loss": 0.2686,
+      "step": 10932
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017995902657529675,
+      "loss": 0.401,
+      "step": 10933
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017994042711709616,
+      "loss": 0.3456,
+      "step": 10934
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017992182717944453,
+      "loss": 0.8192,
+      "step": 10935
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017990322676263978,
+      "loss": 0.585,
+      "step": 10936
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001798846258669797,
+      "loss": 0.615,
+      "step": 10937
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017986602449276222,
+      "loss": 0.4986,
+      "step": 10938
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017984742264028517,
+      "loss": 0.0318,
+      "step": 10939
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017982882030984643,
+      "loss": 0.4296,
+      "step": 10940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017981021750174388,
+      "loss": 0.092,
+      "step": 10941
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017979161421627547,
+      "loss": 0.599,
+      "step": 10942
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001797730104537391,
+      "loss": 0.4,
+      "step": 10943
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001797544062144326,
+      "loss": 0.5289,
+      "step": 10944
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017973580149865394,
+      "loss": 0.3373,
+      "step": 10945
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017971719630670104,
+      "loss": 0.003,
+      "step": 10946
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017969859063887188,
+      "loss": 0.0081,
+      "step": 10947
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017967998449546437,
+      "loss": 0.64,
+      "step": 10948
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017966137787677638,
+      "loss": 0.849,
+      "step": 10949
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017964277078310597,
+      "loss": 0.1841,
+      "step": 10950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017962416321475112,
+      "loss": 0.3383,
+      "step": 10951
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017960555517200968,
+      "loss": 0.0035,
+      "step": 10952
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017958694665517975,
+      "loss": 0.5361,
+      "step": 10953
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001795683376645593,
+      "loss": 0.91,
+      "step": 10954
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017954972820044627,
+      "loss": 1.1063,
+      "step": 10955
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017953111826313865,
+      "loss": 0.0061,
+      "step": 10956
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017951250785293456,
+      "loss": 0.0015,
+      "step": 10957
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017949389697013193,
+      "loss": 0.319,
+      "step": 10958
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017947528561502876,
+      "loss": 0.8697,
+      "step": 10959
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017945667378792317,
+      "loss": 0.297,
+      "step": 10960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017943806148911312,
+      "loss": 0.3234,
+      "step": 10961
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017941944871889675,
+      "loss": 3.3981,
+      "step": 10962
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017940083547757202,
+      "loss": 0.2697,
+      "step": 10963
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017938222176543705,
+      "loss": 0.9823,
+      "step": 10964
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017936360758278992,
+      "loss": 0.2661,
+      "step": 10965
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017934499292992866,
+      "loss": 0.1019,
+      "step": 10966
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017932637780715136,
+      "loss": 0.2315,
+      "step": 10967
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017930776221475615,
+      "loss": 0.0063,
+      "step": 10968
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001792891461530411,
+      "loss": 0.0106,
+      "step": 10969
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017927052962230433,
+      "loss": 0.3966,
+      "step": 10970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017925191262284395,
+      "loss": 0.0362,
+      "step": 10971
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017923329515495812,
+      "loss": 0.1611,
+      "step": 10972
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017921467721894493,
+      "loss": 0.3981,
+      "step": 10973
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017919605881510257,
+      "loss": 0.3752,
+      "step": 10974
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017917743994372905,
+      "loss": 0.3156,
+      "step": 10975
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017915882060512268,
+      "loss": 1.1382,
+      "step": 10976
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017914020079958158,
+      "loss": 0.7768,
+      "step": 10977
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017912158052740387,
+      "loss": 0.1754,
+      "step": 10978
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001791029597888877,
+      "loss": 1.9818,
+      "step": 10979
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001790843385843314,
+      "loss": 0.5966,
+      "step": 10980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017906571691403303,
+      "loss": 1.1225,
+      "step": 10981
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017904709477829082,
+      "loss": 2.0323,
+      "step": 10982
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000179028472177403,
+      "loss": 0.4101,
+      "step": 10983
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017900984911166772,
+      "loss": 0.0043,
+      "step": 10984
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017899122558138328,
+      "loss": 0.0016,
+      "step": 10985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017897260158684784,
+      "loss": 1.0115,
+      "step": 10986
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017895397712835968,
+      "loss": 0.1013,
+      "step": 10987
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017893535220621707,
+      "loss": 0.9787,
+      "step": 10988
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001789167268207182,
+      "loss": 0.0009,
+      "step": 10989
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017889810097216128,
+      "loss": 0.4557,
+      "step": 10990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017887947466084476,
+      "loss": 0.7611,
+      "step": 10991
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017886084788706671,
+      "loss": 0.9258,
+      "step": 10992
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001788422206511255,
+      "loss": 0.475,
+      "step": 10993
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017882359295331947,
+      "loss": 2.7423,
+      "step": 10994
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001788049647939468,
+      "loss": 0.6919,
+      "step": 10995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001787863361733059,
+      "loss": 0.2783,
+      "step": 10996
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017876770709169496,
+      "loss": 0.2234,
+      "step": 10997
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001787490775494124,
+      "loss": 0.5803,
+      "step": 10998
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001787304475467565,
+      "loss": 0.2881,
+      "step": 10999
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017871181708402567,
+      "loss": 2.6667,
+      "step": 11000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017869318616151812,
+      "loss": 0.5855,
+      "step": 11001
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017867455477953219,
+      "loss": 0.719,
+      "step": 11002
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017865592293836641,
+      "loss": 0.3094,
+      "step": 11003
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017863729063831898,
+      "loss": 0.3327,
+      "step": 11004
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001786186578796883,
+      "loss": 0.7458,
+      "step": 11005
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017860002466277283,
+      "loss": 0.9742,
+      "step": 11006
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017858139098787085,
+      "loss": 0.0015,
+      "step": 11007
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017856275685528076,
+      "loss": 0.0542,
+      "step": 11008
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000178544122265301,
+      "loss": 0.3753,
+      "step": 11009
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017852548721823,
+      "loss": 0.3924,
+      "step": 11010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017850685171436612,
+      "loss": 0.0045,
+      "step": 11011
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001784882157540078,
+      "loss": 0.2409,
+      "step": 11012
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017846957933745342,
+      "loss": 1.1514,
+      "step": 11013
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017845094246500147,
+      "loss": 0.1213,
+      "step": 11014
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017843230513695042,
+      "loss": 0.5458,
+      "step": 11015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017841366735359859,
+      "loss": 0.4614,
+      "step": 11016
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017839502911524456,
+      "loss": 0.3121,
+      "step": 11017
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017837639042218677,
+      "loss": 0.4273,
+      "step": 11018
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017835775127472367,
+      "loss": 0.5702,
+      "step": 11019
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001783391116731537,
+      "loss": 1.3389,
+      "step": 11020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017832047161777542,
+      "loss": 0.0369,
+      "step": 11021
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017830183110888732,
+      "loss": 0.2963,
+      "step": 11022
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017828319014678786,
+      "loss": 0.2738,
+      "step": 11023
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001782645487317755,
+      "loss": 0.0013,
+      "step": 11024
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017824590686414885,
+      "loss": 0.0063,
+      "step": 11025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017822726454420643,
+      "loss": 0.1965,
+      "step": 11026
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017820862177224668,
+      "loss": 0.7163,
+      "step": 11027
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017818997854856817,
+      "loss": 0.2557,
+      "step": 11028
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001781713348734695,
+      "loss": 0.0085,
+      "step": 11029
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001781526907472492,
+      "loss": 0.2682,
+      "step": 11030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017813404617020577,
+      "loss": 0.7913,
+      "step": 11031
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017811540114263784,
+      "loss": 0.3647,
+      "step": 11032
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017809675566484391,
+      "loss": 0.7586,
+      "step": 11033
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017807810973712265,
+      "loss": 1.1192,
+      "step": 11034
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017805946335977256,
+      "loss": 0.2349,
+      "step": 11035
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017804081653309233,
+      "loss": 1.1887,
+      "step": 11036
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017802216925738047,
+      "loss": 0.0037,
+      "step": 11037
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001780035215329357,
+      "loss": 0.4288,
+      "step": 11038
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017798487336005643,
+      "loss": 0.037,
+      "step": 11039
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017796622473904147,
+      "loss": 0.33,
+      "step": 11040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017794757567018944,
+      "loss": 0.5806,
+      "step": 11041
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017792892615379888,
+      "loss": 0.0025,
+      "step": 11042
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001779102761901685,
+      "loss": 0.0543,
+      "step": 11043
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017789162577959693,
+      "loss": 0.2795,
+      "step": 11044
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017787297492238282,
+      "loss": 0.8237,
+      "step": 11045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017785432361882485,
+      "loss": 0.7962,
+      "step": 11046
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001778356718692217,
+      "loss": 0.5622,
+      "step": 11047
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017781701967387202,
+      "loss": 0.8073,
+      "step": 11048
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017779836703307457,
+      "loss": 0.5252,
+      "step": 11049
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017777971394712796,
+      "loss": 0.4283,
+      "step": 11050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001777610604163309,
+      "loss": 0.6623,
+      "step": 11051
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017774240644098218,
+      "loss": 0.4463,
+      "step": 11052
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001777237520213804,
+      "loss": 0.8834,
+      "step": 11053
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017770509715782433,
+      "loss": 0.0058,
+      "step": 11054
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017768644185061277,
+      "loss": 0.5278,
+      "step": 11055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017766778610004433,
+      "loss": 0.1841,
+      "step": 11056
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001776491299064179,
+      "loss": 1.0174,
+      "step": 11057
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017763047327003207,
+      "loss": 0.4596,
+      "step": 11058
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017761181619118573,
+      "loss": 0.778,
+      "step": 11059
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017759315867017756,
+      "loss": 0.6079,
+      "step": 11060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017757450070730638,
+      "loss": 1.3788,
+      "step": 11061
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017755584230287093,
+      "loss": 0.0076,
+      "step": 11062
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017753718345717005,
+      "loss": 0.3981,
+      "step": 11063
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001775185241705025,
+      "loss": 2.2091,
+      "step": 11064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001774998644431671,
+      "loss": 1.3359,
+      "step": 11065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001774812042754626,
+      "loss": 0.5198,
+      "step": 11066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001774625436676879,
+      "loss": 0.4766,
+      "step": 11067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017744388262014174,
+      "loss": 0.3429,
+      "step": 11068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000177425221133123,
+      "loss": 0.0044,
+      "step": 11069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017740655920693053,
+      "loss": 0.1446,
+      "step": 11070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017738789684186313,
+      "loss": 0.3349,
+      "step": 11071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017736923403821963,
+      "loss": 0.5578,
+      "step": 11072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017735057079629895,
+      "loss": 0.0395,
+      "step": 11073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017733190711639995,
+      "loss": 0.3744,
+      "step": 11074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001773132429988215,
+      "loss": 0.0035,
+      "step": 11075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001772945784438624,
+      "loss": 0.2175,
+      "step": 11076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017727591345182157,
+      "loss": 1.7577,
+      "step": 11077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017725724802299797,
+      "loss": 0.4702,
+      "step": 11078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017723858215769043,
+      "loss": 2.0156,
+      "step": 11079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001772199158561979,
+      "loss": 0.8322,
+      "step": 11080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017720124911881923,
+      "loss": 0.4877,
+      "step": 11081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017718258194585342,
+      "loss": 0.643,
+      "step": 11082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017716391433759932,
+      "loss": 0.0034,
+      "step": 11083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001771452462943559,
+      "loss": 0.3107,
+      "step": 11084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017712657781642212,
+      "loss": 0.671,
+      "step": 11085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001771079089040969,
+      "loss": 0.4273,
+      "step": 11086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001770892395576792,
+      "loss": 0.7485,
+      "step": 11087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017707056977746797,
+      "loss": 0.0649,
+      "step": 11088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001770518995637622,
+      "loss": 0.0843,
+      "step": 11089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001770332289168609,
+      "loss": 0.4912,
+      "step": 11090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017701455783706297,
+      "loss": 0.0042,
+      "step": 11091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017699588632466744,
+      "loss": 0.2811,
+      "step": 11092
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001769772143799733,
+      "loss": 0.6801,
+      "step": 11093
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001769585420032796,
+      "loss": 0.0033,
+      "step": 11094
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017693986919488524,
+      "loss": 0.4963,
+      "step": 11095
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017692119595508934,
+      "loss": 2.0785,
+      "step": 11096
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001769025222841909,
+      "loss": 0.6176,
+      "step": 11097
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017688384818248895,
+      "loss": 0.3136,
+      "step": 11098
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017686517365028256,
+      "loss": 0.4483,
+      "step": 11099
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001768464986878706,
+      "loss": 0.2043,
+      "step": 11100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017682782329555237,
+      "loss": 0.5281,
+      "step": 11101
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017680914747362676,
+      "loss": 0.0041,
+      "step": 11102
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017679047122239286,
+      "loss": 0.1001,
+      "step": 11103
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017677179454214984,
+      "loss": 0.5336,
+      "step": 11104
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001767531174331967,
+      "loss": 0.0026,
+      "step": 11105
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001767344398958325,
+      "loss": 0.3789,
+      "step": 11106
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017671576193035638,
+      "loss": 2.4629,
+      "step": 11107
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017669708353706745,
+      "loss": 0.6677,
+      "step": 11108
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017667840471626478,
+      "loss": 0.0018,
+      "step": 11109
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001766597254682475,
+      "loss": 0.1873,
+      "step": 11110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001766410457933147,
+      "loss": 0.0025,
+      "step": 11111
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017662236569176553,
+      "loss": 0.6184,
+      "step": 11112
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001766036851638992,
+      "loss": 0.4008,
+      "step": 11113
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017658500421001473,
+      "loss": 0.8572,
+      "step": 11114
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017656632283041126,
+      "loss": 0.4685,
+      "step": 11115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017654764102538811,
+      "loss": 0.4352,
+      "step": 11116
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017652895879524428,
+      "loss": 0.3993,
+      "step": 11117
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017651027614027897,
+      "loss": 0.1686,
+      "step": 11118
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017649159306079137,
+      "loss": 1.3772,
+      "step": 11119
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001764729095570807,
+      "loss": 0.0028,
+      "step": 11120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017645422562944608,
+      "loss": 0.3916,
+      "step": 11121
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017643554127818678,
+      "loss": 0.2238,
+      "step": 11122
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017641685650360194,
+      "loss": 0.095,
+      "step": 11123
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001763981713059908,
+      "loss": 1.0231,
+      "step": 11124
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017637948568565256,
+      "loss": 0.0046,
+      "step": 11125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001763607996428864,
+      "loss": 0.2736,
+      "step": 11126
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017634211317799168,
+      "loss": 0.8612,
+      "step": 11127
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017632342629126752,
+      "loss": 1.0108,
+      "step": 11128
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017630473898301317,
+      "loss": 0.6823,
+      "step": 11129
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017628605125352792,
+      "loss": 0.6202,
+      "step": 11130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017626736310311097,
+      "loss": 0.2833,
+      "step": 11131
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001762486745320617,
+      "loss": 1.8417,
+      "step": 11132
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017622998554067923,
+      "loss": 0.3999,
+      "step": 11133
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017621129612926294,
+      "loss": 0.3438,
+      "step": 11134
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017619260629811207,
+      "loss": 0.2538,
+      "step": 11135
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017617391604752596,
+      "loss": 0.7775,
+      "step": 11136
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001761552253778038,
+      "loss": 0.6106,
+      "step": 11137
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000176136534289245,
+      "loss": 0.6111,
+      "step": 11138
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017611784278214883,
+      "loss": 0.3612,
+      "step": 11139
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017609915085681457,
+      "loss": 0.198,
+      "step": 11140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017608045851354157,
+      "loss": 0.7212,
+      "step": 11141
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017606176575262923,
+      "loss": 0.1399,
+      "step": 11142
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017604307257437681,
+      "loss": 0.3208,
+      "step": 11143
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017602437897908365,
+      "loss": 0.429,
+      "step": 11144
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001760056849670491,
+      "loss": 0.9761,
+      "step": 11145
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017598699053857257,
+      "loss": 1.1709,
+      "step": 11146
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017596829569395335,
+      "loss": 0.9104,
+      "step": 11147
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001759496004334909,
+      "loss": 0.1859,
+      "step": 11148
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017593090475748449,
+      "loss": 0.0008,
+      "step": 11149
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017591220866623356,
+      "loss": 0.5564,
+      "step": 11150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017589351216003757,
+      "loss": 0.5463,
+      "step": 11151
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017587481523919575,
+      "loss": 0.6097,
+      "step": 11152
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017585611790400766,
+      "loss": 0.0014,
+      "step": 11153
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017583742015477263,
+      "loss": 0.5151,
+      "step": 11154
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001758187219917901,
+      "loss": 0.4997,
+      "step": 11155
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017580002341535946,
+      "loss": 2.709,
+      "step": 11156
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017578132442578015,
+      "loss": 0.2454,
+      "step": 11157
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017576262502335167,
+      "loss": 0.4615,
+      "step": 11158
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017574392520837342,
+      "loss": 0.4452,
+      "step": 11159
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001757252249811448,
+      "loss": 1.1819,
+      "step": 11160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001757065243419653,
+      "loss": 0.2673,
+      "step": 11161
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017568782329113446,
+      "loss": 0.0086,
+      "step": 11162
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017566912182895165,
+      "loss": 0.9744,
+      "step": 11163
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017565041995571632,
+      "loss": 0.3296,
+      "step": 11164
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001756317176717281,
+      "loss": 0.7302,
+      "step": 11165
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017561301497728637,
+      "loss": 0.2883,
+      "step": 11166
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017559431187269057,
+      "loss": 0.3449,
+      "step": 11167
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017557560835824036,
+      "loss": 0.3929,
+      "step": 11168
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017555690443423516,
+      "loss": 0.5214,
+      "step": 11169
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017553820010097446,
+      "loss": 0.1134,
+      "step": 11170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017551949535875782,
+      "loss": 0.4971,
+      "step": 11171
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017550079020788473,
+      "loss": 0.2857,
+      "step": 11172
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001754820846486548,
+      "loss": 0.2505,
+      "step": 11173
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017546337868136756,
+      "loss": 0.6878,
+      "step": 11174
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017544467230632243,
+      "loss": 4.5127,
+      "step": 11175
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017542596552381915,
+      "loss": 0.7272,
+      "step": 11176
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017540725833415716,
+      "loss": 0.3914,
+      "step": 11177
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017538855073763603,
+      "loss": 0.605,
+      "step": 11178
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017536984273455538,
+      "loss": 0.4873,
+      "step": 11179
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001753511343252148,
+      "loss": 0.5667,
+      "step": 11180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001753324255099138,
+      "loss": 0.9228,
+      "step": 11181
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017531371628895205,
+      "loss": 0.3135,
+      "step": 11182
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017529500666262914,
+      "loss": 0.8742,
+      "step": 11183
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017527629663124464,
+      "loss": 0.8858,
+      "step": 11184
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017525758619509825,
+      "loss": 0.1255,
+      "step": 11185
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017523887535448942,
+      "loss": 0.98,
+      "step": 11186
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017522016410971791,
+      "loss": 0.2613,
+      "step": 11187
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017520145246108338,
+      "loss": 0.054,
+      "step": 11188
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017518274040888537,
+      "loss": 0.5707,
+      "step": 11189
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017516402795342352,
+      "loss": 0.9155,
+      "step": 11190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001751453150949976,
+      "loss": 0.3323,
+      "step": 11191
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017512660183390718,
+      "loss": 0.0073,
+      "step": 11192
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017510788817045193,
+      "loss": 0.6789,
+      "step": 11193
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017508917410493153,
+      "loss": 0.4919,
+      "step": 11194
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017507045963764564,
+      "loss": 0.0088,
+      "step": 11195
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017505174476889398,
+      "loss": 0.3338,
+      "step": 11196
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017503302949897626,
+      "loss": 0.2481,
+      "step": 11197
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017501431382819209,
+      "loss": 0.8812,
+      "step": 11198
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017499559775684126,
+      "loss": 0.0262,
+      "step": 11199
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017497688128522348,
+      "loss": 0.3546,
+      "step": 11200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017495816441363832,
+      "loss": 0.1888,
+      "step": 11201
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001749394471423857,
+      "loss": 0.2384,
+      "step": 11202
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017492072947176526,
+      "loss": 0.2821,
+      "step": 11203
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017490201140207674,
+      "loss": 0.9567,
+      "step": 11204
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017488329293361982,
+      "loss": 0.2854,
+      "step": 11205
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017486457406669435,
+      "loss": 0.8414,
+      "step": 11206
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017484585480160005,
+      "loss": 0.7201,
+      "step": 11207
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017482713513863668,
+      "loss": 0.5046,
+      "step": 11208
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017480841507810398,
+      "loss": 0.7478,
+      "step": 11209
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017478969462030174,
+      "loss": 0.3715,
+      "step": 11210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001747709737655298,
+      "loss": 0.3416,
+      "step": 11211
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017475225251408784,
+      "loss": 0.3555,
+      "step": 11212
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017473353086627568,
+      "loss": 0.5695,
+      "step": 11213
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017471480882239322,
+      "loss": 0.7797,
+      "step": 11214
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017469608638274014,
+      "loss": 0.6786,
+      "step": 11215
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017467736354761625,
+      "loss": 0.5546,
+      "step": 11216
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001746586403173215,
+      "loss": 0.0909,
+      "step": 11217
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017463991669215562,
+      "loss": 0.5176,
+      "step": 11218
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017462119267241843,
+      "loss": 0.3171,
+      "step": 11219
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001746024682584098,
+      "loss": 0.7376,
+      "step": 11220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017458374345042955,
+      "loss": 0.2021,
+      "step": 11221
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017456501824877752,
+      "loss": 0.1359,
+      "step": 11222
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001745462926537537,
+      "loss": 0.6381,
+      "step": 11223
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017452756666565773,
+      "loss": 0.7888,
+      "step": 11224
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017450884028478962,
+      "loss": 0.4312,
+      "step": 11225
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017449011351144925,
+      "loss": 0.222,
+      "step": 11226
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017447138634593643,
+      "loss": 1.4936,
+      "step": 11227
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017445265878855113,
+      "loss": 0.8488,
+      "step": 11228
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001744339308395931,
+      "loss": 0.6904,
+      "step": 11229
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017441520249936244,
+      "loss": 0.3021,
+      "step": 11230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001743964737681589,
+      "loss": 1.242,
+      "step": 11231
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017437774464628244,
+      "loss": 0.4001,
+      "step": 11232
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000174359015134033,
+      "loss": 0.7659,
+      "step": 11233
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017434028523171053,
+      "loss": 0.2023,
+      "step": 11234
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017432155493961486,
+      "loss": 1.7289,
+      "step": 11235
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017430282425804603,
+      "loss": 0.3935,
+      "step": 11236
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017428409318730398,
+      "loss": 0.279,
+      "step": 11237
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017426536172768855,
+      "loss": 0.1267,
+      "step": 11238
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017424662987949977,
+      "loss": 0.0102,
+      "step": 11239
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017422789764303767,
+      "loss": 0.1772,
+      "step": 11240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001742091650186021,
+      "loss": 0.3952,
+      "step": 11241
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017419043200649306,
+      "loss": 0.9961,
+      "step": 11242
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001741716986070106,
+      "loss": 0.0856,
+      "step": 11243
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017415296482045463,
+      "loss": 0.5177,
+      "step": 11244
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017413423064712517,
+      "loss": 0.6685,
+      "step": 11245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001741154960873223,
+      "loss": 0.8316,
+      "step": 11246
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017409676114134588,
+      "loss": 0.2912,
+      "step": 11247
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017407802580949602,
+      "loss": 0.4793,
+      "step": 11248
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017405929009207277,
+      "loss": 0.6618,
+      "step": 11249
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017404055398937597,
+      "loss": 0.1485,
+      "step": 11250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017402181750170585,
+      "loss": 0.569,
+      "step": 11251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017400308062936242,
+      "loss": 0.5483,
+      "step": 11252
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017398434337264563,
+      "loss": 0.9323,
+      "step": 11253
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017396560573185556,
+      "loss": 0.5695,
+      "step": 11254
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017394686770729227,
+      "loss": 0.3265,
+      "step": 11255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017392812929925588,
+      "loss": 0.413,
+      "step": 11256
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017390939050804637,
+      "loss": 0.3497,
+      "step": 11257
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017389065133396387,
+      "loss": 0.9483,
+      "step": 11258
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017387191177730844,
+      "loss": 0.2141,
+      "step": 11259
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001738531718383802,
+      "loss": 0.8493,
+      "step": 11260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001738344315174792,
+      "loss": 0.5232,
+      "step": 11261
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001738156908149055,
+      "loss": 0.0023,
+      "step": 11262
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017379694973095934,
+      "loss": 0.5246,
+      "step": 11263
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001737782082659407,
+      "loss": 0.6238,
+      "step": 11264
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017375946642014968,
+      "loss": 0.8134,
+      "step": 11265
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017374072419388655,
+      "loss": 0.4835,
+      "step": 11266
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017372198158745132,
+      "loss": 0.0046,
+      "step": 11267
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017370323860114417,
+      "loss": 0.354,
+      "step": 11268
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017368449523526524,
+      "loss": 0.9506,
+      "step": 11269
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017366575149011461,
+      "loss": 0.7111,
+      "step": 11270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001736470073659925,
+      "loss": 0.572,
+      "step": 11271
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017362826286319914,
+      "loss": 0.701,
+      "step": 11272
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017360951798203451,
+      "loss": 1.0215,
+      "step": 11273
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001735907727227989,
+      "loss": 0.2617,
+      "step": 11274
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001735720270857925,
+      "loss": 0.9565,
+      "step": 11275
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017355328107131546,
+      "loss": 0.5516,
+      "step": 11276
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017353453467966793,
+      "loss": 0.1968,
+      "step": 11277
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001735157879111502,
+      "loss": 0.7744,
+      "step": 11278
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017349704076606242,
+      "loss": 0.0052,
+      "step": 11279
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017347829324470477,
+      "loss": 0.833,
+      "step": 11280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001734595453473775,
+      "loss": 0.9849,
+      "step": 11281
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001734407970743808,
+      "loss": 0.5683,
+      "step": 11282
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017342204842601494,
+      "loss": 0.3259,
+      "step": 11283
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001734032994025801,
+      "loss": 0.0043,
+      "step": 11284
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017338455000437655,
+      "loss": 0.5168,
+      "step": 11285
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017336580023170458,
+      "loss": 0.3653,
+      "step": 11286
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017334705008486434,
+      "loss": 0.9485,
+      "step": 11287
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017332829956415608,
+      "loss": 0.1167,
+      "step": 11288
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017330954866988022,
+      "loss": 0.8442,
+      "step": 11289
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017329079740233686,
+      "loss": 0.4286,
+      "step": 11290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017327204576182633,
+      "loss": 0.5797,
+      "step": 11291
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017325329374864892,
+      "loss": 1.1943,
+      "step": 11292
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017323454136310493,
+      "loss": 0.2747,
+      "step": 11293
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001732157886054946,
+      "loss": 0.3275,
+      "step": 11294
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017319703547611832,
+      "loss": 0.4431,
+      "step": 11295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017317828197527623,
+      "loss": 0.2281,
+      "step": 11296
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001731595281032688,
+      "loss": 0.7793,
+      "step": 11297
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001731407738603963,
+      "loss": 0.5556,
+      "step": 11298
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017312201924695898,
+      "loss": 0.3464,
+      "step": 11299
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017310326426325727,
+      "loss": 0.7021,
+      "step": 11300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017308450890959147,
+      "loss": 0.0051,
+      "step": 11301
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017306575318626186,
+      "loss": 0.2778,
+      "step": 11302
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017304699709356884,
+      "loss": 0.6113,
+      "step": 11303
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017302824063181277,
+      "loss": 0.0071,
+      "step": 11304
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017300948380129396,
+      "loss": 0.9708,
+      "step": 11305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017299072660231283,
+      "loss": 0.3971,
+      "step": 11306
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001729719690351697,
+      "loss": 0.5388,
+      "step": 11307
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017295321110016496,
+      "loss": 0.507,
+      "step": 11308
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017293445279759897,
+      "loss": 0.485,
+      "step": 11309
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001729156941277722,
+      "loss": 0.2792,
+      "step": 11310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017289693509098492,
+      "loss": 0.2317,
+      "step": 11311
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017287817568753763,
+      "loss": 0.2538,
+      "step": 11312
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017285941591773068,
+      "loss": 2.2159,
+      "step": 11313
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017284065578186445,
+      "loss": 1.832,
+      "step": 11314
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017282189528023946,
+      "loss": 0.7313,
+      "step": 11315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017280313441315603,
+      "loss": 0.124,
+      "step": 11316
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017278437318091463,
+      "loss": 0.2443,
+      "step": 11317
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001727656115838157,
+      "loss": 0.0239,
+      "step": 11318
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017274684962215964,
+      "loss": 0.2199,
+      "step": 11319
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017272808729624698,
+      "loss": 0.6886,
+      "step": 11320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017270932460637808,
+      "loss": 0.6004,
+      "step": 11321
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017269056155285337,
+      "loss": 0.3414,
+      "step": 11322
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017267179813597344,
+      "loss": 1.1344,
+      "step": 11323
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001726530343560387,
+      "loss": 0.2027,
+      "step": 11324
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017263427021334957,
+      "loss": 0.3438,
+      "step": 11325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017261550570820658,
+      "loss": 0.0011,
+      "step": 11326
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017259674084091022,
+      "loss": 0.4644,
+      "step": 11327
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017257797561176096,
+      "loss": 0.0029,
+      "step": 11328
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001725592100210593,
+      "loss": 0.8366,
+      "step": 11329
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017254044406910576,
+      "loss": 0.3631,
+      "step": 11330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001725216777562008,
+      "loss": 0.5366,
+      "step": 11331
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000172502911082645,
+      "loss": 0.3473,
+      "step": 11332
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017248414404873888,
+      "loss": 0.6082,
+      "step": 11333
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001724653766547829,
+      "loss": 0.334,
+      "step": 11334
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017244660890107765,
+      "loss": 0.2224,
+      "step": 11335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017242784078792364,
+      "loss": 1.2532,
+      "step": 11336
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001724090723156214,
+      "loss": 0.5513,
+      "step": 11337
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017239030348447154,
+      "loss": 0.3571,
+      "step": 11338
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017237153429477451,
+      "loss": 0.5467,
+      "step": 11339
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017235276474683098,
+      "loss": 1.1258,
+      "step": 11340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017233399484094147,
+      "loss": 1.6953,
+      "step": 11341
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017231522457740654,
+      "loss": 0.2987,
+      "step": 11342
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017229645395652678,
+      "loss": 0.2093,
+      "step": 11343
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017227768297860274,
+      "loss": 0.2361,
+      "step": 11344
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017225891164393507,
+      "loss": 0.5115,
+      "step": 11345
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017224013995282433,
+      "loss": 0.6408,
+      "step": 11346
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017222136790557118,
+      "loss": 0.2033,
+      "step": 11347
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017220259550247607,
+      "loss": 0.3438,
+      "step": 11348
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001721838227438398,
+      "loss": 0.4076,
+      "step": 11349
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017216504962996288,
+      "loss": 0.9188,
+      "step": 11350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017214627616114593,
+      "loss": 0.1673,
+      "step": 11351
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001721275023376896,
+      "loss": 0.2708,
+      "step": 11352
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017210872815989458,
+      "loss": 0.4844,
+      "step": 11353
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017208995362806142,
+      "loss": 2.6055,
+      "step": 11354
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017207117874249085,
+      "loss": 0.2589,
+      "step": 11355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017205240350348343,
+      "loss": 0.8835,
+      "step": 11356
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017203362791133988,
+      "loss": 0.0032,
+      "step": 11357
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001720148519663608,
+      "loss": 0.6651,
+      "step": 11358
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000171996075668847,
+      "loss": 0.1565,
+      "step": 11359
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017197729901909895,
+      "loss": 0.8889,
+      "step": 11360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017195852201741752,
+      "loss": 0.5572,
+      "step": 11361
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001719397446641033,
+      "loss": 0.0035,
+      "step": 11362
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017192096695945697,
+      "loss": 0.6925,
+      "step": 11363
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001719021889037793,
+      "loss": 0.4874,
+      "step": 11364
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001718834104973709,
+      "loss": 1.4131,
+      "step": 11365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017186463174053256,
+      "loss": 0.2784,
+      "step": 11366
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017184585263356494,
+      "loss": 0.2561,
+      "step": 11367
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017182707317676873,
+      "loss": 0.1002,
+      "step": 11368
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017180829337044476,
+      "loss": 0.8682,
+      "step": 11369
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017178951321489372,
+      "loss": 0.706,
+      "step": 11370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017177073271041626,
+      "loss": 0.003,
+      "step": 11371
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017175195185731323,
+      "loss": 0.2841,
+      "step": 11372
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017173317065588534,
+      "loss": 0.3584,
+      "step": 11373
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017171438910643333,
+      "loss": 0.6256,
+      "step": 11374
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017169560720925795,
+      "loss": 0.1862,
+      "step": 11375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017167682496465998,
+      "loss": 0.007,
+      "step": 11376
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017165804237294021,
+      "loss": 0.3928,
+      "step": 11377
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017163925943439935,
+      "loss": 0.7229,
+      "step": 11378
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001716204761493383,
+      "loss": 0.1526,
+      "step": 11379
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017160169251805775,
+      "loss": 0.4282,
+      "step": 11380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001715829085408585,
+      "loss": 0.4185,
+      "step": 11381
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017156412421804134,
+      "loss": 0.1592,
+      "step": 11382
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017154533954990712,
+      "loss": 0.2652,
+      "step": 11383
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017152655453675663,
+      "loss": 0.4988,
+      "step": 11384
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017150776917889072,
+      "loss": 0.4114,
+      "step": 11385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017148898347661004,
+      "loss": 0.2028,
+      "step": 11386
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017147019743021563,
+      "loss": 1.1008,
+      "step": 11387
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017145141104000822,
+      "loss": 0.4762,
+      "step": 11388
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017143262430628867,
+      "loss": 0.5033,
+      "step": 11389
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017141383722935776,
+      "loss": 0.1117,
+      "step": 11390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001713950498095164,
+      "loss": 0.2967,
+      "step": 11391
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017137626204706544,
+      "loss": 0.1304,
+      "step": 11392
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017135747394230575,
+      "loss": 0.1179,
+      "step": 11393
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017133868549553812,
+      "loss": 0.4985,
+      "step": 11394
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017131989670706347,
+      "loss": 0.2655,
+      "step": 11395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017130110757718274,
+      "loss": 0.5979,
+      "step": 11396
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017128231810619663,
+      "loss": 0.3286,
+      "step": 11397
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001712635282944062,
+      "loss": 0.3908,
+      "step": 11398
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001712447381421123,
+      "loss": 0.2598,
+      "step": 11399
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017122594764961576,
+      "loss": 0.2793,
+      "step": 11400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001712071568172175,
+      "loss": 2.168,
+      "step": 11401
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017118836564521855,
+      "loss": 0.9336,
+      "step": 11402
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017116957413391962,
+      "loss": 0.3906,
+      "step": 11403
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017115078228362176,
+      "loss": 0.5769,
+      "step": 11404
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017113199009462587,
+      "loss": 0.0076,
+      "step": 11405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017111319756723292,
+      "loss": 0.2778,
+      "step": 11406
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017109440470174374,
+      "loss": 0.1868,
+      "step": 11407
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017107561149845937,
+      "loss": 0.6178,
+      "step": 11408
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017105681795768067,
+      "loss": 0.1376,
+      "step": 11409
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017103802407970862,
+      "loss": 0.3285,
+      "step": 11410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017101922986484426,
+      "loss": 0.6133,
+      "step": 11411
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001710004353133884,
+      "loss": 0.3397,
+      "step": 11412
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017098164042564213,
+      "loss": 0.7638,
+      "step": 11413
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017096284520190638,
+      "loss": 0.341,
+      "step": 11414
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017094404964248212,
+      "loss": 0.0575,
+      "step": 11415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017092525374767033,
+      "loss": 0.3515,
+      "step": 11416
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017090645751777198,
+      "loss": 0.1897,
+      "step": 11417
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017088766095308812,
+      "loss": 0.2541,
+      "step": 11418
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001708688640539197,
+      "loss": 1.3683,
+      "step": 11419
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017085006682056778,
+      "loss": 0.1509,
+      "step": 11420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017083126925333328,
+      "loss": 0.4679,
+      "step": 11421
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017081247135251735,
+      "loss": 0.0054,
+      "step": 11422
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017079367311842084,
+      "loss": 0.0101,
+      "step": 11423
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017077487455134486,
+      "loss": 0.5751,
+      "step": 11424
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017075607565159053,
+      "loss": 0.3439,
+      "step": 11425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017073727641945874,
+      "loss": 0.8758,
+      "step": 11426
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017071847685525059,
+      "loss": 0.9097,
+      "step": 11427
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017069967695926714,
+      "loss": 0.7581,
+      "step": 11428
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001706808767318094,
+      "loss": 0.9685,
+      "step": 11429
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001706620761731785,
+      "loss": 1.2163,
+      "step": 11430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001706432752836754,
+      "loss": 0.8364,
+      "step": 11431
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017062447406360128,
+      "loss": 0.5834,
+      "step": 11432
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017060567251325713,
+      "loss": 0.2081,
+      "step": 11433
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017058687063294413,
+      "loss": 0.0251,
+      "step": 11434
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001705680684229632,
+      "loss": 0.3897,
+      "step": 11435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017054926588361554,
+      "loss": 0.3517,
+      "step": 11436
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001705304630152023,
+      "loss": 0.217,
+      "step": 11437
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017051165981802444,
+      "loss": 0.6293,
+      "step": 11438
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017049285629238314,
+      "loss": 0.5738,
+      "step": 11439
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001704740524385795,
+      "loss": 0.3992,
+      "step": 11440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017045524825691467,
+      "loss": 0.2469,
+      "step": 11441
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001704364437476897,
+      "loss": 0.0021,
+      "step": 11442
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017041763891120575,
+      "loss": 0.0024,
+      "step": 11443
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000170398833747764,
+      "loss": 3.004,
+      "step": 11444
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017038002825766547,
+      "loss": 0.3182,
+      "step": 11445
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017036122244121142,
+      "loss": 0.5232,
+      "step": 11446
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017034241629870298,
+      "loss": 0.1432,
+      "step": 11447
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017032360983044125,
+      "loss": 0.6262,
+      "step": 11448
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017030480303672742,
+      "loss": 0.2832,
+      "step": 11449
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017028599591786256,
+      "loss": 1.0324,
+      "step": 11450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017026718847414803,
+      "loss": 0.7588,
+      "step": 11451
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017024838070588485,
+      "loss": 0.0315,
+      "step": 11452
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017022957261337424,
+      "loss": 0.5231,
+      "step": 11453
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017021076419691737,
+      "loss": 1.0724,
+      "step": 11454
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017019195545681546,
+      "loss": 0.1754,
+      "step": 11455
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017017314639336967,
+      "loss": 0.2297,
+      "step": 11456
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001701543370068813,
+      "loss": 0.4557,
+      "step": 11457
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017013552729765134,
+      "loss": 0.251,
+      "step": 11458
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001701167172659812,
+      "loss": 0.8681,
+      "step": 11459
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017009790691217204,
+      "loss": 0.341,
+      "step": 11460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017007909623652502,
+      "loss": 0.1599,
+      "step": 11461
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017006028523934143,
+      "loss": 0.6662,
+      "step": 11462
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017004147392092248,
+      "loss": 0.7385,
+      "step": 11463
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001700226622815694,
+      "loss": 0.5619,
+      "step": 11464
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017000385032158344,
+      "loss": 0.2779,
+      "step": 11465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016998503804126582,
+      "loss": 0.3066,
+      "step": 11466
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016996622544091782,
+      "loss": 0.0524,
+      "step": 11467
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001699474125208407,
+      "loss": 0.8836,
+      "step": 11468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016992859928133567,
+      "loss": 0.0039,
+      "step": 11469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016990978572270404,
+      "loss": 0.0317,
+      "step": 11470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016989097184524711,
+      "loss": 0.2886,
+      "step": 11471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016987215764926612,
+      "loss": 0.5201,
+      "step": 11472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016985334313506228,
+      "loss": 0.4142,
+      "step": 11473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000169834528302937,
+      "loss": 0.0242,
+      "step": 11474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001698157131531915,
+      "loss": 1.6837,
+      "step": 11475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016979689768612706,
+      "loss": 0.4292,
+      "step": 11476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016977808190204508,
+      "loss": 0.8759,
+      "step": 11477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016975926580124677,
+      "loss": 1.3409,
+      "step": 11478
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016974044938403347,
+      "loss": 0.2725,
+      "step": 11479
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016972163265070652,
+      "loss": 0.1803,
+      "step": 11480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016970281560156717,
+      "loss": 0.1699,
+      "step": 11481
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016968399823691683,
+      "loss": 0.6152,
+      "step": 11482
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016966518055705685,
+      "loss": 0.7694,
+      "step": 11483
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016964636256228845,
+      "loss": 0.5982,
+      "step": 11484
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016962754425291305,
+      "loss": 0.736,
+      "step": 11485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016960872562923201,
+      "loss": 0.3002,
+      "step": 11486
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016958990669154662,
+      "loss": 0.625,
+      "step": 11487
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016957108744015828,
+      "loss": 0.2759,
+      "step": 11488
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001695522678753684,
+      "loss": 0.7451,
+      "step": 11489
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016953344799747823,
+      "loss": 0.7166,
+      "step": 11490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016951462780678925,
+      "loss": 1.2163,
+      "step": 11491
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016949580730360277,
+      "loss": 3.9286,
+      "step": 11492
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694769864882202,
+      "loss": 1.2188,
+      "step": 11493
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694581653609429,
+      "loss": 0.2267,
+      "step": 11494
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016943934392207231,
+      "loss": 0.734,
+      "step": 11495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694205221719098,
+      "loss": 0.7526,
+      "step": 11496
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694017001107568,
+      "loss": 0.2254,
+      "step": 11497
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016938287773891468,
+      "loss": 0.3444,
+      "step": 11498
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001693640550566848,
+      "loss": 0.0339,
+      "step": 11499
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001693452320643687,
+      "loss": 0.022,
+      "step": 11500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016932640876226774,
+      "loss": 0.506,
+      "step": 11501
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016930758515068336,
+      "loss": 0.1392,
+      "step": 11502
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001692887612299169,
+      "loss": 0.3018,
+      "step": 11503
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016926993700026996,
+      "loss": 0.2467,
+      "step": 11504
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016925111246204384,
+      "loss": 0.9027,
+      "step": 11505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016923228761554012,
+      "loss": 0.0062,
+      "step": 11506
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001692134624610601,
+      "loss": 0.3019,
+      "step": 11507
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016919463699890534,
+      "loss": 0.5807,
+      "step": 11508
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001691758112293773,
+      "loss": 0.3312,
+      "step": 11509
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001691569851527774,
+      "loss": 0.004,
+      "step": 11510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016913815876940715,
+      "loss": 0.1951,
+      "step": 11511
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016911933207956802,
+      "loss": 1.5522,
+      "step": 11512
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016910050508356145,
+      "loss": 0.0035,
+      "step": 11513
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016908167778168892,
+      "loss": 0.7723,
+      "step": 11514
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016906285017425206,
+      "loss": 0.8097,
+      "step": 11515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016904402226155217,
+      "loss": 1.4436,
+      "step": 11516
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001690251940438909,
+      "loss": 0.1792,
+      "step": 11517
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016900636552156966,
+      "loss": 0.8725,
+      "step": 11518
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016898753669489008,
+      "loss": 0.4447,
+      "step": 11519
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001689687075641535,
+      "loss": 0.619,
+      "step": 11520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016894987812966162,
+      "loss": 0.0036,
+      "step": 11521
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001689310483917158,
+      "loss": 2.9376,
+      "step": 11522
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016891221835061774,
+      "loss": 0.3678,
+      "step": 11523
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016889338800666882,
+      "loss": 1.1475,
+      "step": 11524
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001688745573601706,
+      "loss": 0.5922,
+      "step": 11525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016885572641142478,
+      "loss": 0.0078,
+      "step": 11526
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016883689516073276,
+      "loss": 0.4765,
+      "step": 11527
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001688180636083961,
+      "loss": 0.0424,
+      "step": 11528
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016879923175471646,
+      "loss": 0.0082,
+      "step": 11529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016878039959999529,
+      "loss": 0.0163,
+      "step": 11530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016876156714453418,
+      "loss": 0.1069,
+      "step": 11531
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001687427343886348,
+      "loss": 0.2811,
+      "step": 11532
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001687239013325986,
+      "loss": 1.2421,
+      "step": 11533
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016870506797672723,
+      "loss": 1.5838,
+      "step": 11534
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001686862343213223,
+      "loss": 0.111,
+      "step": 11535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001686674003666854,
+      "loss": 1.096,
+      "step": 11536
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016864856611311797,
+      "loss": 1.1409,
+      "step": 11537
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016862973156092187,
+      "loss": 1.013,
+      "step": 11538
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016861089671039853,
+      "loss": 1.3809,
+      "step": 11539
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001685920615618496,
+      "loss": 0.377,
+      "step": 11540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016857322611557672,
+      "loss": 0.0028,
+      "step": 11541
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016855439037188148,
+      "loss": 0.5722,
+      "step": 11542
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016853555433106557,
+      "loss": 0.0341,
+      "step": 11543
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016851671799343054,
+      "loss": 0.1579,
+      "step": 11544
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016849788135927805,
+      "loss": 1.5413,
+      "step": 11545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016847904442890977,
+      "loss": 0.8795,
+      "step": 11546
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016846020720262734,
+      "loss": 1.3014,
+      "step": 11547
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016844136968073236,
+      "loss": 0.6329,
+      "step": 11548
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016842253186352655,
+      "loss": 1.0597,
+      "step": 11549
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016840369375131155,
+      "loss": 0.8238,
+      "step": 11550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016838485534438898,
+      "loss": 3.4025,
+      "step": 11551
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016836601664306055,
+      "loss": 0.0302,
+      "step": 11552
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016834717764762795,
+      "loss": 0.2743,
+      "step": 11553
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016832833835839282,
+      "loss": 0.9773,
+      "step": 11554
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016830949877565688,
+      "loss": 0.2505,
+      "step": 11555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016829065889972176,
+      "loss": 0.87,
+      "step": 11556
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016827181873088919,
+      "loss": 0.5137,
+      "step": 11557
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016825297826946094,
+      "loss": 0.2198,
+      "step": 11558
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016823413751573853,
+      "loss": 0.4607,
+      "step": 11559
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001682152964700238,
+      "loss": 1.3282,
+      "step": 11560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016819645513261848,
+      "loss": 0.2218,
+      "step": 11561
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016817761350382423,
+      "loss": 0.4942,
+      "step": 11562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016815877158394272,
+      "loss": 0.0519,
+      "step": 11563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001681399293732758,
+      "loss": 0.1784,
+      "step": 11564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016812108687212513,
+      "loss": 0.0127,
+      "step": 11565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016810224408079244,
+      "loss": 0.3227,
+      "step": 11566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016808340099957946,
+      "loss": 0.3394,
+      "step": 11567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016806455762878798,
+      "loss": 0.6101,
+      "step": 11568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001680457139687197,
+      "loss": 0.0521,
+      "step": 11569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016802687001967644,
+      "loss": 0.5339,
+      "step": 11570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016800802578195983,
+      "loss": 0.2408,
+      "step": 11571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016798918125587173,
+      "loss": 0.678,
+      "step": 11572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016797033644171396,
+      "loss": 0.4232,
+      "step": 11573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016795149133978814,
+      "loss": 0.7215,
+      "step": 11574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001679326459503962,
+      "loss": 0.4797,
+      "step": 11575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016791380027383982,
+      "loss": 0.6304,
+      "step": 11576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016789495431042078,
+      "loss": 0.0979,
+      "step": 11577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016787610806044096,
+      "loss": 1.512,
+      "step": 11578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016785726152420205,
+      "loss": 0.2894,
+      "step": 11579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001678384147020059,
+      "loss": 0.8378,
+      "step": 11580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016781956759415434,
+      "loss": 0.0241,
+      "step": 11581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016780072020094913,
+      "loss": 0.3154,
+      "step": 11582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001677818725226921,
+      "loss": 0.6793,
+      "step": 11583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001677630245596851,
+      "loss": 2.1945,
+      "step": 11584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001677441763122299,
+      "loss": 0.4243,
+      "step": 11585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001677253277806283,
+      "loss": 0.2407,
+      "step": 11586
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016770647896518226,
+      "loss": 0.7149,
+      "step": 11587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016768762986619349,
+      "loss": 0.1202,
+      "step": 11588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016766878048396386,
+      "loss": 0.547,
+      "step": 11589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016764993081879524,
+      "loss": 0.3977,
+      "step": 11590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016763108087098947,
+      "loss": 0.3878,
+      "step": 11591
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001676122306408484,
+      "loss": 0.7307,
+      "step": 11592
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016759338012867388,
+      "loss": 0.3256,
+      "step": 11593
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016757452933476778,
+      "loss": 1.0933,
+      "step": 11594
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016755567825943197,
+      "loss": 0.7523,
+      "step": 11595
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016753682690296836,
+      "loss": 0.2171,
+      "step": 11596
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001675179752656787,
+      "loss": 0.4834,
+      "step": 11597
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016749912334786503,
+      "loss": 0.2111,
+      "step": 11598
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016748027114982917,
+      "loss": 1.0662,
+      "step": 11599
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016746141867187298,
+      "loss": 0.437,
+      "step": 11600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016744256591429837,
+      "loss": 0.4426,
+      "step": 11601
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001674237128774072,
+      "loss": 0.3729,
+      "step": 11602
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001674048595615015,
+      "loss": 0.232,
+      "step": 11603
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016738600596688302,
+      "loss": 0.1609,
+      "step": 11604
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016736715209385379,
+      "loss": 0.2248,
+      "step": 11605
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001673482979427157,
+      "loss": 0.7621,
+      "step": 11606
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001673294435137706,
+      "loss": 2.1316,
+      "step": 11607
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016731058880732052,
+      "loss": 0.5441,
+      "step": 11608
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016729173382366733,
+      "loss": 0.5291,
+      "step": 11609
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000167272878563113,
+      "loss": 0.6452,
+      "step": 11610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016725402302595938,
+      "loss": 0.2095,
+      "step": 11611
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001672351672125085,
+      "loss": 2.6006,
+      "step": 11612
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001672163111230623,
+      "loss": 0.0664,
+      "step": 11613
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016719745475792268,
+      "loss": 0.2883,
+      "step": 11614
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016717859811739167,
+      "loss": 0.5812,
+      "step": 11615
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001671597412017712,
+      "loss": 0.0071,
+      "step": 11616
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016714088401136316,
+      "loss": 0.8524,
+      "step": 11617
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016712202654646966,
+      "loss": 0.6948,
+      "step": 11618
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001671031688073926,
+      "loss": 0.0053,
+      "step": 11619
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016708431079443388,
+      "loss": 0.0026,
+      "step": 11620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016706545250789562,
+      "loss": 0.0147,
+      "step": 11621
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001670465939480798,
+      "loss": 0.0096,
+      "step": 11622
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016702773511528826,
+      "loss": 0.2868,
+      "step": 11623
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016700887600982315,
+      "loss": 0.5168,
+      "step": 11624
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016699001663198642,
+      "loss": 0.4683,
+      "step": 11625
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016697115698208008,
+      "loss": 0.2379,
+      "step": 11626
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001669522970604061,
+      "loss": 0.0032,
+      "step": 11627
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016693343686726656,
+      "loss": 0.0096,
+      "step": 11628
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001669145764029634,
+      "loss": 0.0059,
+      "step": 11629
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016689571566779873,
+      "loss": 0.2949,
+      "step": 11630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016687685466207448,
+      "loss": 0.9223,
+      "step": 11631
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016685799338609275,
+      "loss": 0.416,
+      "step": 11632
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001668391318401556,
+      "loss": 0.3045,
+      "step": 11633
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000166820270024565,
+      "loss": 0.3308,
+      "step": 11634
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016680140793962298,
+      "loss": 0.352,
+      "step": 11635
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001667825455856317,
+      "loss": 0.0025,
+      "step": 11636
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016676368296289306,
+      "loss": 1.6922,
+      "step": 11637
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016674482007170924,
+      "loss": 0.6762,
+      "step": 11638
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016672595691238223,
+      "loss": 0.0021,
+      "step": 11639
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016670709348521416,
+      "loss": 0.2898,
+      "step": 11640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016668822979050706,
+      "loss": 0.0045,
+      "step": 11641
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016666936582856299,
+      "loss": 0.266,
+      "step": 11642
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000166650501599684,
+      "loss": 2.017,
+      "step": 11643
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016663163710417227,
+      "loss": 0.8524,
+      "step": 11644
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016661277234232986,
+      "loss": 0.5632,
+      "step": 11645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016659390731445874,
+      "loss": 0.8126,
+      "step": 11646
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016657504202086115,
+      "loss": 0.0251,
+      "step": 11647
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016655617646183917,
+      "loss": 0.4918,
+      "step": 11648
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016653731063769484,
+      "loss": 0.3505,
+      "step": 11649
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016651844454873029,
+      "loss": 0.4836,
+      "step": 11650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016649957819524767,
+      "loss": 0.3176,
+      "step": 11651
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016648071157754905,
+      "loss": 0.5492,
+      "step": 11652
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001664618446959366,
+      "loss": 0.7828,
+      "step": 11653
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016644297755071232,
+      "loss": 0.5658,
+      "step": 11654
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001664241101421785,
+      "loss": 0.379,
+      "step": 11655
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001664052424706372,
+      "loss": 0.3006,
+      "step": 11656
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016638637453639057,
+      "loss": 0.4209,
+      "step": 11657
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016636750633974073,
+      "loss": 0.6482,
+      "step": 11658
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016634863788098986,
+      "loss": 0.5408,
+      "step": 11659
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001663297691604401,
+      "loss": 0.001,
+      "step": 11660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016631090017839352,
+      "loss": 0.0343,
+      "step": 11661
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001662920309351524,
+      "loss": 0.6495,
+      "step": 11662
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016627316143101887,
+      "loss": 0.0114,
+      "step": 11663
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016625429166629508,
+      "loss": 0.7151,
+      "step": 11664
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016623542164128318,
+      "loss": 0.0214,
+      "step": 11665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001662165513562854,
+      "loss": 0.5094,
+      "step": 11666
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016619768081160385,
+      "loss": 0.0028,
+      "step": 11667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016617881000754078,
+      "loss": 0.4114,
+      "step": 11668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016615993894439834,
+      "loss": 0.011,
+      "step": 11669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001661410676224787,
+      "loss": 0.2859,
+      "step": 11670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016612219604208414,
+      "loss": 1.1286,
+      "step": 11671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016610332420351675,
+      "loss": 0.4656,
+      "step": 11672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016608445210707884,
+      "loss": 0.6844,
+      "step": 11673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016606557975307255,
+      "loss": 2.8946,
+      "step": 11674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016604670714180013,
+      "loss": 0.4517,
+      "step": 11675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016602783427356377,
+      "loss": 0.3531,
+      "step": 11676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016600896114866566,
+      "loss": 2.4551,
+      "step": 11677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001659900877674081,
+      "loss": 0.4767,
+      "step": 11678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001659712141300933,
+      "loss": 0.4138,
+      "step": 11679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016595234023702342,
+      "loss": 0.0032,
+      "step": 11680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016593346608850077,
+      "loss": 1.3577,
+      "step": 11681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001659145916848276,
+      "loss": 0.5815,
+      "step": 11682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016589571702630614,
+      "loss": 0.5397,
+      "step": 11683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001658768421132386,
+      "loss": 0.2452,
+      "step": 11684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016585796694592727,
+      "loss": 1.0204,
+      "step": 11685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016583909152467438,
+      "loss": 0.0015,
+      "step": 11686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016582021584978223,
+      "loss": 0.2382,
+      "step": 11687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016580133992155304,
+      "loss": 0.1446,
+      "step": 11688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016578246374028912,
+      "loss": 0.8097,
+      "step": 11689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016576358730629274,
+      "loss": 1.3707,
+      "step": 11690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016574471061986615,
+      "loss": 0.0215,
+      "step": 11691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016572583368131168,
+      "loss": 0.3585,
+      "step": 11692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016570695649093155,
+      "loss": 0.0701,
+      "step": 11693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016568807904902814,
+      "loss": 0.2624,
+      "step": 11694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001656692013559036,
+      "loss": 0.4105,
+      "step": 11695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016565032341186035,
+      "loss": 0.0999,
+      "step": 11696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001656314452172007,
+      "loss": 0.4446,
+      "step": 11697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001656125667722269,
+      "loss": 0.3417,
+      "step": 11698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016559368807724125,
+      "loss": 0.0868,
+      "step": 11699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001655748091325461,
+      "loss": 0.0832,
+      "step": 11700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016555592993844375,
+      "loss": 0.1333,
+      "step": 11701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016553705049523654,
+      "loss": 0.752,
+      "step": 11702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016551817080322678,
+      "loss": 1.7192,
+      "step": 11703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016549929086271677,
+      "loss": 0.6162,
+      "step": 11704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016548041067400895,
+      "loss": 0.7031,
+      "step": 11705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016546153023740554,
+      "loss": 0.0022,
+      "step": 11706
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016544264955320893,
+      "loss": 0.0014,
+      "step": 11707
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016542376862172144,
+      "loss": 0.3928,
+      "step": 11708
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001654048874432455,
+      "loss": 0.6544,
+      "step": 11709
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016538600601808336,
+      "loss": 0.7333,
+      "step": 11710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016536712434653747,
+      "loss": 5.2402,
+      "step": 11711
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016534824242891012,
+      "loss": 0.5168,
+      "step": 11712
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016532936026550368,
+      "loss": 3.5358,
+      "step": 11713
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016531047785662057,
+      "loss": 0.8879,
+      "step": 11714
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016529159520256315,
+      "loss": 0.8249,
+      "step": 11715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016527271230363372,
+      "loss": 0.1124,
+      "step": 11716
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016525382916013478,
+      "loss": 0.0024,
+      "step": 11717
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016523494577236862,
+      "loss": 0.3687,
+      "step": 11718
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016521606214063773,
+      "loss": 2.0686,
+      "step": 11719
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001651971782652444,
+      "loss": 0.5135,
+      "step": 11720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016517829414649103,
+      "loss": 0.0253,
+      "step": 11721
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001651594097846801,
+      "loss": 0.3029,
+      "step": 11722
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000165140525180114,
+      "loss": 0.2893,
+      "step": 11723
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016512164033309507,
+      "loss": 3.3125,
+      "step": 11724
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016510275524392575,
+      "loss": 0.4862,
+      "step": 11725
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001650838699129085,
+      "loss": 1.7607,
+      "step": 11726
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016506498434034576,
+      "loss": 0.3458,
+      "step": 11727
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016504609852653984,
+      "loss": 0.0499,
+      "step": 11728
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016502721247179324,
+      "loss": 0.9508,
+      "step": 11729
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001650083261764084,
+      "loss": 0.2598,
+      "step": 11730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016498943964068772,
+      "loss": 0.1855,
+      "step": 11731
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001649705528649337,
+      "loss": 0.0923,
+      "step": 11732
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001649516658494487,
+      "loss": 0.396,
+      "step": 11733
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016493277859453518,
+      "loss": 0.454,
+      "step": 11734
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016491389110049573,
+      "loss": 1.0137,
+      "step": 11735
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016489500336763262,
+      "loss": 0.607,
+      "step": 11736
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001648761153962484,
+      "loss": 0.4405,
+      "step": 11737
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016485722718664552,
+      "loss": 0.2618,
+      "step": 11738
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016483833873912645,
+      "loss": 0.3748,
+      "step": 11739
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016481945005399366,
+      "loss": 0.7552,
+      "step": 11740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016480056113154959,
+      "loss": 0.3668,
+      "step": 11741
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016478167197209677,
+      "loss": 0.3076,
+      "step": 11742
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016476278257593767,
+      "loss": 0.0087,
+      "step": 11743
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001647438929433747,
+      "loss": 0.7638,
+      "step": 11744
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016472500307471044,
+      "loss": 0.1329,
+      "step": 11745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016470611297024742,
+      "loss": 0.4102,
+      "step": 11746
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016468722263028802,
+      "loss": 0.0098,
+      "step": 11747
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016466833205513475,
+      "loss": 0.0331,
+      "step": 11748
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016464944124509022,
+      "loss": 0.5919,
+      "step": 11749
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016463055020045686,
+      "loss": 0.2733,
+      "step": 11750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016461165892153718,
+      "loss": 0.8916,
+      "step": 11751
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016459276740863374,
+      "loss": 0.3721,
+      "step": 11752
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000164573875662049,
+      "loss": 0.5417,
+      "step": 11753
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016455498368208552,
+      "loss": 0.8086,
+      "step": 11754
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016453609146904584,
+      "loss": 0.2775,
+      "step": 11755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016451719902323242,
+      "loss": 0.3818,
+      "step": 11756
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016449830634494787,
+      "loss": 0.1205,
+      "step": 11757
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016447941343449476,
+      "loss": 1.2826,
+      "step": 11758
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001644605202921755,
+      "loss": 1.4683,
+      "step": 11759
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016444162691829276,
+      "loss": 0.0252,
+      "step": 11760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000164422733313149,
+      "loss": 0.8931,
+      "step": 11761
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016440383947704683,
+      "loss": 0.4776,
+      "step": 11762
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016438494541028877,
+      "loss": 0.003,
+      "step": 11763
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016436605111317743,
+      "loss": 0.2732,
+      "step": 11764
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016434715658601533,
+      "loss": 0.326,
+      "step": 11765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016432826182910504,
+      "loss": 0.6211,
+      "step": 11766
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016430936684274915,
+      "loss": 0.4641,
+      "step": 11767
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001642904716272502,
+      "loss": 0.0411,
+      "step": 11768
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016427157618291086,
+      "loss": 0.0024,
+      "step": 11769
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016425268051003356,
+      "loss": 0.0027,
+      "step": 11770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000164233784608921,
+      "loss": 0.5297,
+      "step": 11771
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016421488847987578,
+      "loss": 0.871,
+      "step": 11772
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016419599212320044,
+      "loss": 2.9122,
+      "step": 11773
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016417709553919756,
+      "loss": 0.8549,
+      "step": 11774
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001641581987281698,
+      "loss": 0.0048,
+      "step": 11775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016413930169041974,
+      "loss": 0.939,
+      "step": 11776
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016412040442624994,
+      "loss": 0.2356,
+      "step": 11777
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016410150693596312,
+      "loss": 0.0079,
+      "step": 11778
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016408260921986178,
+      "loss": 0.3763,
+      "step": 11779
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016406371127824858,
+      "loss": 0.6222,
+      "step": 11780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016404481311142622,
+      "loss": 0.2423,
+      "step": 11781
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001640259147196972,
+      "loss": 1.558,
+      "step": 11782
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001640070161033642,
+      "loss": 0.2564,
+      "step": 11783
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016398811726272988,
+      "loss": 0.9545,
+      "step": 11784
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016396921819809685,
+      "loss": 0.2549,
+      "step": 11785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016395031890976776,
+      "loss": 0.394,
+      "step": 11786
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016393141939804524,
+      "loss": 0.4043,
+      "step": 11787
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000163912519663232,
+      "loss": 0.4513,
+      "step": 11788
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016389361970563056,
+      "loss": 0.0524,
+      "step": 11789
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016387471952554366,
+      "loss": 0.8354,
+      "step": 11790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016385581912327398,
+      "loss": 0.1623,
+      "step": 11791
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016383691849912414,
+      "loss": 0.2411,
+      "step": 11792
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016381801765339682,
+      "loss": 0.0088,
+      "step": 11793
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016379911658639465,
+      "loss": 0.3124,
+      "step": 11794
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016378021529842043,
+      "loss": 0.6375,
+      "step": 11795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016376131378977667,
+      "loss": 0.3285,
+      "step": 11796
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016374241206076613,
+      "loss": 2.3477,
+      "step": 11797
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016372351011169153,
+      "loss": 1.0136,
+      "step": 11798
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016370460794285546,
+      "loss": 0.9223,
+      "step": 11799
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016368570555456066,
+      "loss": 0.6815,
+      "step": 11800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016366680294710984,
+      "loss": 0.1269,
+      "step": 11801
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001636479001208057,
+      "loss": 0.197,
+      "step": 11802
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016362899707595088,
+      "loss": 0.6939,
+      "step": 11803
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016361009381284813,
+      "loss": 0.4523,
+      "step": 11804
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001635911903318002,
+      "loss": 0.1384,
+      "step": 11805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016357228663310973,
+      "loss": 0.122,
+      "step": 11806
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016355338271707947,
+      "loss": 1.8025,
+      "step": 11807
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016353447858401208,
+      "loss": 0.7042,
+      "step": 11808
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016351557423421038,
+      "loss": 0.373,
+      "step": 11809
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016349666966797703,
+      "loss": 0.7326,
+      "step": 11810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016347776488561476,
+      "loss": 0.2656,
+      "step": 11811
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016345885988742632,
+      "loss": 1.2681,
+      "step": 11812
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016343995467371444,
+      "loss": 0.0481,
+      "step": 11813
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016342104924478183,
+      "loss": 0.4578,
+      "step": 11814
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001634021436009313,
+      "loss": 0.5992,
+      "step": 11815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016338323774246554,
+      "loss": 0.0699,
+      "step": 11816
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016336433166968728,
+      "loss": 0.2847,
+      "step": 11817
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016334542538289936,
+      "loss": 0.3224,
+      "step": 11818
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016332651888240442,
+      "loss": 0.6437,
+      "step": 11819
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016330761216850533,
+      "loss": 0.373,
+      "step": 11820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001632887052415048,
+      "loss": 0.8208,
+      "step": 11821
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001632697981017056,
+      "loss": 1.7829,
+      "step": 11822
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016325089074941047,
+      "loss": 0.7597,
+      "step": 11823
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016323198318492224,
+      "loss": 0.2256,
+      "step": 11824
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016321307540854364,
+      "loss": 0.0667,
+      "step": 11825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016319416742057745,
+      "loss": 0.2636,
+      "step": 11826
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001631752592213265,
+      "loss": 0.754,
+      "step": 11827
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016315635081109355,
+      "loss": 0.2358,
+      "step": 11828
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016313744219018136,
+      "loss": 0.3704,
+      "step": 11829
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016311853335889282,
+      "loss": 0.6703,
+      "step": 11830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016309962431753055,
+      "loss": 0.6712,
+      "step": 11831
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001630807150663975,
+      "loss": 0.4585,
+      "step": 11832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001630618056057965,
+      "loss": 0.379,
+      "step": 11833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001630428959360302,
+      "loss": 1.2048,
+      "step": 11834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001630239860574015,
+      "loss": 0.0023,
+      "step": 11835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016300507597021325,
+      "loss": 0.002,
+      "step": 11836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001629861656747682,
+      "loss": 0.3972,
+      "step": 11837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001629672551713692,
+      "loss": 0.7285,
+      "step": 11838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016294834446031907,
+      "loss": 0.8782,
+      "step": 11839
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016292943354192065,
+      "loss": 0.0439,
+      "step": 11840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016291052241647675,
+      "loss": 1.7705,
+      "step": 11841
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016289161108429018,
+      "loss": 0.3706,
+      "step": 11842
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016287269954566385,
+      "loss": 0.108,
+      "step": 11843
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001628537878009005,
+      "loss": 0.5016,
+      "step": 11844
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001628348758503031,
+      "loss": 0.4287,
+      "step": 11845
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016281596369417437,
+      "loss": 0.207,
+      "step": 11846
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016279705133281726,
+      "loss": 0.9302,
+      "step": 11847
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016277813876653456,
+      "loss": 0.4989,
+      "step": 11848
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016275922599562912,
+      "loss": 2.2168,
+      "step": 11849
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016274031302040384,
+      "loss": 0.4281,
+      "step": 11850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016272139984116157,
+      "loss": 0.1402,
+      "step": 11851
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016270248645820518,
+      "loss": 0.6196,
+      "step": 11852
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016268357287183753,
+      "loss": 0.0311,
+      "step": 11853
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001626646590823615,
+      "loss": 0.3473,
+      "step": 11854
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016264574509007997,
+      "loss": 1.4074,
+      "step": 11855
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001626268308952958,
+      "loss": 0.0039,
+      "step": 11856
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016260791649831186,
+      "loss": 0.3397,
+      "step": 11857
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016258900189943106,
+      "loss": 0.0086,
+      "step": 11858
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016257008709895634,
+      "loss": 0.4941,
+      "step": 11859
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001625511720971905,
+      "loss": 0.2491,
+      "step": 11860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016253225689443648,
+      "loss": 0.3478,
+      "step": 11861
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016251334149099722,
+      "loss": 0.4635,
+      "step": 11862
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001624944258871755,
+      "loss": 0.3862,
+      "step": 11863
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016247551008327432,
+      "loss": 3.4719,
+      "step": 11864
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001624565940795966,
+      "loss": 0.4238,
+      "step": 11865
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001624376778764452,
+      "loss": 0.5661,
+      "step": 11866
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016241876147412307,
+      "loss": 0.5565,
+      "step": 11867
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623998448729331,
+      "loss": 0.009,
+      "step": 11868
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623809280731782,
+      "loss": 0.343,
+      "step": 11869
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623620110751614,
+      "loss": 1.5506,
+      "step": 11870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016234309387918547,
+      "loss": 0.211,
+      "step": 11871
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623241764855534,
+      "loss": 5.0328,
+      "step": 11872
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623052588945682,
+      "loss": 0.1975,
+      "step": 11873
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016228634110653273,
+      "loss": 0.3088,
+      "step": 11874
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016226742312174994,
+      "loss": 0.3409,
+      "step": 11875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016224850494052278,
+      "loss": 0.4818,
+      "step": 11876
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016222958656315418,
+      "loss": 0.389,
+      "step": 11877
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016221066798994714,
+      "loss": 0.316,
+      "step": 11878
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016219174922120456,
+      "loss": 0.8027,
+      "step": 11879
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001621728302572294,
+      "loss": 0.5238,
+      "step": 11880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001621539110983246,
+      "loss": 0.1945,
+      "step": 11881
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016213499174479325,
+      "loss": 0.2094,
+      "step": 11882
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016211607219693814,
+      "loss": 0.0133,
+      "step": 11883
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016209715245506234,
+      "loss": 0.7997,
+      "step": 11884
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001620782325194688,
+      "loss": 0.1731,
+      "step": 11885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016205931239046052,
+      "loss": 1.1439,
+      "step": 11886
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016204039206834037,
+      "loss": 0.57,
+      "step": 11887
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001620214715534115,
+      "loss": 0.0256,
+      "step": 11888
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016200255084597678,
+      "loss": 0.5726,
+      "step": 11889
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016198362994633922,
+      "loss": 0.0042,
+      "step": 11890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016196470885480177,
+      "loss": 0.4635,
+      "step": 11891
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001619457875716675,
+      "loss": 0.2988,
+      "step": 11892
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016192686609723937,
+      "loss": 0.6422,
+      "step": 11893
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016190794443182043,
+      "loss": 2.5918,
+      "step": 11894
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016188902257571354,
+      "loss": 0.5158,
+      "step": 11895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016187010052922185,
+      "loss": 0.3944,
+      "step": 11896
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016185117829264832,
+      "loss": 0.6811,
+      "step": 11897
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016183225586629592,
+      "loss": 0.3025,
+      "step": 11898
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001618133332504677,
+      "loss": 0.3594,
+      "step": 11899
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016179441044546672,
+      "loss": 0.5061,
+      "step": 11900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016177548745159592,
+      "loss": 0.2574,
+      "step": 11901
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016175656426915835,
+      "loss": 0.5715,
+      "step": 11902
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016173764089845707,
+      "loss": 0.1851,
+      "step": 11903
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001617187173397951,
+      "loss": 0.552,
+      "step": 11904
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016169979359347548,
+      "loss": 0.4302,
+      "step": 11905
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016168086965980113,
+      "loss": 1.0799,
+      "step": 11906
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016166194553907527,
+      "loss": 0.344,
+      "step": 11907
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016164302123160087,
+      "loss": 0.4959,
+      "step": 11908
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001616240967376809,
+      "loss": 0.6692,
+      "step": 11909
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016160517205761842,
+      "loss": 0.4256,
+      "step": 11910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016158624719171663,
+      "loss": 0.7689,
+      "step": 11911
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016156732214027845,
+      "loss": 0.3891,
+      "step": 11912
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016154839690360695,
+      "loss": 0.113,
+      "step": 11913
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001615294714820052,
+      "loss": 0.2029,
+      "step": 11914
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016151054587577627,
+      "loss": 0.2482,
+      "step": 11915
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016149162008522323,
+      "loss": 0.6882,
+      "step": 11916
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016147269411064913,
+      "loss": 0.3126,
+      "step": 11917
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016145376795235707,
+      "loss": 1.0376,
+      "step": 11918
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016143484161065008,
+      "loss": 0.007,
+      "step": 11919
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016141591508583134,
+      "loss": 0.7056,
+      "step": 11920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016139698837820376,
+      "loss": 0.0036,
+      "step": 11921
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016137806148807055,
+      "loss": 0.2608,
+      "step": 11922
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016135913441573478,
+      "loss": 0.5656,
+      "step": 11923
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016134020716149949,
+      "loss": 2.548,
+      "step": 11924
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001613212797256678,
+      "loss": 0.4015,
+      "step": 11925
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016130235210854284,
+      "loss": 0.2879,
+      "step": 11926
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016128342431042763,
+      "loss": 0.2723,
+      "step": 11927
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016126449633162534,
+      "loss": 0.6015,
+      "step": 11928
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016124556817243905,
+      "loss": 0.7677,
+      "step": 11929
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016122663983317184,
+      "loss": 0.5602,
+      "step": 11930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001612077113141269,
+      "loss": 0.2891,
+      "step": 11931
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016118878261560724,
+      "loss": 0.3893,
+      "step": 11932
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000161169853737916,
+      "loss": 5.9749,
+      "step": 11933
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016115092468135635,
+      "loss": 0.0183,
+      "step": 11934
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016113199544623134,
+      "loss": 0.9021,
+      "step": 11935
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016111306603284415,
+      "loss": 0.2781,
+      "step": 11936
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016109413644149788,
+      "loss": 0.0017,
+      "step": 11937
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016107520667249568,
+      "loss": 0.742,
+      "step": 11938
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016105627672614067,
+      "loss": 0.1052,
+      "step": 11939
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016103734660273596,
+      "loss": 0.0394,
+      "step": 11940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016101841630258474,
+      "loss": 0.0029,
+      "step": 11941
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016099948582599008,
+      "loss": 0.5033,
+      "step": 11942
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001609805551732552,
+      "loss": 0.0034,
+      "step": 11943
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016096162434468318,
+      "loss": 0.3046,
+      "step": 11944
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001609426933405772,
+      "loss": 0.3956,
+      "step": 11945
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016092376216124041,
+      "loss": 0.3277,
+      "step": 11946
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000160904830806976,
+      "loss": 0.3938,
+      "step": 11947
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016088589927808704,
+      "loss": 1.1056,
+      "step": 11948
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016086696757487676,
+      "loss": 1.3504,
+      "step": 11949
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016084803569764832,
+      "loss": 0.0156,
+      "step": 11950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016082910364670483,
+      "loss": 0.4044,
+      "step": 11951
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016081017142234951,
+      "loss": 0.3253,
+      "step": 11952
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016079123902488552,
+      "loss": 0.5141,
+      "step": 11953
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016077230645461605,
+      "loss": 0.3014,
+      "step": 11954
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016075337371184426,
+      "loss": 0.0018,
+      "step": 11955
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001607344407968733,
+      "loss": 0.3317,
+      "step": 11956
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016071550771000644,
+      "loss": 0.3519,
+      "step": 11957
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016069657445154674,
+      "loss": 1.4226,
+      "step": 11958
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016067764102179746,
+      "loss": 1.9639,
+      "step": 11959
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016065870742106183,
+      "loss": 0.7632,
+      "step": 11960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016063977364964296,
+      "loss": 0.1967,
+      "step": 11961
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001606208397078441,
+      "loss": 0.204,
+      "step": 11962
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001606019055959684,
+      "loss": 0.2393,
+      "step": 11963
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016058297131431914,
+      "loss": 0.8762,
+      "step": 11964
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016056403686319945,
+      "loss": 0.4484,
+      "step": 11965
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016054510224291257,
+      "loss": 0.7913,
+      "step": 11966
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001605261674537617,
+      "loss": 0.2756,
+      "step": 11967
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016050723249605005,
+      "loss": 0.212,
+      "step": 11968
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016048829737008087,
+      "loss": 1.0749,
+      "step": 11969
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016046936207615728,
+      "loss": 0.1141,
+      "step": 11970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016045042661458263,
+      "loss": 1.6873,
+      "step": 11971
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001604314909856601,
+      "loss": 0.0027,
+      "step": 11972
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016041255518969288,
+      "loss": 0.3684,
+      "step": 11973
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001603936192269842,
+      "loss": 0.1593,
+      "step": 11974
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016037468309783727,
+      "loss": 0.0135,
+      "step": 11975
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016035574680255542,
+      "loss": 0.3776,
+      "step": 11976
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016033681034144178,
+      "loss": 0.2172,
+      "step": 11977
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016031787371479966,
+      "loss": 0.8456,
+      "step": 11978
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016029893692293227,
+      "loss": 0.3996,
+      "step": 11979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001602799999661429,
+      "loss": 0.1315,
+      "step": 11980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001602610628447347,
+      "loss": 1.9333,
+      "step": 11981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000160242125559011,
+      "loss": 0.0387,
+      "step": 11982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016022318810927506,
+      "loss": 0.5164,
+      "step": 11983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016020425049583008,
+      "loss": 2.6294,
+      "step": 11984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016018531271897932,
+      "loss": 0.3101,
+      "step": 11985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016016637477902612,
+      "loss": 0.636,
+      "step": 11986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016014743667627364,
+      "loss": 0.9038,
+      "step": 11987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001601284984110252,
+      "loss": 0.6706,
+      "step": 11988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016010955998358407,
+      "loss": 0.0039,
+      "step": 11989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016009062139425348,
+      "loss": 0.379,
+      "step": 11990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016007168264333673,
+      "loss": 0.5886,
+      "step": 11991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016005274373113716,
+      "loss": 0.3648,
+      "step": 11992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016003380465795793,
+      "loss": 0.1097,
+      "step": 11993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016001486542410238,
+      "loss": 0.0037,
+      "step": 11994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015999592602987384,
+      "loss": 1.005,
+      "step": 11995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015997698647557553,
+      "loss": 0.3594,
+      "step": 11996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015995804676151068,
+      "loss": 0.5651,
+      "step": 11997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015993910688798276,
+      "loss": 1.877,
+      "step": 11998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001599201668552949,
+      "loss": 0.0054,
+      "step": 11999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015990122666375046,
+      "loss": 0.4084,
+      "step": 12000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015988228631365277,
+      "loss": 0.9974,
+      "step": 12001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015986334580530504,
+      "loss": 0.3915,
+      "step": 12002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015984440513901066,
+      "loss": 0.6688,
+      "step": 12003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001598254643150729,
+      "loss": 0.3752,
+      "step": 12004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015980652333379508,
+      "loss": 0.5279,
+      "step": 12005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015978758219548054,
+      "loss": 0.544,
+      "step": 12006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015976864090043252,
+      "loss": 0.3465,
+      "step": 12007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015974969944895434,
+      "loss": 0.3585,
+      "step": 12008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015973075784134943,
+      "loss": 0.815,
+      "step": 12009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159711816077921,
+      "loss": 0.7576,
+      "step": 12010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001596928741589724,
+      "loss": 0.2045,
+      "step": 12011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159673932084807,
+      "loss": 0.1649,
+      "step": 12012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015965498985572808,
+      "loss": 0.0089,
+      "step": 12013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159636047472039,
+      "loss": 0.6258,
+      "step": 12014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001596171049340431,
+      "loss": 0.494,
+      "step": 12015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001595981622420436,
+      "loss": 0.0419,
+      "step": 12016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015957921939634403,
+      "loss": 0.1344,
+      "step": 12017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015956027639724762,
+      "loss": 3.0753,
+      "step": 12018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015954133324505768,
+      "loss": 0.0966,
+      "step": 12019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015952238994007765,
+      "loss": 0.3902,
+      "step": 12020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001595034464826109,
+      "loss": 0.4913,
+      "step": 12021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001594845028729606,
+      "loss": 0.6472,
+      "step": 12022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001594655591114303,
+      "loss": 0.4168,
+      "step": 12023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015944661519832324,
+      "loss": 0.3021,
+      "step": 12024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015942767113394282,
+      "loss": 0.4781,
+      "step": 12025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001594087269185924,
+      "loss": 0.1161,
+      "step": 12026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015938978255257535,
+      "loss": 0.009,
+      "step": 12027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159370838036195,
+      "loss": 1.6083,
+      "step": 12028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015935189336975476,
+      "loss": 0.0586,
+      "step": 12029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159332948553558,
+      "loss": 0.7136,
+      "step": 12030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015931400358790802,
+      "loss": 0.0016,
+      "step": 12031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015929505847310833,
+      "loss": 4.686,
+      "step": 12032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015927611320946217,
+      "loss": 0.9536,
+      "step": 12033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015925716779727295,
+      "loss": 0.3403,
+      "step": 12034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015923822223684418,
+      "loss": 1.2447,
+      "step": 12035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015921927652847906,
+      "loss": 0.2926,
+      "step": 12036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001592003306724811,
+      "loss": 0.0019,
+      "step": 12037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015918138466915365,
+      "loss": 0.3406,
+      "step": 12038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001591624385188001,
+      "loss": 0.1532,
+      "step": 12039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015914349222172385,
+      "loss": 1.1974,
+      "step": 12040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015912454577822832,
+      "loss": 0.9887,
+      "step": 12041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015910559918861682,
+      "loss": 0.0061,
+      "step": 12042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015908665245319283,
+      "loss": 0.2414,
+      "step": 12043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015906770557225978,
+      "loss": 0.5789,
+      "step": 12044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015904875854612103,
+      "loss": 0.1894,
+      "step": 12045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015902981137507993,
+      "loss": 1.3916,
+      "step": 12046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015901086405944003,
+      "loss": 0.7283,
+      "step": 12047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015899191659950464,
+      "loss": 0.0025,
+      "step": 12048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001589729689955772,
+      "loss": 1.0291,
+      "step": 12049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001589540212479611,
+      "loss": 1.5586,
+      "step": 12050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001589350733569598,
+      "loss": 0.2298,
+      "step": 12051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015891612532287672,
+      "loss": 0.4011,
+      "step": 12052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015889717714601527,
+      "loss": 0.6721,
+      "step": 12053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015887822882667887,
+      "loss": 0.0302,
+      "step": 12054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015885928036517098,
+      "loss": 0.5612,
+      "step": 12055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000158840331761795,
+      "loss": 0.0017,
+      "step": 12056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015882138301685436,
+      "loss": 0.498,
+      "step": 12057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015880243413065256,
+      "loss": 0.3317,
+      "step": 12058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015878348510349295,
+      "loss": 0.3942,
+      "step": 12059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015876453593567901,
+      "loss": 0.6164,
+      "step": 12060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015874558662751418,
+      "loss": 0.4644,
+      "step": 12061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015872663717930193,
+      "loss": 0.3039,
+      "step": 12062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015870768759134566,
+      "loss": 0.0044,
+      "step": 12063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015868873786394882,
+      "loss": 0.6052,
+      "step": 12064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001586697879974149,
+      "loss": 0.2864,
+      "step": 12065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015865083799204734,
+      "loss": 0.4052,
+      "step": 12066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015863188784814963,
+      "loss": 0.1069,
+      "step": 12067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015861293756602512,
+      "loss": 0.4886,
+      "step": 12068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015859398714597738,
+      "loss": 0.3292,
+      "step": 12069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001585750365883099,
+      "loss": 0.7439,
+      "step": 12070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015855608589332597,
+      "loss": 0.0017,
+      "step": 12071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001585371350613292,
+      "loss": 0.0469,
+      "step": 12072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015851818409262305,
+      "loss": 0.4383,
+      "step": 12073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015849923298751092,
+      "loss": 2.4257,
+      "step": 12074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015848028174629636,
+      "loss": 0.0046,
+      "step": 12075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015846133036928283,
+      "loss": 0.1661,
+      "step": 12076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015844237885677376,
+      "loss": 0.7,
+      "step": 12077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015842342720907265,
+      "loss": 0.0013,
+      "step": 12078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015840447542648308,
+      "loss": 0.001,
+      "step": 12079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015838552350930834,
+      "loss": 0.0016,
+      "step": 12080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015836657145785208,
+      "loss": 0.2127,
+      "step": 12081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015834761927241775,
+      "loss": 0.0576,
+      "step": 12082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015832866695330878,
+      "loss": 0.2256,
+      "step": 12083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015830971450082876,
+      "loss": 0.0383,
+      "step": 12084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015829076191528111,
+      "loss": 0.5345,
+      "step": 12085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015827180919696934,
+      "loss": 0.0037,
+      "step": 12086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015825285634619697,
+      "loss": 0.0115,
+      "step": 12087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001582339033632675,
+      "loss": 0.2759,
+      "step": 12088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015821495024848442,
+      "loss": 0.2553,
+      "step": 12089
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015819599700215125,
+      "loss": 0.5362,
+      "step": 12090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015817704362457144,
+      "loss": 0.376,
+      "step": 12091
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015815809011604856,
+      "loss": 0.1453,
+      "step": 12092
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015813913647688617,
+      "loss": 0.9176,
+      "step": 12093
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015812018270738772,
+      "loss": 0.4875,
+      "step": 12094
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015810122880785662,
+      "loss": 0.4925,
+      "step": 12095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001580822747785966,
+      "loss": 0.0012,
+      "step": 12096
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015806332061991108,
+      "loss": 3.1709,
+      "step": 12097
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015804436633210352,
+      "loss": 1.0824,
+      "step": 12098
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015802541191547756,
+      "loss": 0.3773,
+      "step": 12099
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015800645737033668,
+      "loss": 0.4712,
+      "step": 12100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015798750269698436,
+      "loss": 0.2925,
+      "step": 12101
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015796854789572418,
+      "loss": 1.2088,
+      "step": 12102
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015794959296685967,
+      "loss": 4.0822,
+      "step": 12103
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015793063791069437,
+      "loss": 0.5371,
+      "step": 12104
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015791168272753182,
+      "loss": 0.0078,
+      "step": 12105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001578927274176755,
+      "loss": 0.419,
+      "step": 12106
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015787377198142903,
+      "loss": 0.2713,
+      "step": 12107
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015785481641909595,
+      "loss": 0.2116,
+      "step": 12108
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001578358607309797,
+      "loss": 0.9605,
+      "step": 12109
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001578169049173839,
+      "loss": 0.3226,
+      "step": 12110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015779794897861216,
+      "loss": 0.1343,
+      "step": 12111
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015777899291496796,
+      "loss": 0.3931,
+      "step": 12112
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015776003672675484,
+      "loss": 0.1659,
+      "step": 12113
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015774108041427637,
+      "loss": 0.5716,
+      "step": 12114
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015772212397783611,
+      "loss": 3.0157,
+      "step": 12115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001577031674177377,
+      "loss": 0.0024,
+      "step": 12116
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015768421073428452,
+      "loss": 0.3566,
+      "step": 12117
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015766525392778028,
+      "loss": 0.3073,
+      "step": 12118
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015764629699852853,
+      "loss": 0.3366,
+      "step": 12119
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015762733994683278,
+      "loss": 0.3884,
+      "step": 12120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001576083827729966,
+      "loss": 0.0349,
+      "step": 12121
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015758942547732365,
+      "loss": 0.7955,
+      "step": 12122
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001575704680601174,
+      "loss": 0.135,
+      "step": 12123
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015755151052168148,
+      "loss": 0.1702,
+      "step": 12124
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015753255286231945,
+      "loss": 0.1942,
+      "step": 12125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015751359508233488,
+      "loss": 0.8864,
+      "step": 12126
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001574946371820314,
+      "loss": 2.0548,
+      "step": 12127
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015747567916171254,
+      "loss": 0.3936,
+      "step": 12128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015745672102168186,
+      "loss": 0.52,
+      "step": 12129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015743776276224298,
+      "loss": 0.3937,
+      "step": 12130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015741880438369956,
+      "loss": 0.2421,
+      "step": 12131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015739984588635506,
+      "loss": 0.2462,
+      "step": 12132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015738088727051318,
+      "loss": 0.0036,
+      "step": 12133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015736192853647743,
+      "loss": 0.0009,
+      "step": 12134
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015734296968455144,
+      "loss": 0.7788,
+      "step": 12135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015732401071503883,
+      "loss": 0.2273,
+      "step": 12136
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015730505162824318,
+      "loss": 1.0501,
+      "step": 12137
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015728609242446811,
+      "loss": 0.0026,
+      "step": 12138
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015726713310401717,
+      "loss": 0.3063,
+      "step": 12139
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000157248173667194,
+      "loss": 0.4608,
+      "step": 12140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001572292141143022,
+      "loss": 0.5086,
+      "step": 12141
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015721025444564546,
+      "loss": 0.9906,
+      "step": 12142
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015719129466152724,
+      "loss": 0.003,
+      "step": 12143
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015717233476225122,
+      "loss": 0.8993,
+      "step": 12144
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001571533747481211,
+      "loss": 0.5545,
+      "step": 12145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015713441461944033,
+      "loss": 0.6797,
+      "step": 12146
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015711545437651264,
+      "loss": 0.2271,
+      "step": 12147
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015709649401964167,
+      "loss": 0.2901,
+      "step": 12148
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000157077533549131,
+      "loss": 0.002,
+      "step": 12149
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015705857296528423,
+      "loss": 0.1328,
+      "step": 12150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000157039612268405,
+      "loss": 0.4929,
+      "step": 12151
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001570206514587969,
+      "loss": 0.817,
+      "step": 12152
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015700169053676366,
+      "loss": 0.1211,
+      "step": 12153
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015698272950260888,
+      "loss": 1.9209,
+      "step": 12154
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015696376835663608,
+      "loss": 0.3086,
+      "step": 12155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015694480709914905,
+      "loss": 0.5654,
+      "step": 12156
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015692584573045136,
+      "loss": 0.2815,
+      "step": 12157
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015690688425084665,
+      "loss": 2.627,
+      "step": 12158
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001568879226606385,
+      "loss": 1.5153,
+      "step": 12159
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015686896096013065,
+      "loss": 0.3094,
+      "step": 12160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001568499991496267,
+      "loss": 0.9373,
+      "step": 12161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015683103722943025,
+      "loss": 0.0051,
+      "step": 12162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015681207519984504,
+      "loss": 0.9751,
+      "step": 12163
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015679311306117461,
+      "loss": 0.0008,
+      "step": 12164
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015677415081372272,
+      "loss": 0.287,
+      "step": 12165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015675518845779295,
+      "loss": 0.5968,
+      "step": 12166
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015673622599368897,
+      "loss": 0.8253,
+      "step": 12167
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001567172634217145,
+      "loss": 0.0008,
+      "step": 12168
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001566983007421731,
+      "loss": 0.2635,
+      "step": 12169
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001566793379553684,
+      "loss": 0.4626,
+      "step": 12170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015666037506160423,
+      "loss": 0.3111,
+      "step": 12171
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001566414120611841,
+      "loss": 0.2311,
+      "step": 12172
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001566224489544117,
+      "loss": 0.9055,
+      "step": 12173
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015660348574159075,
+      "loss": 0.033,
+      "step": 12174
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001565845224230249,
+      "loss": 0.4951,
+      "step": 12175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001565655589990178,
+      "loss": 0.0039,
+      "step": 12176
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015654659546987315,
+      "loss": 0.4781,
+      "step": 12177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001565276318358945,
+      "loss": 0.3697,
+      "step": 12178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001565086680973857,
+      "loss": 0.6775,
+      "step": 12179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015648970425465036,
+      "loss": 0.4834,
+      "step": 12180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015647074030799207,
+      "loss": 0.5004,
+      "step": 12181
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015645177625771465,
+      "loss": 0.7319,
+      "step": 12182
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015643281210412174,
+      "loss": 0.8148,
+      "step": 12183
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015641384784751696,
+      "loss": 0.223,
+      "step": 12184
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000156394883488204,
+      "loss": 0.1904,
+      "step": 12185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015637591902648663,
+      "loss": 0.0012,
+      "step": 12186
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015635695446266845,
+      "loss": 0.0015,
+      "step": 12187
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015633798979705323,
+      "loss": 0.7969,
+      "step": 12188
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015631902502994457,
+      "loss": 0.7389,
+      "step": 12189
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015630006016164623,
+      "loss": 0.8319,
+      "step": 12190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015628109519246185,
+      "loss": 0.043,
+      "step": 12191
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015626213012269523,
+      "loss": 0.3203,
+      "step": 12192
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001562431649526499,
+      "loss": 0.2972,
+      "step": 12193
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015622419968262973,
+      "loss": 0.3703,
+      "step": 12194
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015620523431293833,
+      "loss": 0.428,
+      "step": 12195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015618626884387934,
+      "loss": 0.0033,
+      "step": 12196
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015616730327575664,
+      "loss": 0.0049,
+      "step": 12197
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001561483376088738,
+      "loss": 0.7111,
+      "step": 12198
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015612937184353453,
+      "loss": 1.3782,
+      "step": 12199
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001561104059800426,
+      "loss": 0.322,
+      "step": 12200
     }
   ],
   "logging_steps": 1,
   "max_steps": 24926,
   "num_train_epochs": 1,
   "save_steps": 200,
-  "total_flos": 1.6759298804692255e+18,
+  "total_flos": 2.6954895321745e+18,
   "trial_name": null,
   "trial_params": null
 }