End of training

all_results.json

@@ -1,12 +1,12 @@
 {
     "epoch": 20.0,
     "eval_accuracy": 0.9152691968225949,
-    "eval_loss": 0.3084510564804077,
-    "eval_runtime": 142.0303,
-    "eval_samples_per_second": 47.863,
-    "eval_steps_per_second": 1.5,
-    "train_loss": 0.8349186476609461,
-    "train_runtime": 26368.6047,
-    "train_samples_per_second": 38.754,
-    "train_steps_per_second": 0.605
+    "eval_loss": 0.30927804112434387,
+    "eval_runtime": 150.9277,
+    "eval_samples_per_second": 45.041,
+    "eval_steps_per_second": 1.411,
+    "train_loss": 0.8353394886007285,
+    "train_runtime": 26863.0447,
+    "train_samples_per_second": 38.04,
+    "train_steps_per_second": 0.594
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 20.0,
     "eval_accuracy": 0.9152691968225949,
-    "eval_loss": 0.3084510564804077,
-    "eval_runtime": 142.0303,
-    "eval_samples_per_second": 47.863,
-    "eval_steps_per_second": 1.5
+    "eval_loss": 0.30927804112434387,
+    "eval_runtime": 150.9277,
+    "eval_samples_per_second": 45.041,
+    "eval_steps_per_second": 1.411
 }

train_results.json

@@ -1,7 +1,7 @@
 {
     "epoch": 20.0,
-    "train_loss": 0.8349186476609461,
-    "train_runtime": 26368.6047,
-    "train_samples_per_second": 38.754,
-    "train_steps_per_second": 0.605
+    "train_loss": 0.8353394886007285,
+    "train_runtime": 26863.0447,
+    "train_samples_per_second": 38.04,
+    "train_steps_per_second": 0.594
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
   "best_metric": 0.9152691968225949,
-  "best_model_checkpoint": "trillsson3-ft-keyword-spotting-13/checkpoint-13566",
+  "best_model_checkpoint": "trillsson3-ft-keyword-spotting-13/checkpoint-11970",
   "epoch": 19.999373825923605,
   "global_step": 15960,
   "is_hyper_param_search": false,
@@ -10,1145 +10,1145 @@
     {
       "epoch": 0.13,
       "learning_rate": 1.7669172932330825e-05,
-      "loss": 7.8644,
+      "loss": 7.8872,
       "step": 100
     },
     {
       "epoch": 0.25,
       "learning_rate": 3.646616541353383e-05,
-      "loss": 6.0388,
+      "loss": 6.0261,
       "step": 200
     },
     {
       "epoch": 0.38,
       "learning_rate": 5.526315789473683e-05,
-      "loss": 3.8396,
+      "loss": 3.8766,
       "step": 300
     },
     {
       "epoch": 0.5,
       "learning_rate": 7.406015037593984e-05,
-      "loss": 3.0482,
+      "loss": 3.1036,
       "step": 400
     },
     {
       "epoch": 0.63,
       "learning_rate": 9.285714285714286e-05,
-      "loss": 2.6758,
+      "loss": 2.5949,
       "step": 500
     },
     {
       "epoch": 0.75,
       "learning_rate": 0.00011165413533834586,
-      "loss": 2.1056,
+      "loss": 2.158,
       "step": 600
     },
     {
       "epoch": 0.88,
       "learning_rate": 0.00013045112781954885,
-      "loss": 1.8064,
+      "loss": 1.7756,
       "step": 700
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.6403353927625772,
-      "eval_loss": 0.9358639717102051,
-      "eval_runtime": 141.8393,
-      "eval_samples_per_second": 47.927,
-      "eval_steps_per_second": 1.502,
+      "eval_accuracy": 0.6395998823183289,
+      "eval_loss": 0.9283319711685181,
+      "eval_runtime": 153.4542,
+      "eval_samples_per_second": 44.3,
+      "eval_steps_per_second": 1.388,
       "step": 798
     },
     {
       "epoch": 1.0,
       "learning_rate": 0.00014924812030075185,
-      "loss": 1.4335,
+      "loss": 1.4696,
       "step": 800
     },
     {
       "epoch": 1.13,
       "learning_rate": 0.00016804511278195486,
-      "loss": 1.2967,
+      "loss": 1.2914,
       "step": 900
     },
     {
       "epoch": 1.25,
       "learning_rate": 0.0001868421052631579,
-      "loss": 1.1709,
+      "loss": 1.1715,
       "step": 1000
     },
     {
       "epoch": 1.38,
       "learning_rate": 0.0002056390977443609,
-      "loss": 1.0607,
+      "loss": 1.087,
       "step": 1100
     },
     {
       "epoch": 1.5,
       "learning_rate": 0.0002244360902255639,
-      "loss": 0.9944,
+      "loss": 0.978,
       "step": 1200
     },
     {
       "epoch": 1.63,
       "learning_rate": 0.0002432330827067669,
-      "loss": 0.9323,
+      "loss": 0.92,
       "step": 1300
     },
     {
       "epoch": 1.75,
       "learning_rate": 0.0002620300751879699,
-      "loss": 0.8718,
+      "loss": 0.8856,
       "step": 1400
     },
     {
       "epoch": 1.88,
       "learning_rate": 0.0002808270676691729,
-      "loss": 0.8601,
+      "loss": 0.8631,
       "step": 1500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.8527508090614887,
-      "eval_loss": 0.4831967353820801,
-      "eval_runtime": 141.9778,
-      "eval_samples_per_second": 47.881,
-      "eval_steps_per_second": 1.5,
+      "eval_accuracy": 0.8573109738158282,
+      "eval_loss": 0.4884476065635681,
+      "eval_runtime": 147.3242,
+      "eval_samples_per_second": 46.143,
+      "eval_steps_per_second": 1.446,
       "step": 1596
     },
     {
       "epoch": 2.01,
       "learning_rate": 0.0002996240601503759,
-      "loss": 0.836,
+      "loss": 0.837,
       "step": 1600
     },
     {
       "epoch": 2.13,
       "learning_rate": 0.000297953216374269,
-      "loss": 0.8312,
+      "loss": 0.8428,
       "step": 1700
     },
     {
       "epoch": 2.26,
       "learning_rate": 0.0002958646616541353,
-      "loss": 0.8016,
+      "loss": 0.7937,
       "step": 1800
     },
     {
       "epoch": 2.38,
       "learning_rate": 0.0002937761069340017,
-      "loss": 0.7724,
+      "loss": 0.7866,
       "step": 1900
     },
     {
       "epoch": 2.51,
       "learning_rate": 0.00029168755221386795,
-      "loss": 0.7621,
+      "loss": 0.7648,
       "step": 2000
     },
     {
       "epoch": 2.63,
       "learning_rate": 0.00028959899749373433,
-      "loss": 0.7624,
+      "loss": 0.7622,
       "step": 2100
     },
     {
       "epoch": 2.76,
       "learning_rate": 0.00028751044277360065,
-      "loss": 0.7549,
+      "loss": 0.7636,
       "step": 2200
     },
     {
       "epoch": 2.88,
       "learning_rate": 0.000285421888053467,
-      "loss": 0.7585,
+      "loss": 0.7551,
       "step": 2300
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.8854074727861135,
-      "eval_loss": 0.39516541361808777,
-      "eval_runtime": 141.91,
-      "eval_samples_per_second": 47.904,
-      "eval_steps_per_second": 1.501,
+      "eval_accuracy": 0.8832009414533687,
+      "eval_loss": 0.3967166543006897,
+      "eval_runtime": 146.488,
+      "eval_samples_per_second": 46.407,
+      "eval_steps_per_second": 1.454,
       "step": 2394
     },
     {
       "epoch": 3.01,
       "learning_rate": 0.0002833333333333333,
-      "loss": 0.7495,
+      "loss": 0.7545,
       "step": 2400
     },
     {
       "epoch": 3.13,
       "learning_rate": 0.0002812447786131997,
-      "loss": 0.7343,
+      "loss": 0.7217,
       "step": 2500
     },
     {
       "epoch": 3.26,
       "learning_rate": 0.00027915622389306595,
-      "loss": 0.7065,
+      "loss": 0.7158,
       "step": 2600
     },
     {
       "epoch": 3.38,
       "learning_rate": 0.0002770676691729323,
-      "loss": 0.7109,
+      "loss": 0.7164,
       "step": 2700
     },
     {
       "epoch": 3.51,
       "learning_rate": 0.00027497911445279865,
-      "loss": 0.687,
+      "loss": 0.6992,
       "step": 2800
     },
     {
       "epoch": 3.63,
       "learning_rate": 0.000272890559732665,
-      "loss": 0.7325,
+      "loss": 0.7291,
       "step": 2900
     },
     {
       "epoch": 3.76,
       "learning_rate": 0.0002708020050125313,
-      "loss": 0.7026,
+      "loss": 0.6976,
       "step": 3000
     },
     {
       "epoch": 3.88,
       "learning_rate": 0.0002687134502923976,
-      "loss": 0.7026,
+      "loss": 0.6968,
       "step": 3100
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.9049720506031186,
-      "eval_loss": 0.36231401562690735,
-      "eval_runtime": 141.8088,
-      "eval_samples_per_second": 47.938,
-      "eval_steps_per_second": 1.502,
+      "eval_accuracy": 0.8989408649602825,
+      "eval_loss": 0.36444538831710815,
+      "eval_runtime": 146.8663,
+      "eval_samples_per_second": 46.287,
+      "eval_steps_per_second": 1.45,
       "step": 3192
     },
     {
       "epoch": 4.01,
       "learning_rate": 0.00026662489557226395,
-      "loss": 0.73,
+      "loss": 0.7219,
       "step": 3200
     },
     {
       "epoch": 4.14,
       "learning_rate": 0.0002645363408521303,
-      "loss": 0.6916,
+      "loss": 0.6859,
       "step": 3300
     },
     {
       "epoch": 4.26,
       "learning_rate": 0.00026244778613199665,
-      "loss": 0.7066,
+      "loss": 0.7076,
       "step": 3400
     },
     {
       "epoch": 4.39,
       "learning_rate": 0.00026035923141186297,
-      "loss": 0.6909,
+      "loss": 0.6935,
       "step": 3500
     },
     {
       "epoch": 4.51,
       "learning_rate": 0.0002582706766917293,
-      "loss": 0.7258,
+      "loss": 0.7098,
       "step": 3600
     },
     {
       "epoch": 4.64,
       "learning_rate": 0.0002561821219715956,
-      "loss": 0.6908,
+      "loss": 0.6902,
       "step": 3700
     },
     {
       "epoch": 4.76,
       "learning_rate": 0.00025409356725146194,
-      "loss": 0.6977,
+      "loss": 0.6824,
       "step": 3800
     },
     {
       "epoch": 4.89,
       "learning_rate": 0.0002520050125313283,
-      "loss": 0.6924,
+      "loss": 0.67,
       "step": 3900
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.903501029714622,
-      "eval_loss": 0.3456359803676605,
-      "eval_runtime": 141.2792,
-      "eval_samples_per_second": 48.117,
-      "eval_steps_per_second": 1.508,
+      "eval_accuracy": 0.9057075610473668,
+      "eval_loss": 0.342781126499176,
+      "eval_runtime": 148.4976,
+      "eval_samples_per_second": 45.779,
+      "eval_steps_per_second": 1.434,
       "step": 3990
     },
     {
       "epoch": 5.01,
       "learning_rate": 0.00024991645781119465,
-      "loss": 0.7063,
+      "loss": 0.6909,
       "step": 4000
     },
     {
       "epoch": 5.14,
       "learning_rate": 0.00024782790309106097,
-      "loss": 0.714,
+      "loss": 0.7105,
       "step": 4100
     },
     {
       "epoch": 5.26,
       "learning_rate": 0.0002457393483709273,
-      "loss": 0.7052,
+      "loss": 0.7022,
       "step": 4200
     },
     {
       "epoch": 5.39,
       "learning_rate": 0.00024365079365079364,
-      "loss": 0.6786,
+      "loss": 0.6867,
       "step": 4300
     },
     {
       "epoch": 5.51,
-      "learning_rate": 0.00024158312447786128,
-      "loss": 0.6846,
+      "learning_rate": 0.00024156223893065997,
+      "loss": 0.6881,
       "step": 4400
     },
     {
       "epoch": 5.64,
-      "learning_rate": 0.00023949456975772763,
-      "loss": 0.697,
+      "learning_rate": 0.00023947368421052632,
+      "loss": 0.6882,
       "step": 4500
     },
     {
       "epoch": 5.76,
-      "learning_rate": 0.00023740601503759396,
-      "loss": 0.6739,
+      "learning_rate": 0.00023738512949039262,
+      "loss": 0.6711,
       "step": 4600
     },
     {
       "epoch": 5.89,
-      "learning_rate": 0.0002353174603174603,
-      "loss": 0.6816,
+      "learning_rate": 0.00023529657477025897,
+      "loss": 0.6854,
       "step": 4700
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.9005589879376287,
-      "eval_loss": 0.34049057960510254,
-      "eval_runtime": 141.6045,
-      "eval_samples_per_second": 48.007,
-      "eval_steps_per_second": 1.504,
+      "eval_accuracy": 0.902618417181524,
+      "eval_loss": 0.34081852436065674,
+      "eval_runtime": 149.584,
+      "eval_samples_per_second": 45.446,
+      "eval_steps_per_second": 1.424,
       "step": 4788
     },
     {
       "epoch": 6.02,
-      "learning_rate": 0.00023322890559732663,
-      "loss": 0.6973,
+      "learning_rate": 0.0002332080200501253,
+      "loss": 0.6815,
       "step": 4800
     },
     {
       "epoch": 6.14,
-      "learning_rate": 0.00023114035087719298,
-      "loss": 0.6901,
+      "learning_rate": 0.00023111946532999164,
+      "loss": 0.6863,
       "step": 4900
     },
     {
       "epoch": 6.27,
-      "learning_rate": 0.00022905179615705928,
-      "loss": 0.6833,
+      "learning_rate": 0.00022903091060985797,
+      "loss": 0.6843,
       "step": 5000
     },
     {
       "epoch": 6.39,
-      "learning_rate": 0.00022696324143692563,
-      "loss": 0.6515,
+      "learning_rate": 0.0002269423558897243,
+      "loss": 0.6677,
       "step": 5100
     },
     {
       "epoch": 6.52,
-      "learning_rate": 0.00022487468671679195,
-      "loss": 0.7043,
+      "learning_rate": 0.00022485380116959061,
+      "loss": 0.7096,
       "step": 5200
     },
     {
       "epoch": 6.64,
-      "learning_rate": 0.0002227861319966583,
-      "loss": 0.6675,
+      "learning_rate": 0.00022276524644945696,
+      "loss": 0.6534,
       "step": 5300
     },
     {
       "epoch": 6.77,
-      "learning_rate": 0.00022069757727652463,
-      "loss": 0.6833,
+      "learning_rate": 0.0002206766917293233,
+      "loss": 0.6773,
       "step": 5400
     },
     {
       "epoch": 6.89,
-      "learning_rate": 0.00021860902255639098,
-      "loss": 0.6461,
+      "learning_rate": 0.00021858813700918964,
+      "loss": 0.6701,
       "step": 5500
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.9004118858487791,
-      "eval_loss": 0.33838745951652527,
-      "eval_runtime": 142.5146,
-      "eval_samples_per_second": 47.7,
-      "eval_steps_per_second": 1.495,
+      "eval_accuracy": 0.901294498381877,
+      "eval_loss": 0.3358515202999115,
+      "eval_runtime": 149.9158,
+      "eval_samples_per_second": 45.345,
+      "eval_steps_per_second": 1.421,
       "step": 5586
     },
     {
       "epoch": 7.02,
-      "learning_rate": 0.00021652046783625728,
-      "loss": 0.6867,
+      "learning_rate": 0.00021649958228905596,
+      "loss": 0.6695,
       "step": 5600
     },
     {
       "epoch": 7.14,
       "learning_rate": 0.00021443191311612363,
-      "loss": 0.6969,
+      "loss": 0.7058,
       "step": 5700
     },
     {
       "epoch": 7.27,
       "learning_rate": 0.00021234335839598995,
-      "loss": 0.6741,
+      "loss": 0.6577,
       "step": 5800
     },
     {
       "epoch": 7.39,
       "learning_rate": 0.0002102548036758563,
-      "loss": 0.6698,
+      "loss": 0.6799,
       "step": 5900
     },
     {
       "epoch": 7.52,
       "learning_rate": 0.00020816624895572263,
-      "loss": 0.6678,
+      "loss": 0.6727,
       "step": 6000
     },
     {
       "epoch": 7.64,
       "learning_rate": 0.00020607769423558895,
-      "loss": 0.6446,
+      "loss": 0.6397,
       "step": 6100
     },
     {
       "epoch": 7.77,
       "learning_rate": 0.00020398913951545527,
-      "loss": 0.6887,
+      "loss": 0.6845,
       "step": 6200
     },
     {
       "epoch": 7.89,
       "learning_rate": 0.00020190058479532163,
-      "loss": 0.6697,
+      "loss": 0.6734,
       "step": 6300
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.9045307443365695,
-      "eval_loss": 0.3271743953227997,
-      "eval_runtime": 143.1828,
-      "eval_samples_per_second": 47.478,
-      "eval_steps_per_second": 1.488,
+      "eval_accuracy": 0.9058546631362165,
+      "eval_loss": 0.3285204768180847,
+      "eval_runtime": 149.6806,
+      "eval_samples_per_second": 45.417,
+      "eval_steps_per_second": 1.423,
       "step": 6384
     },
     {
       "epoch": 8.02,
-      "learning_rate": 0.0001998329156223893,
-      "loss": 0.676,
+      "learning_rate": 0.00019981203007518795,
+      "loss": 0.6753,
       "step": 6400
     },
     {
       "epoch": 8.15,
-      "learning_rate": 0.00019774436090225564,
-      "loss": 0.6687,
+      "learning_rate": 0.0001977234753550543,
+      "loss": 0.6674,
       "step": 6500
     },
     {
       "epoch": 8.27,
-      "learning_rate": 0.00019565580618212194,
-      "loss": 0.6582,
+      "learning_rate": 0.00019563492063492062,
+      "loss": 0.6704,
       "step": 6600
     },
     {
       "epoch": 8.4,
-      "learning_rate": 0.0001935672514619883,
-      "loss": 0.6945,
+      "learning_rate": 0.00019354636591478695,
+      "loss": 0.686,
       "step": 6700
     },
     {
       "epoch": 8.52,
-      "learning_rate": 0.0001914786967418546,
-      "loss": 0.6876,
+      "learning_rate": 0.00019145781119465327,
+      "loss": 0.6792,
       "step": 6800
     },
     {
       "epoch": 8.65,
-      "learning_rate": 0.00018939014202172096,
-      "loss": 0.6754,
+      "learning_rate": 0.00018936925647451962,
+      "loss": 0.6698,
       "step": 6900
     },
     {
       "epoch": 8.77,
-      "learning_rate": 0.0001873015873015873,
-      "loss": 0.6757,
+      "learning_rate": 0.00018728070175438595,
+      "loss": 0.6755,
       "step": 7000
     },
     {
       "epoch": 8.9,
-      "learning_rate": 0.0001852130325814536,
-      "loss": 0.6575,
+      "learning_rate": 0.0001851921470342523,
+      "loss": 0.6581,
       "step": 7100
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.910856134157105,
-      "eval_loss": 0.3237280249595642,
-      "eval_runtime": 142.0574,
-      "eval_samples_per_second": 47.854,
-      "eval_steps_per_second": 1.499,
+      "eval_accuracy": 0.9095322153574581,
+      "eval_loss": 0.31989586353302,
+      "eval_runtime": 148.5082,
+      "eval_samples_per_second": 45.775,
+      "eval_steps_per_second": 1.434,
       "step": 7182
     },
     {
       "epoch": 9.02,
-      "learning_rate": 0.00018312447786131994,
-      "loss": 0.6676,
+      "learning_rate": 0.0001831035923141186,
+      "loss": 0.6601,
       "step": 7200
     },
     {
       "epoch": 9.15,
-      "learning_rate": 0.00018103592314118629,
-      "loss": 0.6514,
+      "learning_rate": 0.00018101503759398495,
+      "loss": 0.6746,
       "step": 7300
     },
     {
       "epoch": 9.27,
-      "learning_rate": 0.0001789473684210526,
-      "loss": 0.6948,
+      "learning_rate": 0.00017892648287385127,
+      "loss": 0.6805,
       "step": 7400
     },
     {
       "epoch": 9.4,
-      "learning_rate": 0.00017685881370091896,
-      "loss": 0.6489,
+      "learning_rate": 0.00017683792815371762,
+      "loss": 0.6777,
       "step": 7500
     },
     {
       "epoch": 9.52,
-      "learning_rate": 0.00017477025898078526,
-      "loss": 0.6642,
+      "learning_rate": 0.00017474937343358395,
+      "loss": 0.6619,
       "step": 7600
     },
     {
       "epoch": 9.65,
-      "learning_rate": 0.0001726817042606516,
-      "loss": 0.6609,
+      "learning_rate": 0.0001726608187134503,
+      "loss": 0.6625,
       "step": 7700
     },
     {
       "epoch": 9.77,
-      "learning_rate": 0.00017059314954051793,
-      "loss": 0.666,
+      "learning_rate": 0.0001705722639933166,
+      "loss": 0.665,
       "step": 7800
     },
     {
       "epoch": 9.9,
-      "learning_rate": 0.00016850459482038428,
-      "loss": 0.6634,
+      "learning_rate": 0.00016848370927318294,
+      "loss": 0.6557,
       "step": 7900
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.902618417181524,
-      "eval_loss": 0.32581621408462524,
-      "eval_runtime": 143.1323,
-      "eval_samples_per_second": 47.495,
-      "eval_steps_per_second": 1.488,
+      "eval_accuracy": 0.8986466607825832,
+      "eval_loss": 0.3300594091415405,
+      "eval_runtime": 148.7316,
+      "eval_samples_per_second": 45.706,
+      "eval_steps_per_second": 1.432,
       "step": 7980
     },
     {
       "epoch": 10.03,
-      "learning_rate": 0.0001664160401002506,
-      "loss": 0.659,
+      "learning_rate": 0.00016639515455304927,
+      "loss": 0.6688,
       "step": 8000
     },
     {
       "epoch": 10.15,
       "learning_rate": 0.00016432748538011696,
-      "loss": 0.6515,
+      "loss": 0.6642,
       "step": 8100
     },
     {
       "epoch": 10.28,
       "learning_rate": 0.00016223893065998326,
-      "loss": 0.6513,
+      "loss": 0.6542,
       "step": 8200
     },
     {
       "epoch": 10.4,
       "learning_rate": 0.0001601503759398496,
-      "loss": 0.6516,
+      "loss": 0.6461,
       "step": 8300
     },
     {
       "epoch": 10.53,
       "learning_rate": 0.00015806182121971593,
-      "loss": 0.6812,
+      "loss": 0.6721,
       "step": 8400
     },
     {
       "epoch": 10.65,
       "learning_rate": 0.00015597326649958228,
-      "loss": 0.6584,
+      "loss": 0.6642,
       "step": 8500
     },
     {
       "epoch": 10.78,
       "learning_rate": 0.0001538847117794486,
-      "loss": 0.6548,
+      "loss": 0.6529,
       "step": 8600
     },
     {
       "epoch": 10.9,
       "learning_rate": 0.00015179615705931496,
-      "loss": 0.6604,
+      "loss": 0.6768,
       "step": 8700
     },
     {
       "epoch": 11.0,
-      "eval_accuracy": 0.9042365401588702,
-      "eval_loss": 0.3179236054420471,
-      "eval_runtime": 144.9391,
-      "eval_samples_per_second": 46.902,
-      "eval_steps_per_second": 1.47,
+      "eval_accuracy": 0.9046778464254193,
+      "eval_loss": 0.31744641065597534,
+      "eval_runtime": 148.9285,
+      "eval_samples_per_second": 45.646,
+      "eval_steps_per_second": 1.43,
       "step": 8778
     },
     {
       "epoch": 11.03,
       "learning_rate": 0.00014970760233918125,
-      "loss": 0.7029,
+      "loss": 0.7036,
       "step": 8800
     },
     {
       "epoch": 11.15,
-      "learning_rate": 0.00014763993316624894,
-      "loss": 0.6572,
+      "learning_rate": 0.0001476190476190476,
+      "loss": 0.6825,
       "step": 8900
     },
     {
       "epoch": 11.28,
-      "learning_rate": 0.00014555137844611527,
-      "loss": 0.6593,
+      "learning_rate": 0.00014553049289891393,
+      "loss": 0.6693,
       "step": 9000
     },
     {
       "epoch": 11.4,
-      "learning_rate": 0.00014346282372598162,
-      "loss": 0.6585,
+      "learning_rate": 0.00014344193817878025,
+      "loss": 0.661,
       "step": 9100
     },
     {
       "epoch": 11.53,
-      "learning_rate": 0.00014137426900584794,
-      "loss": 0.6797,
+      "learning_rate": 0.0001413533834586466,
+      "loss": 0.6625,
       "step": 9200
     },
     {
       "epoch": 11.65,
-      "learning_rate": 0.00013928571428571427,
-      "loss": 0.6646,
+      "learning_rate": 0.00013926482873851293,
+      "loss": 0.6523,
       "step": 9300
     },
     {
       "epoch": 11.78,
-      "learning_rate": 0.00013719715956558062,
-      "loss": 0.6978,
+      "learning_rate": 0.00013717627401837925,
+      "loss": 0.6765,
       "step": 9400
     },
     {
       "epoch": 11.9,
-      "learning_rate": 0.00013510860484544694,
-      "loss": 0.6483,
+      "learning_rate": 0.0001350877192982456,
+      "loss": 0.6459,
       "step": 9500
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.9058546631362165,
-      "eval_loss": 0.3203018307685852,
-      "eval_runtime": 145.5825,
-      "eval_samples_per_second": 46.695,
-      "eval_steps_per_second": 1.463,
+      "eval_accuracy": 0.903059723448073,
+      "eval_loss": 0.3192310631275177,
+      "eval_runtime": 149.2235,
+      "eval_samples_per_second": 45.556,
+      "eval_steps_per_second": 1.427,
       "step": 9576
     },
     {
       "epoch": 12.03,
-      "learning_rate": 0.00013302005012531327,
-      "loss": 0.6659,
+      "learning_rate": 0.00013299916457811193,
+      "loss": 0.6518,
       "step": 9600
     },
     {
       "epoch": 12.16,
-      "learning_rate": 0.00013093149540517962,
-      "loss": 0.6641,
+      "learning_rate": 0.00013091060985797825,
+      "loss": 0.6644,
       "step": 9700
     },
     {
       "epoch": 12.28,
-      "learning_rate": 0.00012884294068504594,
-      "loss": 0.6587,
+      "learning_rate": 0.0001288220551378446,
+      "loss": 0.6565,
       "step": 9800
     },
     {
       "epoch": 12.41,
-      "learning_rate": 0.00012675438596491227,
-      "loss": 0.6521,
+      "learning_rate": 0.00012673350041771093,
+      "loss": 0.6647,
       "step": 9900
     },
     {
       "epoch": 12.53,
-      "learning_rate": 0.00012466583124477862,
-      "loss": 0.6697,
+      "learning_rate": 0.00012464494569757725,
+      "loss": 0.6651,
       "step": 10000
     },
     {
       "epoch": 12.66,
-      "learning_rate": 0.00012257727652464494,
-      "loss": 0.6604,
+      "learning_rate": 0.0001225563909774436,
+      "loss": 0.6697,
       "step": 10100
     },
     {
       "epoch": 12.78,
-      "learning_rate": 0.00012048872180451128,
-      "loss": 0.6298,
+      "learning_rate": 0.00012046783625730992,
+      "loss": 0.6559,
       "step": 10200
     },
     {
       "epoch": 12.91,
-      "learning_rate": 0.0001184001670843776,
-      "loss": 0.6578,
+      "learning_rate": 0.00011837928153717626,
+      "loss": 0.6607,
       "step": 10300
     },
     {
       "epoch": 13.0,
-      "eval_accuracy": 0.9089438070020595,
-      "eval_loss": 0.3160444498062134,
-      "eval_runtime": 145.1947,
-      "eval_samples_per_second": 46.82,
-      "eval_steps_per_second": 1.467,
+      "eval_accuracy": 0.9065901735804649,
+      "eval_loss": 0.3172565996646881,
+      "eval_runtime": 149.6048,
+      "eval_samples_per_second": 45.44,
+      "eval_steps_per_second": 1.424,
       "step": 10374
     },
     {
       "epoch": 13.03,
-      "learning_rate": 0.00011631161236424394,
-      "loss": 0.6578,
+      "learning_rate": 0.00011629072681704259,
+      "loss": 0.6435,
       "step": 10400
     },
     {
       "epoch": 13.16,
-      "learning_rate": 0.00011422305764411028,
-      "loss": 0.6588,
+      "learning_rate": 0.00011420217209690892,
+      "loss": 0.6727,
       "step": 10500
     },
     {
       "epoch": 13.28,
-      "learning_rate": 0.0001121345029239766,
+      "learning_rate": 0.00011211361737677526,
       "loss": 0.6701,
       "step": 10600
     },
     {
       "epoch": 13.41,
       "learning_rate": 0.00011004594820384294,
-      "loss": 0.6476,
+      "loss": 0.6474,
       "step": 10700
     },
     {
       "epoch": 13.53,
       "learning_rate": 0.00010795739348370928,
-      "loss": 0.6363,
+      "loss": 0.6307,
       "step": 10800
     },
     {
       "epoch": 13.66,
       "learning_rate": 0.0001058688387635756,
-      "loss": 0.6569,
+      "loss": 0.6748,
       "step": 10900
     },
     {
       "epoch": 13.78,
       "learning_rate": 0.00010378028404344194,
-      "loss": 0.6646,
+      "loss": 0.6382,
       "step": 11000
     },
     {
       "epoch": 13.91,
-      "learning_rate": 0.0001017126148705096,
-      "loss": 0.654,
+      "learning_rate": 0.00010169172932330827,
+      "loss": 0.656,
       "step": 11100
     },
     {
       "epoch": 14.0,
-      "eval_accuracy": 0.9090909090909091,
-      "eval_loss": 0.3138688802719116,
-      "eval_runtime": 144.1517,
-      "eval_samples_per_second": 47.159,
-      "eval_steps_per_second": 1.478,
+      "eval_accuracy": 0.9093851132686084,
+      "eval_loss": 0.3141985833644867,
+      "eval_runtime": 152.4058,
+      "eval_samples_per_second": 44.605,
+      "eval_steps_per_second": 1.398,
       "step": 11172
     },
     {
       "epoch": 14.04,
-      "learning_rate": 9.962406015037594e-05,
-      "loss": 0.6437,
+      "learning_rate": 9.96031746031746e-05,
+      "loss": 0.6329,
       "step": 11200
     },
     {
       "epoch": 14.16,
-      "learning_rate": 9.753550543024226e-05,
-      "loss": 0.671,
+      "learning_rate": 9.751461988304094e-05,
+      "loss": 0.6499,
       "step": 11300
     },
     {
       "epoch": 14.29,
-      "learning_rate": 9.54469507101086e-05,
-      "loss": 0.6526,
+      "learning_rate": 9.542606516290727e-05,
+      "loss": 0.6651,
       "step": 11400
     },
     {
       "epoch": 14.41,
-      "learning_rate": 9.335839598997494e-05,
-      "loss": 0.647,
+      "learning_rate": 9.33375104427736e-05,
+      "loss": 0.6593,
       "step": 11500
     },
     {
       "epoch": 14.54,
-      "learning_rate": 9.126984126984126e-05,
-      "loss": 0.6658,
+      "learning_rate": 9.124895572263994e-05,
+      "loss": 0.6637,
       "step": 11600
     },
     {
       "epoch": 14.66,
-      "learning_rate": 8.91812865497076e-05,
-      "loss": 0.6508,
+      "learning_rate": 8.916040100250626e-05,
+      "loss": 0.6527,
       "step": 11700
     },
     {
       "epoch": 14.79,
-      "learning_rate": 8.709273182957394e-05,
-      "loss": 0.6234,
+      "learning_rate": 8.70718462823726e-05,
+      "loss": 0.6358,
       "step": 11800
     },
     {
       "epoch": 14.91,
-      "learning_rate": 8.500417710944026e-05,
-      "loss": 0.6418,
+      "learning_rate": 8.498329156223893e-05,
+      "loss": 0.6302,
       "step": 11900
     },
     {
       "epoch": 15.0,
-      "eval_accuracy": 0.9124742571344513,
-      "eval_loss": 0.30906587839126587,
-      "eval_runtime": 143.6897,
-      "eval_samples_per_second": 47.31,
-      "eval_steps_per_second": 1.482,
+      "eval_accuracy": 0.9152691968225949,
+      "eval_loss": 0.30927804112434387,
+      "eval_runtime": 149.4384,
+      "eval_samples_per_second": 45.49,
+      "eval_steps_per_second": 1.425,
       "step": 11970
     },
     {
       "epoch": 15.04,
-      "learning_rate": 8.29156223893066e-05,
-      "loss": 0.6882,
+      "learning_rate": 8.289473684210526e-05,
+      "loss": 0.6972,
       "step": 12000
     },
     {
       "epoch": 15.16,
-      "learning_rate": 8.082706766917294e-05,
-      "loss": 0.6246,
+      "learning_rate": 8.08061821219716e-05,
+      "loss": 0.6475,
       "step": 12100
     },
     {
       "epoch": 15.29,
-      "learning_rate": 7.873851294903926e-05,
-      "loss": 0.6462,
+      "learning_rate": 7.871762740183793e-05,
+      "loss": 0.6497,
       "step": 12200
     },
     {
       "epoch": 15.41,
-      "learning_rate": 7.66499582289056e-05,
-      "loss": 0.6391,
+      "learning_rate": 7.662907268170426e-05,
+      "loss": 0.6496,
       "step": 12300
     },
     {
       "epoch": 15.54,
-      "learning_rate": 7.456140350877192e-05,
-      "loss": 0.6373,
+      "learning_rate": 7.454051796157058e-05,
+      "loss": 0.6397,
       "step": 12400
     },
     {
       "epoch": 15.66,
-      "learning_rate": 7.247284878863826e-05,
-      "loss": 0.6593,
+      "learning_rate": 7.245196324143692e-05,
+      "loss": 0.6485,
       "step": 12500
     },
     {
       "epoch": 15.79,
-      "learning_rate": 7.038429406850458e-05,
-      "loss": 0.6799,
+      "learning_rate": 7.036340852130326e-05,
+      "loss": 0.6785,
       "step": 12600
     },
     {
       "epoch": 15.91,
-      "learning_rate": 6.829573934837092e-05,
-      "loss": 0.6394,
+      "learning_rate": 6.827485380116958e-05,
+      "loss": 0.636,
       "step": 12700
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.9029126213592233,
-      "eval_loss": 0.32229486107826233,
-      "eval_runtime": 143.4982,
-      "eval_samples_per_second": 47.373,
-      "eval_steps_per_second": 1.484,
+      "eval_accuracy": 0.9043836422477199,
+      "eval_loss": 0.3184495270252228,
+      "eval_runtime": 150.904,
+      "eval_samples_per_second": 45.049,
+      "eval_steps_per_second": 1.411,
       "step": 12768
     },
     {
       "epoch": 16.04,
-      "learning_rate": 6.620718462823724e-05,
-      "loss": 0.6501,
+      "learning_rate": 6.618629908103592e-05,
+      "loss": 0.6455,
       "step": 12800
     },
     {
       "epoch": 16.17,
-      "learning_rate": 6.411862990810358e-05,
-      "loss": 0.6517,
+      "learning_rate": 6.409774436090225e-05,
+      "loss": 0.629,
       "step": 12900
     },
     {
       "epoch": 16.29,
-      "learning_rate": 6.203007518796992e-05,
-      "loss": 0.6452,
+      "learning_rate": 6.200918964076858e-05,
+      "loss": 0.6345,
       "step": 13000
     },
     {
       "epoch": 16.42,
-      "learning_rate": 5.994152046783625e-05,
-      "loss": 0.6472,
+      "learning_rate": 5.9920634920634916e-05,
+      "loss": 0.6492,
       "step": 13100
     },
     {
       "epoch": 16.54,
-      "learning_rate": 5.785296574770258e-05,
-      "loss": 0.6504,
+      "learning_rate": 5.783208020050125e-05,
+      "loss": 0.6486,
       "step": 13200
     },
     {
       "epoch": 16.67,
-      "learning_rate": 5.578529657477026e-05,
-      "loss": 0.6369,
+      "learning_rate": 5.5743525480367584e-05,
+      "loss": 0.6422,
       "step": 13300
     },
     {
       "epoch": 16.79,
-      "learning_rate": 5.369674185463659e-05,
-      "loss": 0.6703,
+      "learning_rate": 5.367585630743525e-05,
+      "loss": 0.6606,
       "step": 13400
     },
     {
       "epoch": 16.92,
-      "learning_rate": 5.160818713450292e-05,
-      "loss": 0.637,
+      "learning_rate": 5.158730158730158e-05,
+      "loss": 0.6327,
       "step": 13500
     },
     {
       "epoch": 17.0,
-      "eval_accuracy": 0.9152691968225949,
-      "eval_loss": 0.3084510564804077,
-      "eval_runtime": 142.2965,
-      "eval_samples_per_second": 47.773,
-      "eval_steps_per_second": 1.497,
+      "eval_accuracy": 0.911738746690203,
+      "eval_loss": 0.3104247748851776,
+      "eval_runtime": 150.9624,
+      "eval_samples_per_second": 45.031,
+      "eval_steps_per_second": 1.411,
       "step": 13566
     },
     {
       "epoch": 17.04,
-      "learning_rate": 4.951963241436926e-05,
-      "loss": 0.6336,
+      "learning_rate": 4.949874686716791e-05,
+      "loss": 0.6315,
       "step": 13600
     },
     {
       "epoch": 17.17,
-      "learning_rate": 4.743107769423559e-05,
-      "loss": 0.6343,
+      "learning_rate": 4.741019214703425e-05,
+      "loss": 0.6487,
       "step": 13700
     },
     {
       "epoch": 17.29,
-      "learning_rate": 4.534252297410192e-05,
-      "loss": 0.6277,
+      "learning_rate": 4.532163742690058e-05,
+      "loss": 0.6465,
       "step": 13800
     },
     {
       "epoch": 17.42,
-      "learning_rate": 4.325396825396825e-05,
-      "loss": 0.6402,
+      "learning_rate": 4.323308270676691e-05,
+      "loss": 0.6309,
       "step": 13900
     },
     {
       "epoch": 17.54,
-      "learning_rate": 4.1165413533834586e-05,
-      "loss": 0.6435,
+      "learning_rate": 4.1144527986633246e-05,
+      "loss": 0.6369,
       "step": 14000
     },
     {
       "epoch": 17.67,
-      "learning_rate": 3.907685881370092e-05,
-      "loss": 0.6751,
+      "learning_rate": 3.905597326649958e-05,
+      "loss": 0.655,
       "step": 14100
     },
     {
       "epoch": 17.79,
-      "learning_rate": 3.698830409356725e-05,
-      "loss": 0.6788,
+      "learning_rate": 3.696741854636591e-05,
+      "loss": 0.6558,
       "step": 14200
     },
     {
       "epoch": 17.92,
-      "learning_rate": 3.489974937343358e-05,
-      "loss": 0.6258,
+      "learning_rate": 3.4878863826232245e-05,
+      "loss": 0.6428,
       "step": 14300
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.9068843777581642,
-      "eval_loss": 0.31823334097862244,
-      "eval_runtime": 144.3331,
-      "eval_samples_per_second": 47.099,
-      "eval_steps_per_second": 1.476,
+      "eval_accuracy": 0.9083553986466608,
+      "eval_loss": 0.315799742937088,
+      "eval_runtime": 150.6062,
+      "eval_samples_per_second": 45.138,
+      "eval_steps_per_second": 1.414,
       "step": 14364
     },
     {
       "epoch": 18.05,
-      "learning_rate": 3.281119465329991e-05,
-      "loss": 0.6569,
+      "learning_rate": 3.2790309106098576e-05,
+      "loss": 0.6466,
       "step": 14400
     },
     {
       "epoch": 18.17,
-      "learning_rate": 3.072263993316625e-05,
-      "loss": 0.636,
+      "learning_rate": 3.070175438596491e-05,
+      "loss": 0.6219,
       "step": 14500
     },
     {
       "epoch": 18.3,
-      "learning_rate": 2.8634085213032577e-05,
-      "loss": 0.637,
+      "learning_rate": 2.861319966583124e-05,
+      "loss": 0.6351,
       "step": 14600
     },
     {
       "epoch": 18.42,
-      "learning_rate": 2.654553049289891e-05,
-      "loss": 0.6577,
+      "learning_rate": 2.6524644945697575e-05,
+      "loss": 0.6638,
       "step": 14700
     },
     {
       "epoch": 18.55,
-      "learning_rate": 2.4456975772765242e-05,
-      "loss": 0.6397,
+      "learning_rate": 2.4436090225563906e-05,
+      "loss": 0.6414,
       "step": 14800
     },
     {
       "epoch": 18.67,
-      "learning_rate": 2.2368421052631576e-05,
-      "loss": 0.6459,
+      "learning_rate": 2.234753550543024e-05,
+      "loss": 0.6617,
       "step": 14900
     },
     {
       "epoch": 18.8,
-      "learning_rate": 2.0279866332497907e-05,
-      "loss": 0.6389,
+      "learning_rate": 2.025898078529657e-05,
+      "loss": 0.6359,
       "step": 15000
     },
     {
       "epoch": 18.92,
-      "learning_rate": 1.819131161236424e-05,
-      "loss": 0.6438,
+      "learning_rate": 1.8170426065162904e-05,
+      "loss": 0.6515,
       "step": 15100
     },
     {
       "epoch": 19.0,
-      "eval_accuracy": 0.9077669902912622,
-      "eval_loss": 0.31267043948173523,
-      "eval_runtime": 142.4378,
-      "eval_samples_per_second": 47.726,
-      "eval_steps_per_second": 1.495,
+      "eval_accuracy": 0.9096793174463077,
+      "eval_loss": 0.312863290309906,
+      "eval_runtime": 150.6838,
+      "eval_samples_per_second": 45.114,
+      "eval_steps_per_second": 1.414,
       "step": 15162
     },
     {
       "epoch": 19.05,
-      "learning_rate": 1.6102756892230575e-05,
-      "loss": 0.6462,
+      "learning_rate": 1.608187134502924e-05,
+      "loss": 0.6385,
       "step": 15200
     },
     {
       "epoch": 19.17,
-      "learning_rate": 1.4014202172096908e-05,
-      "loss": 0.6364,
+      "learning_rate": 1.3993316624895571e-05,
+      "loss": 0.634,
       "step": 15300
     },
     {
       "epoch": 19.3,
-      "learning_rate": 1.192564745196324e-05,
-      "loss": 0.6412,
+      "learning_rate": 1.1904761904761903e-05,
+      "loss": 0.6509,
       "step": 15400
     },
     {
       "epoch": 19.42,
-      "learning_rate": 9.837092731829572e-06,
-      "loss": 0.6138,
+      "learning_rate": 9.816207184628236e-06,
+      "loss": 0.6388,
       "step": 15500
     },
     {
       "epoch": 19.55,
-      "learning_rate": 7.748538011695905e-06,
-      "loss": 0.6652,
+      "learning_rate": 7.727652464494568e-06,
+      "loss": 0.6493,
       "step": 15600
     },
     {
       "epoch": 19.67,
-      "learning_rate": 5.659983291562238e-06,
-      "loss": 0.64,
+      "learning_rate": 5.6390977443609015e-06,
+      "loss": 0.6526,
       "step": 15700
     },
     {
       "epoch": 19.8,
-      "learning_rate": 3.571428571428571e-06,
-      "loss": 0.6232,
+      "learning_rate": 3.5505430242272343e-06,
+      "loss": 0.6309,
       "step": 15800
     },
     {
       "epoch": 19.92,
-      "learning_rate": 1.4828738512949038e-06,
-      "loss": 0.6569,
+      "learning_rate": 1.4619883040935671e-06,
+      "loss": 0.6441,
       "step": 15900
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.9114445425125037,
-      "eval_loss": 0.3101291060447693,
-      "eval_runtime": 141.8004,
-      "eval_samples_per_second": 47.941,
-      "eval_steps_per_second": 1.502,
+      "eval_accuracy": 0.909973521624007,
+      "eval_loss": 0.31152603030204773,
+      "eval_runtime": 152.3497,
+      "eval_samples_per_second": 44.621,
+      "eval_steps_per_second": 1.398,
       "step": 15960
     },
     {
       "epoch": 20.0,
       "step": 15960,
       "total_flos": 0.0,
-      "train_loss": 0.8349186476609461,
-      "train_runtime": 26368.6047,
-      "train_samples_per_second": 38.754,
-      "train_steps_per_second": 0.605
+      "train_loss": 0.8353394886007285,
+      "train_runtime": 26863.0447,
+      "train_samples_per_second": 38.04,
+      "train_steps_per_second": 0.594
     }
   ],
   "max_steps": 15960,