End of training

all_results.json

@@ -1,12 +1,12 @@
 {
     "epoch": 4.0,
-    "eval_loss": 0.8656262755393982,
-    "eval_runtime": 72.4132,
-    "eval_samples_per_second": 16.668,
-    "eval_steps_per_second": 2.085,
-    "total_flos": 7811792839259640.0,
-    "train_loss": 0.2173631976611841,
-    "train_runtime": 16487.0261,
-    "train_samples_per_second": 2.634,
-    "train_steps_per_second": 1.317
+    "eval_loss": 0.8028334379196167,
+    "eval_runtime": 59.9412,
+    "eval_samples_per_second": 16.466,
+    "eval_steps_per_second": 2.069,
+    "total_flos": 1.349309672235756e+16,
+    "train_loss": 0.20338401917281557,
+    "train_runtime": 27801.3632,
+    "train_samples_per_second": 2.698,
+    "train_steps_per_second": 1.349
 }

eval_results.json

@@ -1,7 +1,7 @@
 {
     "epoch": 4.0,
-    "eval_loss": 0.8656262755393982,
-    "eval_runtime": 72.4132,
-    "eval_samples_per_second": 16.668,
-    "eval_steps_per_second": 2.085
+    "eval_loss": 0.8028334379196167,
+    "eval_runtime": 59.9412,
+    "eval_samples_per_second": 16.466,
+    "eval_steps_per_second": 2.069
 }

runs/Aug23_19-03-57_wolf-creek/events.out.tfevents.1724489607.wolf-creek.3846561.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98c6dd5e42240689e9fa1b421a9dab3d55b7f0b7ace195a446848085f1d8b043
+size 316

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 4.0,
-    "total_flos": 7811792839259640.0,
-    "train_loss": 0.2173631976611841,
-    "train_runtime": 16487.0261,
-    "train_samples_per_second": 2.634,
-    "train_steps_per_second": 1.317
+    "total_flos": 1.349309672235756e+16,
+    "train_loss": 0.20338401917281557,
+    "train_runtime": 27801.3632,
+    "train_samples_per_second": 2.698,
+    "train_steps_per_second": 1.349
 }

trainer_state.json

@@ -1,670 +1,1150 @@
 {
-  "best_metric": 0.8656262755393982,
-  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e4l58-l/checkpoint-21500",
+  "best_metric": 0.8028334379196167,
+  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e4l58-l/checkpoint-37500",
   "epoch": 4.0,
   "eval_steps": 500,
-  "global_step": 21712,
+  "global_step": 37504,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.09211495946941783,
-      "grad_norm": 20.04043960571289,
-      "learning_rate": 4.8848563006632274e-08,
-      "loss": 0.3758,
+      "epoch": 0.05332764505119454,
+      "grad_norm": 217.55360412597656,
+      "learning_rate": 4.933340443686007e-08,
+      "loss": 0.4667,
       "step": 500
     },
     {
-      "epoch": 0.09211495946941783,
-      "eval_loss": 1.4184927940368652,
-      "eval_runtime": 74.3562,
-      "eval_samples_per_second": 16.233,
-      "eval_steps_per_second": 2.031,
+      "epoch": 0.05332764505119454,
+      "eval_loss": 1.4426143169403076,
+      "eval_runtime": 61.9901,
+      "eval_samples_per_second": 15.922,
+      "eval_steps_per_second": 2.0,
       "step": 500
     },
     {
-      "epoch": 0.18422991893883567,
-      "grad_norm": 74.25025939941406,
-      "learning_rate": 4.7697126013264556e-08,
-      "loss": 0.4103,
+      "epoch": 0.10665529010238908,
+      "grad_norm": 73.73340606689453,
+      "learning_rate": 4.8666808873720136e-08,
+      "loss": 0.4532,
       "step": 1000
     },
     {
-      "epoch": 0.18422991893883567,
-      "eval_loss": 1.3500770330429077,
-      "eval_runtime": 75.7013,
-      "eval_samples_per_second": 15.944,
-      "eval_steps_per_second": 1.995,
+      "epoch": 0.10665529010238908,
+      "eval_loss": 1.3815597295761108,
+      "eval_runtime": 62.4624,
+      "eval_samples_per_second": 15.801,
+      "eval_steps_per_second": 1.985,
       "step": 1000
     },
     {
-      "epoch": 0.2763448784082535,
-      "grad_norm": 0.5102410912513733,
-      "learning_rate": 4.6545689019896826e-08,
-      "loss": 0.433,
+      "epoch": 0.1599829351535836,
+      "grad_norm": 471.4011535644531,
+      "learning_rate": 4.80002133105802e-08,
+      "loss": 0.3749,
       "step": 1500
     },
     {
-      "epoch": 0.2763448784082535,
-      "eval_loss": 1.2885024547576904,
-      "eval_runtime": 75.197,
-      "eval_samples_per_second": 16.051,
-      "eval_steps_per_second": 2.008,
+      "epoch": 0.1599829351535836,
+      "eval_loss": 1.3310539722442627,
+      "eval_runtime": 63.3918,
+      "eval_samples_per_second": 15.57,
+      "eval_steps_per_second": 1.956,
       "step": 1500
     },
     {
-      "epoch": 0.36845983787767134,
-      "grad_norm": 0.1667678952217102,
-      "learning_rate": 4.539425202652911e-08,
-      "loss": 0.3424,
+      "epoch": 0.21331058020477817,
+      "grad_norm": 11.962693214416504,
+      "learning_rate": 4.733361774744027e-08,
+      "loss": 0.336,
       "step": 2000
     },
     {
-      "epoch": 0.36845983787767134,
-      "eval_loss": 1.239119052886963,
-      "eval_runtime": 75.387,
-      "eval_samples_per_second": 16.011,
-      "eval_steps_per_second": 2.003,
+      "epoch": 0.21331058020477817,
+      "eval_loss": 1.2890639305114746,
+      "eval_runtime": 63.177,
+      "eval_samples_per_second": 15.623,
+      "eval_steps_per_second": 1.963,
       "step": 2000
     },
     {
-      "epoch": 0.46057479734708917,
-      "grad_norm": 343.8175048828125,
-      "learning_rate": 4.4242815033161385e-08,
-      "loss": 0.3645,
+      "epoch": 0.2666382252559727,
+      "grad_norm": 375.1609191894531,
+      "learning_rate": 4.666702218430034e-08,
+      "loss": 0.3585,
       "step": 2500
     },
     {
-      "epoch": 0.46057479734708917,
-      "eval_loss": 1.1902339458465576,
-      "eval_runtime": 75.9899,
-      "eval_samples_per_second": 15.884,
-      "eval_steps_per_second": 1.987,
+      "epoch": 0.2666382252559727,
+      "eval_loss": 1.2536433935165405,
+      "eval_runtime": 63.1512,
+      "eval_samples_per_second": 15.629,
+      "eval_steps_per_second": 1.964,
       "step": 2500
     },
     {
-      "epoch": 0.552689756816507,
-      "grad_norm": 305.5626525878906,
-      "learning_rate": 4.309137803979366e-08,
-      "loss": 0.3172,
+      "epoch": 0.3199658703071672,
+      "grad_norm": 0.0022936267778277397,
+      "learning_rate": 4.600042662116041e-08,
+      "loss": 0.303,
       "step": 3000
     },
     {
-      "epoch": 0.552689756816507,
-      "eval_loss": 1.1506118774414062,
-      "eval_runtime": 74.8655,
-      "eval_samples_per_second": 16.122,
-      "eval_steps_per_second": 2.017,
+      "epoch": 0.3199658703071672,
+      "eval_loss": 1.2202869653701782,
+      "eval_runtime": 63.5857,
+      "eval_samples_per_second": 15.522,
+      "eval_steps_per_second": 1.95,
       "step": 3000
     },
     {
-      "epoch": 0.6448047162859248,
-      "grad_norm": 65.59048461914062,
-      "learning_rate": 4.193994104642594e-08,
-      "loss": 0.2751,
+      "epoch": 0.37329351535836175,
+      "grad_norm": 15.805087089538574,
+      "learning_rate": 4.5333831058020476e-08,
+      "loss": 0.3242,
       "step": 3500
     },
     {
-      "epoch": 0.6448047162859248,
-      "eval_loss": 1.1169357299804688,
-      "eval_runtime": 76.2936,
-      "eval_samples_per_second": 15.82,
-      "eval_steps_per_second": 1.979,
+      "epoch": 0.37329351535836175,
+      "eval_loss": 1.195627212524414,
+      "eval_runtime": 63.4024,
+      "eval_samples_per_second": 15.567,
+      "eval_steps_per_second": 1.956,
       "step": 3500
     },
     {
-      "epoch": 0.7369196757553427,
-      "grad_norm": 18.106201171875,
-      "learning_rate": 4.078850405305821e-08,
-      "loss": 0.2919,
+      "epoch": 0.42662116040955633,
+      "grad_norm": 5.377908229827881,
+      "learning_rate": 4.4667235494880546e-08,
+      "loss": 0.2427,
       "step": 4000
     },
     {
-      "epoch": 0.7369196757553427,
-      "eval_loss": 1.0920671224594116,
-      "eval_runtime": 74.9101,
-      "eval_samples_per_second": 16.113,
-      "eval_steps_per_second": 2.016,
+      "epoch": 0.42662116040955633,
+      "eval_loss": 1.169384241104126,
+      "eval_runtime": 63.5082,
+      "eval_samples_per_second": 15.541,
+      "eval_steps_per_second": 1.953,
       "step": 4000
     },
     {
-      "epoch": 0.8290346352247605,
-      "grad_norm": 506.40570068359375,
-      "learning_rate": 3.9637067059690496e-08,
-      "loss": 0.2583,
+      "epoch": 0.47994880546075086,
+      "grad_norm": 150.5334930419922,
+      "learning_rate": 4.4000639931740615e-08,
+      "loss": 0.2993,
       "step": 4500
     },
     {
-      "epoch": 0.8290346352247605,
-      "eval_loss": 1.0721209049224854,
-      "eval_runtime": 75.8053,
-      "eval_samples_per_second": 15.922,
-      "eval_steps_per_second": 1.992,
+      "epoch": 0.47994880546075086,
+      "eval_loss": 1.145558476448059,
+      "eval_runtime": 62.3888,
+      "eval_samples_per_second": 15.82,
+      "eval_steps_per_second": 1.988,
       "step": 4500
     },
     {
-      "epoch": 0.9211495946941783,
-      "grad_norm": 1.1508910655975342,
-      "learning_rate": 3.848563006632277e-08,
-      "loss": 0.2679,
+      "epoch": 0.5332764505119454,
+      "grad_norm": 0.12246542423963547,
+      "learning_rate": 4.333404436860068e-08,
+      "loss": 0.3183,
       "step": 5000
     },
     {
-      "epoch": 0.9211495946941783,
-      "eval_loss": 1.0519349575042725,
-      "eval_runtime": 75.226,
-      "eval_samples_per_second": 16.045,
-      "eval_steps_per_second": 2.007,
+      "epoch": 0.5332764505119454,
+      "eval_loss": 1.1201218366622925,
+      "eval_runtime": 63.7154,
+      "eval_samples_per_second": 15.491,
+      "eval_steps_per_second": 1.946,
       "step": 5000
     },
     {
-      "epoch": 1.013264554163596,
-      "grad_norm": 0.1393290013074875,
-      "learning_rate": 3.733419307295505e-08,
-      "loss": 0.2472,
+      "epoch": 0.58660409556314,
+      "grad_norm": 0.3306196630001068,
+      "learning_rate": 4.266744880546075e-08,
+      "loss": 0.307,
       "step": 5500
     },
     {
-      "epoch": 1.013264554163596,
-      "eval_loss": 1.0355629920959473,
-      "eval_runtime": 74.6535,
-      "eval_samples_per_second": 16.168,
-      "eval_steps_per_second": 2.023,
+      "epoch": 0.58660409556314,
+      "eval_loss": 1.0982334613800049,
+      "eval_runtime": 62.4882,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.984,
       "step": 5500
     },
     {
-      "epoch": 1.105379513633014,
-      "grad_norm": 200.34664916992188,
-      "learning_rate": 3.6182756079587324e-08,
-      "loss": 0.26,
+      "epoch": 0.6399317406143344,
+      "grad_norm": 149.38189697265625,
+      "learning_rate": 4.2000853242320816e-08,
+      "loss": 0.2638,
       "step": 6000
     },
     {
-      "epoch": 1.105379513633014,
-      "eval_loss": 1.0176944732666016,
-      "eval_runtime": 74.4374,
-      "eval_samples_per_second": 16.215,
-      "eval_steps_per_second": 2.029,
+      "epoch": 0.6399317406143344,
+      "eval_loss": 1.0779541730880737,
+      "eval_runtime": 63.4101,
+      "eval_samples_per_second": 15.565,
+      "eval_steps_per_second": 1.956,
       "step": 6000
     },
     {
-      "epoch": 1.1974944731024317,
-      "grad_norm": 431.1804504394531,
-      "learning_rate": 3.50313190862196e-08,
-      "loss": 0.2153,
+      "epoch": 0.693259385665529,
+      "grad_norm": 353.0036315917969,
+      "learning_rate": 4.1334257679180886e-08,
+      "loss": 0.2226,
       "step": 6500
     },
     {
-      "epoch": 1.1974944731024317,
-      "eval_loss": 1.0044586658477783,
-      "eval_runtime": 74.3247,
-      "eval_samples_per_second": 16.24,
-      "eval_steps_per_second": 2.032,
+      "epoch": 0.693259385665529,
+      "eval_loss": 1.0613036155700684,
+      "eval_runtime": 62.8292,
+      "eval_samples_per_second": 15.709,
+      "eval_steps_per_second": 1.974,
       "step": 6500
     },
     {
-      "epoch": 1.2896094325718497,
-      "grad_norm": 0.6204941272735596,
-      "learning_rate": 3.387988209285188e-08,
-      "loss": 0.1791,
+      "epoch": 0.7465870307167235,
+      "grad_norm": 0.13188917934894562,
+      "learning_rate": 4.0667662116040955e-08,
+      "loss": 0.2453,
       "step": 7000
     },
     {
-      "epoch": 1.2896094325718497,
-      "eval_loss": 0.9926707148551941,
-      "eval_runtime": 75.7317,
-      "eval_samples_per_second": 15.938,
-      "eval_steps_per_second": 1.994,
+      "epoch": 0.7465870307167235,
+      "eval_loss": 1.0443507432937622,
+      "eval_runtime": 62.7457,
+      "eval_samples_per_second": 15.73,
+      "eval_steps_per_second": 1.976,
       "step": 7000
     },
     {
-      "epoch": 1.3817243920412676,
-      "grad_norm": 52.27273178100586,
-      "learning_rate": 3.272844509948415e-08,
-      "loss": 0.2082,
+      "epoch": 0.7999146757679181,
+      "grad_norm": 59.391639709472656,
+      "learning_rate": 4.000106655290102e-08,
+      "loss": 0.272,
       "step": 7500
     },
     {
-      "epoch": 1.3817243920412676,
-      "eval_loss": 0.9803809523582458,
-      "eval_runtime": 74.9936,
-      "eval_samples_per_second": 16.095,
-      "eval_steps_per_second": 2.014,
+      "epoch": 0.7999146757679181,
+      "eval_loss": 1.0300519466400146,
+      "eval_runtime": 62.3719,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.988,
       "step": 7500
     },
     {
-      "epoch": 1.4738393515106853,
-      "grad_norm": 0.0976635068655014,
-      "learning_rate": 3.1577008106116435e-08,
-      "loss": 0.196,
+      "epoch": 0.8532423208191127,
+      "grad_norm": 495.4797668457031,
+      "learning_rate": 3.9334470989761093e-08,
+      "loss": 0.283,
       "step": 8000
     },
     {
-      "epoch": 1.4738393515106853,
-      "eval_loss": 0.9711655378341675,
-      "eval_runtime": 74.3537,
-      "eval_samples_per_second": 16.233,
-      "eval_steps_per_second": 2.031,
+      "epoch": 0.8532423208191127,
+      "eval_loss": 1.0166898965835571,
+      "eval_runtime": 62.0338,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.999,
       "step": 8000
     },
     {
-      "epoch": 1.565954310980103,
-      "grad_norm": 46.64162826538086,
-      "learning_rate": 3.042557111274871e-08,
-      "loss": 0.1946,
+      "epoch": 0.9065699658703071,
+      "grad_norm": 159.0310821533203,
+      "learning_rate": 3.866787542662116e-08,
+      "loss": 0.2331,
       "step": 8500
     },
     {
-      "epoch": 1.565954310980103,
-      "eval_loss": 0.9621181488037109,
-      "eval_runtime": 75.7205,
-      "eval_samples_per_second": 15.94,
-      "eval_steps_per_second": 1.994,
+      "epoch": 0.9065699658703071,
+      "eval_loss": 1.003544807434082,
+      "eval_runtime": 62.9704,
+      "eval_samples_per_second": 15.674,
+      "eval_steps_per_second": 1.969,
       "step": 8500
     },
     {
-      "epoch": 1.658069270449521,
-      "grad_norm": 0.05690698325634003,
-      "learning_rate": 2.927413411938099e-08,
-      "loss": 0.2422,
+      "epoch": 0.9598976109215017,
+      "grad_norm": 45.40327835083008,
+      "learning_rate": 3.8001279863481225e-08,
+      "loss": 0.2362,
       "step": 9000
     },
     {
-      "epoch": 1.658069270449521,
-      "eval_loss": 0.9536784291267395,
-      "eval_runtime": 75.1745,
-      "eval_samples_per_second": 16.056,
-      "eval_steps_per_second": 2.009,
+      "epoch": 0.9598976109215017,
+      "eval_loss": 0.992453396320343,
+      "eval_runtime": 63.1405,
+      "eval_samples_per_second": 15.632,
+      "eval_steps_per_second": 1.964,
       "step": 9000
     },
     {
-      "epoch": 1.750184229918939,
-      "grad_norm": 0.08470064401626587,
-      "learning_rate": 2.8122697126013263e-08,
-      "loss": 0.2106,
+      "epoch": 1.0132252559726962,
+      "grad_norm": 621.8538818359375,
+      "learning_rate": 3.7334684300341295e-08,
+      "loss": 0.2396,
       "step": 9500
     },
     {
-      "epoch": 1.750184229918939,
-      "eval_loss": 0.9457550644874573,
-      "eval_runtime": 74.5062,
-      "eval_samples_per_second": 16.2,
-      "eval_steps_per_second": 2.027,
+      "epoch": 1.0132252559726962,
+      "eval_loss": 0.9829725027084351,
+      "eval_runtime": 62.3244,
+      "eval_samples_per_second": 15.836,
+      "eval_steps_per_second": 1.99,
       "step": 9500
     },
     {
-      "epoch": 1.8422991893883567,
-      "grad_norm": 1.5745967626571655,
-      "learning_rate": 2.697126013264554e-08,
-      "loss": 0.1801,
+      "epoch": 1.0665529010238908,
+      "grad_norm": 461.60736083984375,
+      "learning_rate": 3.6668088737201364e-08,
+      "loss": 0.2013,
       "step": 10000
     },
     {
-      "epoch": 1.8422991893883567,
-      "eval_loss": 0.9393072724342346,
-      "eval_runtime": 75.9432,
-      "eval_samples_per_second": 15.893,
-      "eval_steps_per_second": 1.988,
+      "epoch": 1.0665529010238908,
+      "eval_loss": 0.9735796451568604,
+      "eval_runtime": 63.0763,
+      "eval_samples_per_second": 15.648,
+      "eval_steps_per_second": 1.966,
       "step": 10000
     },
     {
-      "epoch": 1.9344141488577744,
-      "grad_norm": 0.06661002337932587,
-      "learning_rate": 2.5819823139277818e-08,
-      "loss": 0.2117,
+      "epoch": 1.1198805460750854,
+      "grad_norm": 7.0309906005859375,
+      "learning_rate": 3.6001493174061433e-08,
+      "loss": 0.2082,
       "step": 10500
     },
     {
-      "epoch": 1.9344141488577744,
-      "eval_loss": 0.9308408498764038,
-      "eval_runtime": 75.0811,
-      "eval_samples_per_second": 16.076,
-      "eval_steps_per_second": 2.011,
+      "epoch": 1.1198805460750854,
+      "eval_loss": 0.9639086127281189,
+      "eval_runtime": 63.0238,
+      "eval_samples_per_second": 15.661,
+      "eval_steps_per_second": 1.968,
       "step": 10500
     },
     {
-      "epoch": 2.026529108327192,
-      "grad_norm": 0.236691415309906,
-      "learning_rate": 2.4668386145910094e-08,
-      "loss": 0.2061,
+      "epoch": 1.17320819112628,
+      "grad_norm": 304.01995849609375,
+      "learning_rate": 3.5334897610921496e-08,
+      "loss": 0.2023,
       "step": 11000
     },
     {
-      "epoch": 2.026529108327192,
-      "eval_loss": 0.9237484335899353,
-      "eval_runtime": 74.6134,
-      "eval_samples_per_second": 16.177,
-      "eval_steps_per_second": 2.024,
+      "epoch": 1.17320819112628,
+      "eval_loss": 0.9558469653129578,
+      "eval_runtime": 62.1866,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.994,
       "step": 11000
     },
     {
-      "epoch": 2.1186440677966103,
-      "grad_norm": 8.831416130065918,
-      "learning_rate": 2.3516949152542374e-08,
-      "loss": 0.1878,
+      "epoch": 1.2265358361774745,
+      "grad_norm": 0.0017981340643018484,
+      "learning_rate": 3.4668302047781565e-08,
+      "loss": 0.2331,
       "step": 11500
     },
     {
-      "epoch": 2.1186440677966103,
-      "eval_loss": 0.9166750311851501,
-      "eval_runtime": 75.6886,
-      "eval_samples_per_second": 15.947,
-      "eval_steps_per_second": 1.995,
+      "epoch": 1.2265358361774745,
+      "eval_loss": 0.9465070962905884,
+      "eval_runtime": 63.1547,
+      "eval_samples_per_second": 15.628,
+      "eval_steps_per_second": 1.963,
       "step": 11500
     },
     {
-      "epoch": 2.210759027266028,
-      "grad_norm": 0.787421464920044,
-      "learning_rate": 2.236551215917465e-08,
-      "loss": 0.1655,
+      "epoch": 1.2798634812286689,
+      "grad_norm": 0.8147886395454407,
+      "learning_rate": 3.4001706484641635e-08,
+      "loss": 0.1784,
       "step": 12000
     },
     {
-      "epoch": 2.210759027266028,
-      "eval_loss": 0.9109458327293396,
-      "eval_runtime": 74.9866,
-      "eval_samples_per_second": 16.096,
-      "eval_steps_per_second": 2.014,
+      "epoch": 1.2798634812286689,
+      "eval_loss": 0.9392343163490295,
+      "eval_runtime": 62.9221,
+      "eval_samples_per_second": 15.686,
+      "eval_steps_per_second": 1.971,
       "step": 12000
     },
     {
-      "epoch": 2.3028739867354457,
-      "grad_norm": 535.4310913085938,
-      "learning_rate": 2.1214075165806926e-08,
-      "loss": 0.1946,
+      "epoch": 1.3331911262798635,
+      "grad_norm": 0.01276802271604538,
+      "learning_rate": 3.3335110921501704e-08,
+      "loss": 0.1953,
       "step": 12500
     },
     {
-      "epoch": 2.3028739867354457,
-      "eval_loss": 0.9070548415184021,
-      "eval_runtime": 74.3796,
-      "eval_samples_per_second": 16.228,
-      "eval_steps_per_second": 2.03,
+      "epoch": 1.3331911262798635,
+      "eval_loss": 0.9316345453262329,
+      "eval_runtime": 62.2816,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.991,
       "step": 12500
     },
     {
-      "epoch": 2.3949889462048635,
-      "grad_norm": 0.8420014381408691,
-      "learning_rate": 2.0062638172439202e-08,
-      "loss": 0.1882,
+      "epoch": 1.386518771331058,
+      "grad_norm": 0.07341516762971878,
+      "learning_rate": 3.2668515358361773e-08,
+      "loss": 0.1867,
       "step": 13000
     },
     {
-      "epoch": 2.3949889462048635,
-      "eval_loss": 0.9021016359329224,
-      "eval_runtime": 75.8812,
-      "eval_samples_per_second": 15.906,
-      "eval_steps_per_second": 1.99,
+      "epoch": 1.386518771331058,
+      "eval_loss": 0.9270105957984924,
+      "eval_runtime": 63.2619,
+      "eval_samples_per_second": 15.602,
+      "eval_steps_per_second": 1.96,
       "step": 13000
     },
     {
-      "epoch": 2.4871039056742816,
-      "grad_norm": 1.4041686058044434,
-      "learning_rate": 1.891120117907148e-08,
-      "loss": 0.1871,
+      "epoch": 1.4398464163822526,
+      "grad_norm": 17.55596923828125,
+      "learning_rate": 3.200191979522184e-08,
+      "loss": 0.22,
       "step": 13500
     },
     {
-      "epoch": 2.4871039056742816,
-      "eval_loss": 0.8960274457931519,
-      "eval_runtime": 75.2283,
-      "eval_samples_per_second": 16.044,
-      "eval_steps_per_second": 2.007,
+      "epoch": 1.4398464163822526,
+      "eval_loss": 0.919725775718689,
+      "eval_runtime": 63.3459,
+      "eval_samples_per_second": 15.581,
+      "eval_steps_per_second": 1.958,
       "step": 13500
     },
     {
-      "epoch": 2.5792188651436994,
-      "grad_norm": 327.2938537597656,
-      "learning_rate": 1.7759764185703757e-08,
-      "loss": 0.1419,
+      "epoch": 1.493174061433447,
+      "grad_norm": 3.4780771732330322,
+      "learning_rate": 3.133532423208191e-08,
+      "loss": 0.1656,
       "step": 14000
     },
     {
-      "epoch": 2.5792188651436994,
-      "eval_loss": 0.8913142085075378,
-      "eval_runtime": 74.4503,
-      "eval_samples_per_second": 16.212,
-      "eval_steps_per_second": 2.028,
+      "epoch": 1.493174061433447,
+      "eval_loss": 0.9148014783859253,
+      "eval_runtime": 62.257,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.992,
       "step": 14000
     },
     {
-      "epoch": 2.671333824613117,
-      "grad_norm": 0.01779676228761673,
-      "learning_rate": 1.6608327192336037e-08,
-      "loss": 0.1431,
+      "epoch": 1.5465017064846416,
+      "grad_norm": 158.22390747070312,
+      "learning_rate": 3.066872866894198e-08,
+      "loss": 0.1968,
       "step": 14500
     },
     {
-      "epoch": 2.671333824613117,
-      "eval_loss": 0.8879275321960449,
-      "eval_runtime": 75.8636,
-      "eval_samples_per_second": 15.91,
-      "eval_steps_per_second": 1.99,
+      "epoch": 1.5465017064846416,
+      "eval_loss": 0.9096349477767944,
+      "eval_runtime": 63.3671,
+      "eval_samples_per_second": 15.576,
+      "eval_steps_per_second": 1.957,
       "step": 14500
     },
     {
-      "epoch": 2.7634487840825352,
-      "grad_norm": 1.6428073644638062,
-      "learning_rate": 1.5456890198968313e-08,
-      "loss": 0.1811,
+      "epoch": 1.5998293515358362,
+      "grad_norm": 621.3922729492188,
+      "learning_rate": 3.0002133105802044e-08,
+      "loss": 0.1676,
       "step": 15000
     },
     {
-      "epoch": 2.7634487840825352,
-      "eval_loss": 0.8848333954811096,
-      "eval_runtime": 75.1911,
-      "eval_samples_per_second": 16.052,
-      "eval_steps_per_second": 2.008,
+      "epoch": 1.5998293515358362,
+      "eval_loss": 0.9056846499443054,
+      "eval_runtime": 63.2369,
+      "eval_samples_per_second": 15.608,
+      "eval_steps_per_second": 1.961,
       "step": 15000
     },
     {
-      "epoch": 2.855563743551953,
-      "grad_norm": 0.0006026208866387606,
-      "learning_rate": 1.430545320560059e-08,
-      "loss": 0.1694,
+      "epoch": 1.6531569965870307,
+      "grad_norm": 26.23398208618164,
+      "learning_rate": 2.9335537542662113e-08,
+      "loss": 0.2074,
       "step": 15500
     },
     {
-      "epoch": 2.855563743551953,
-      "eval_loss": 0.8826532959938049,
-      "eval_runtime": 74.6044,
-      "eval_samples_per_second": 16.179,
-      "eval_steps_per_second": 2.024,
+      "epoch": 1.6531569965870307,
+      "eval_loss": 0.8993550539016724,
+      "eval_runtime": 62.3767,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.988,
       "step": 15500
     },
     {
-      "epoch": 2.9476787030213707,
-      "grad_norm": 226.09982299804688,
-      "learning_rate": 1.3154016212232866e-08,
-      "loss": 0.1718,
+      "epoch": 1.7064846416382253,
+      "grad_norm": 283.3314514160156,
+      "learning_rate": 2.8668941979522183e-08,
+      "loss": 0.1847,
       "step": 16000
     },
     {
-      "epoch": 2.9476787030213707,
-      "eval_loss": 0.8798421621322632,
-      "eval_runtime": 76.0825,
-      "eval_samples_per_second": 15.864,
-      "eval_steps_per_second": 1.985,
+      "epoch": 1.7064846416382253,
+      "eval_loss": 0.8954482078552246,
+      "eval_runtime": 63.2168,
+      "eval_samples_per_second": 15.613,
+      "eval_steps_per_second": 1.962,
       "step": 16000
     },
     {
-      "epoch": 3.0397936624907884,
-      "grad_norm": 5.1134706154698506e-05,
-      "learning_rate": 1.2002579218865142e-08,
-      "loss": 0.153,
+      "epoch": 1.75981228668942,
+      "grad_norm": 12.151030540466309,
+      "learning_rate": 2.800234641638225e-08,
+      "loss": 0.1845,
       "step": 16500
     },
     {
-      "epoch": 3.0397936624907884,
-      "eval_loss": 0.8776962757110596,
-      "eval_runtime": 75.3205,
-      "eval_samples_per_second": 16.025,
-      "eval_steps_per_second": 2.005,
+      "epoch": 1.75981228668942,
+      "eval_loss": 0.8900338411331177,
+      "eval_runtime": 63.5581,
+      "eval_samples_per_second": 15.529,
+      "eval_steps_per_second": 1.951,
       "step": 16500
     },
     {
-      "epoch": 3.131908621960206,
-      "grad_norm": 0.017252299934625626,
-      "learning_rate": 1.085114222549742e-08,
-      "loss": 0.1715,
+      "epoch": 1.8131399317406145,
+      "grad_norm": 0.7503092288970947,
+      "learning_rate": 2.7335750853242318e-08,
+      "loss": 0.1721,
       "step": 17000
     },
     {
-      "epoch": 3.131908621960206,
-      "eval_loss": 0.8758523464202881,
-      "eval_runtime": 74.4544,
-      "eval_samples_per_second": 16.211,
-      "eval_steps_per_second": 2.028,
+      "epoch": 1.8131399317406145,
+      "eval_loss": 0.8872589468955994,
+      "eval_runtime": 62.0282,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.999,
       "step": 17000
     },
     {
-      "epoch": 3.2240235814296243,
-      "grad_norm": 68.8683853149414,
-      "learning_rate": 9.699705232129698e-09,
-      "loss": 0.1558,
+      "epoch": 1.8664675767918089,
+      "grad_norm": 180.9355926513672,
+      "learning_rate": 2.6669155290102384e-08,
+      "loss": 0.2627,
       "step": 17500
     },
     {
-      "epoch": 3.2240235814296243,
-      "eval_loss": 0.874191403388977,
-      "eval_runtime": 75.7598,
-      "eval_samples_per_second": 15.932,
-      "eval_steps_per_second": 1.993,
+      "epoch": 1.8664675767918089,
+      "eval_loss": 0.8810189962387085,
+      "eval_runtime": 62.7212,
+      "eval_samples_per_second": 15.736,
+      "eval_steps_per_second": 1.977,
       "step": 17500
     },
     {
-      "epoch": 3.316138540899042,
-      "grad_norm": 0.791266918182373,
-      "learning_rate": 8.548268238761974e-09,
-      "loss": 0.1384,
+      "epoch": 1.9197952218430034,
+      "grad_norm": 0.10503419488668442,
+      "learning_rate": 2.600255972696246e-08,
+      "loss": 0.1623,
       "step": 18000
     },
     {
-      "epoch": 3.316138540899042,
-      "eval_loss": 0.8715030550956726,
-      "eval_runtime": 74.7336,
-      "eval_samples_per_second": 16.151,
-      "eval_steps_per_second": 2.021,
+      "epoch": 1.9197952218430034,
+      "eval_loss": 0.8773773908615112,
+      "eval_runtime": 62.9181,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.971,
       "step": 18000
     },
     {
-      "epoch": 3.4082535003684598,
-      "grad_norm": 6.072669982910156,
-      "learning_rate": 7.396831245394252e-09,
-      "loss": 0.1788,
+      "epoch": 1.9731228668941978,
+      "grad_norm": 477.76715087890625,
+      "learning_rate": 2.5335964163822526e-08,
+      "loss": 0.2162,
       "step": 18500
     },
     {
-      "epoch": 3.4082535003684598,
-      "eval_loss": 0.8695403933525085,
-      "eval_runtime": 75.2877,
-      "eval_samples_per_second": 16.032,
-      "eval_steps_per_second": 2.006,
+      "epoch": 1.9731228668941978,
+      "eval_loss": 0.8713328242301941,
+      "eval_runtime": 62.3477,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.989,
       "step": 18500
     },
     {
-      "epoch": 3.500368459837878,
-      "grad_norm": 0.43182799220085144,
-      "learning_rate": 6.2453942520265294e-09,
-      "loss": 0.1668,
+      "epoch": 2.0264505119453924,
+      "grad_norm": 222.06802368164062,
+      "learning_rate": 2.4669368600682595e-08,
+      "loss": 0.1802,
       "step": 19000
     },
     {
-      "epoch": 3.500368459837878,
-      "eval_loss": 0.8684694766998291,
-      "eval_runtime": 75.7597,
-      "eval_samples_per_second": 15.932,
-      "eval_steps_per_second": 1.993,
+      "epoch": 2.0264505119453924,
+      "eval_loss": 0.8679063320159912,
+      "eval_runtime": 63.0768,
+      "eval_samples_per_second": 15.648,
+      "eval_steps_per_second": 1.966,
       "step": 19000
     },
     {
-      "epoch": 3.5924834193072956,
-      "grad_norm": 0.01970759592950344,
-      "learning_rate": 5.093957258658806e-09,
-      "loss": 0.1697,
+      "epoch": 2.079778156996587,
+      "grad_norm": 48.63727569580078,
+      "learning_rate": 2.400277303754266e-08,
+      "loss": 0.179,
       "step": 19500
     },
     {
-      "epoch": 3.5924834193072956,
-      "eval_loss": 0.8674086928367615,
-      "eval_runtime": 74.5942,
-      "eval_samples_per_second": 16.181,
-      "eval_steps_per_second": 2.024,
+      "epoch": 2.079778156996587,
+      "eval_loss": 0.86326664686203,
+      "eval_runtime": 63.3068,
+      "eval_samples_per_second": 15.591,
+      "eval_steps_per_second": 1.959,
       "step": 19500
     },
     {
-      "epoch": 3.6845983787767134,
-      "grad_norm": 0.0002275915176142007,
-      "learning_rate": 3.942520265291083e-09,
-      "loss": 0.1764,
+      "epoch": 2.1331058020477816,
+      "grad_norm": 0.5827552080154419,
+      "learning_rate": 2.333617747440273e-08,
+      "loss": 0.1549,
       "step": 20000
     },
     {
-      "epoch": 3.6845983787767134,
-      "eval_loss": 0.8666148781776428,
-      "eval_runtime": 74.3994,
-      "eval_samples_per_second": 16.223,
-      "eval_steps_per_second": 2.03,
+      "epoch": 2.1331058020477816,
+      "eval_loss": 0.8606277108192444,
+      "eval_runtime": 62.3116,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.99,
       "step": 20000
     },
     {
-      "epoch": 3.776713338246131,
-      "grad_norm": 199.484375,
-      "learning_rate": 2.79108327192336e-09,
-      "loss": 0.1417,
+      "epoch": 2.186433447098976,
+      "grad_norm": 0.060927845537662506,
+      "learning_rate": 2.2669581911262796e-08,
+      "loss": 0.1742,
       "step": 20500
     },
     {
-      "epoch": 3.776713338246131,
-      "eval_loss": 0.8660197854042053,
-      "eval_runtime": 75.6487,
-      "eval_samples_per_second": 15.955,
-      "eval_steps_per_second": 1.996,
+      "epoch": 2.186433447098976,
+      "eval_loss": 0.8584741353988647,
+      "eval_runtime": 63.1659,
+      "eval_samples_per_second": 15.626,
+      "eval_steps_per_second": 1.963,
       "step": 20500
     },
     {
-      "epoch": 3.868828297715549,
-      "grad_norm": 0.0007738114218227565,
-      "learning_rate": 1.6396462785556374e-09,
-      "loss": 0.1556,
+      "epoch": 2.2397610921501707,
+      "grad_norm": 16.044307708740234,
+      "learning_rate": 2.2002986348122866e-08,
+      "loss": 0.1448,
       "step": 21000
     },
     {
-      "epoch": 3.868828297715549,
-      "eval_loss": 0.8657393455505371,
-      "eval_runtime": 74.9437,
-      "eval_samples_per_second": 16.105,
-      "eval_steps_per_second": 2.015,
+      "epoch": 2.2397610921501707,
+      "eval_loss": 0.8546220064163208,
+      "eval_runtime": 63.0133,
+      "eval_samples_per_second": 15.663,
+      "eval_steps_per_second": 1.968,
       "step": 21000
     },
     {
-      "epoch": 3.960943257184967,
-      "grad_norm": 0.006366326939314604,
-      "learning_rate": 4.882092851879146e-10,
-      "loss": 0.1605,
+      "epoch": 2.2930887372013653,
+      "grad_norm": 0.07948703318834305,
+      "learning_rate": 2.1336390784982935e-08,
+      "loss": 0.2066,
       "step": 21500
     },
     {
-      "epoch": 3.960943257184967,
-      "eval_loss": 0.8656262755393982,
-      "eval_runtime": 74.4332,
-      "eval_samples_per_second": 16.216,
-      "eval_steps_per_second": 2.029,
+      "epoch": 2.2930887372013653,
+      "eval_loss": 0.8512936234474182,
+      "eval_runtime": 62.3887,
+      "eval_samples_per_second": 15.82,
+      "eval_steps_per_second": 1.988,
       "step": 21500
     },
+    {
+      "epoch": 2.34641638225256,
+      "grad_norm": 0.0013235179940238595,
+      "learning_rate": 2.0669795221843004e-08,
+      "loss": 0.1483,
+      "step": 22000
+    },
+    {
+      "epoch": 2.34641638225256,
+      "eval_loss": 0.8480631709098816,
+      "eval_runtime": 63.1614,
+      "eval_samples_per_second": 15.627,
+      "eval_steps_per_second": 1.963,
+      "step": 22000
+    },
+    {
+      "epoch": 2.399744027303754,
+      "grad_norm": 0.001996780978515744,
+      "learning_rate": 2.000319965870307e-08,
+      "loss": 0.1813,
+      "step": 22500
+    },
+    {
+      "epoch": 2.399744027303754,
+      "eval_loss": 0.84471195936203,
+      "eval_runtime": 63.0032,
+      "eval_samples_per_second": 15.666,
+      "eval_steps_per_second": 1.968,
+      "step": 22500
+    },
+    {
+      "epoch": 2.453071672354949,
+      "grad_norm": 0.24369017779827118,
+      "learning_rate": 1.933660409556314e-08,
+      "loss": 0.1617,
+      "step": 23000
+    },
+    {
+      "epoch": 2.453071672354949,
+      "eval_loss": 0.8411402702331543,
+      "eval_runtime": 62.707,
+      "eval_samples_per_second": 15.74,
+      "eval_steps_per_second": 1.977,
+      "step": 23000
+    },
+    {
+      "epoch": 2.506399317406143,
+      "grad_norm": 9.377041816711426,
+      "learning_rate": 1.8670008532423206e-08,
+      "loss": 0.1664,
+      "step": 23500
+    },
+    {
+      "epoch": 2.506399317406143,
+      "eval_loss": 0.8393682837486267,
+      "eval_runtime": 63.143,
+      "eval_samples_per_second": 15.631,
+      "eval_steps_per_second": 1.964,
+      "step": 23500
+    },
+    {
+      "epoch": 2.5597269624573378,
+      "grad_norm": 118.065673828125,
+      "learning_rate": 1.800341296928328e-08,
+      "loss": 0.1786,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5597269624573378,
+      "eval_loss": 0.8357825875282288,
+      "eval_runtime": 63.3183,
+      "eval_samples_per_second": 15.588,
+      "eval_steps_per_second": 1.958,
+      "step": 24000
+    },
+    {
+      "epoch": 2.6130546075085324,
+      "grad_norm": 3.0652036666870117,
+      "learning_rate": 1.7336817406143344e-08,
+      "loss": 0.1465,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6130546075085324,
+      "eval_loss": 0.8330481648445129,
+      "eval_runtime": 63.0565,
+      "eval_samples_per_second": 15.653,
+      "eval_steps_per_second": 1.966,
+      "step": 24500
+    },
+    {
+      "epoch": 2.666382252559727,
+      "grad_norm": 68.4100570678711,
+      "learning_rate": 1.6670221843003414e-08,
+      "loss": 0.1289,
+      "step": 25000
+    },
+    {
+      "epoch": 2.666382252559727,
+      "eval_loss": 0.831382691860199,
+      "eval_runtime": 62.8466,
+      "eval_samples_per_second": 15.705,
+      "eval_steps_per_second": 1.973,
+      "step": 25000
+    },
+    {
+      "epoch": 2.7197098976109215,
+      "grad_norm": 0.03573409095406532,
+      "learning_rate": 1.600362627986348e-08,
+      "loss": 0.1662,
+      "step": 25500
+    },
+    {
+      "epoch": 2.7197098976109215,
+      "eval_loss": 0.829609751701355,
+      "eval_runtime": 63.536,
+      "eval_samples_per_second": 15.535,
+      "eval_steps_per_second": 1.952,
+      "step": 25500
+    },
+    {
+      "epoch": 2.773037542662116,
+      "grad_norm": 85.1174087524414,
+      "learning_rate": 1.533703071672355e-08,
+      "loss": 0.1463,
+      "step": 26000
+    },
+    {
+      "epoch": 2.773037542662116,
+      "eval_loss": 0.8262338638305664,
+      "eval_runtime": 63.1492,
+      "eval_samples_per_second": 15.63,
+      "eval_steps_per_second": 1.964,
+      "step": 26000
+    },
+    {
+      "epoch": 2.8263651877133107,
+      "grad_norm": 1.988323450088501,
+      "learning_rate": 1.4670435153583615e-08,
+      "loss": 0.1471,
+      "step": 26500
+    },
+    {
+      "epoch": 2.8263651877133107,
+      "eval_loss": 0.8249030709266663,
+      "eval_runtime": 63.1063,
+      "eval_samples_per_second": 15.64,
+      "eval_steps_per_second": 1.965,
+      "step": 26500
+    },
+    {
+      "epoch": 2.8796928327645053,
+      "grad_norm": 14.553828239440918,
+      "learning_rate": 1.4003839590443686e-08,
+      "loss": 0.167,
+      "step": 27000
+    },
+    {
+      "epoch": 2.8796928327645053,
+      "eval_loss": 0.821887731552124,
+      "eval_runtime": 63.5261,
+      "eval_samples_per_second": 15.537,
+      "eval_steps_per_second": 1.952,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9330204778157,
+      "grad_norm": 0.4691038131713867,
+      "learning_rate": 1.3337244027303754e-08,
+      "loss": 0.1268,
+      "step": 27500
+    },
+    {
+      "epoch": 2.9330204778157,
+      "eval_loss": 0.8203587532043457,
+      "eval_runtime": 63.3176,
+      "eval_samples_per_second": 15.588,
+      "eval_steps_per_second": 1.958,
+      "step": 27500
+    },
+    {
+      "epoch": 2.986348122866894,
+      "grad_norm": 327.5917663574219,
+      "learning_rate": 1.2670648464163821e-08,
+      "loss": 0.177,
+      "step": 28000
+    },
+    {
+      "epoch": 2.986348122866894,
+      "eval_loss": 0.8176783919334412,
+      "eval_runtime": 63.0051,
+      "eval_samples_per_second": 15.665,
+      "eval_steps_per_second": 1.968,
+      "step": 28000
+    },
+    {
+      "epoch": 3.0396757679180886,
+      "grad_norm": 1.4327431472338503e-06,
+      "learning_rate": 1.200405290102389e-08,
+      "loss": 0.1206,
+      "step": 28500
+    },
+    {
+      "epoch": 3.0396757679180886,
+      "eval_loss": 0.8165684938430786,
+      "eval_runtime": 63.4667,
+      "eval_samples_per_second": 15.551,
+      "eval_steps_per_second": 1.954,
+      "step": 28500
+    },
+    {
+      "epoch": 3.093003412969283,
+      "grad_norm": 0.21452489495277405,
+      "learning_rate": 1.1337457337883958e-08,
+      "loss": 0.1345,
+      "step": 29000
+    },
+    {
+      "epoch": 3.093003412969283,
+      "eval_loss": 0.8155500292778015,
+      "eval_runtime": 63.0809,
+      "eval_samples_per_second": 15.647,
+      "eval_steps_per_second": 1.966,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1463310580204777,
+      "grad_norm": 8.160687446594238,
+      "learning_rate": 1.0670861774744026e-08,
+      "loss": 0.1907,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1463310580204777,
+      "eval_loss": 0.8143629431724548,
+      "eval_runtime": 62.8466,
+      "eval_samples_per_second": 15.705,
+      "eval_steps_per_second": 1.973,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1996587030716723,
+      "grad_norm": 0.04997281730175018,
+      "learning_rate": 1.0004266211604095e-08,
+      "loss": 0.1395,
+      "step": 30000
+    },
+    {
+      "epoch": 3.1996587030716723,
+      "eval_loss": 0.812623918056488,
+      "eval_runtime": 63.0553,
+      "eval_samples_per_second": 15.653,
+      "eval_steps_per_second": 1.967,
+      "step": 30000
+    },
+    {
+      "epoch": 3.252986348122867,
+      "grad_norm": 609.138916015625,
+      "learning_rate": 9.337670648464163e-09,
+      "loss": 0.1511,
+      "step": 30500
+    },
+    {
+      "epoch": 3.252986348122867,
+      "eval_loss": 0.8111589550971985,
+      "eval_runtime": 63.102,
+      "eval_samples_per_second": 15.641,
+      "eval_steps_per_second": 1.965,
+      "step": 30500
+    },
+    {
+      "epoch": 3.3063139931740615,
+      "grad_norm": 6.588051795959473,
+      "learning_rate": 8.671075085324232e-09,
+      "loss": 0.1334,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3063139931740615,
+      "eval_loss": 0.8101842403411865,
+      "eval_runtime": 62.7702,
+      "eval_samples_per_second": 15.724,
+      "eval_steps_per_second": 1.975,
+      "step": 31000
+    },
+    {
+      "epoch": 3.359641638225256,
+      "grad_norm": 0.0009110367391258478,
+      "learning_rate": 8.0044795221843e-09,
+      "loss": 0.1799,
+      "step": 31500
+    },
+    {
+      "epoch": 3.359641638225256,
+      "eval_loss": 0.8090225458145142,
+      "eval_runtime": 63.1535,
+      "eval_samples_per_second": 15.629,
+      "eval_steps_per_second": 1.963,
+      "step": 31500
+    },
+    {
+      "epoch": 3.4129692832764507,
+      "grad_norm": 0.1708969920873642,
+      "learning_rate": 7.3378839590443675e-09,
+      "loss": 0.1289,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4129692832764507,
+      "eval_loss": 0.8080681562423706,
+      "eval_runtime": 63.1177,
+      "eval_samples_per_second": 15.637,
+      "eval_steps_per_second": 1.965,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4662969283276452,
+      "grad_norm": 143.53611755371094,
+      "learning_rate": 6.671288395904437e-09,
+      "loss": 0.1545,
+      "step": 32500
+    },
+    {
+      "epoch": 3.4662969283276452,
+      "eval_loss": 0.8072041273117065,
+      "eval_runtime": 62.7779,
+      "eval_samples_per_second": 15.722,
+      "eval_steps_per_second": 1.975,
+      "step": 32500
+    },
+    {
+      "epoch": 3.51962457337884,
+      "grad_norm": 301.7650146484375,
+      "learning_rate": 6.0046928327645045e-09,
+      "loss": 0.1705,
+      "step": 33000
+    },
+    {
+      "epoch": 3.51962457337884,
+      "eval_loss": 0.8064139485359192,
+      "eval_runtime": 63.0778,
+      "eval_samples_per_second": 15.647,
+      "eval_steps_per_second": 1.966,
+      "step": 33000
+    },
+    {
+      "epoch": 3.572952218430034,
+      "grad_norm": 0.01895347237586975,
+      "learning_rate": 5.338097269624573e-09,
+      "loss": 0.1424,
+      "step": 33500
+    },
+    {
+      "epoch": 3.572952218430034,
+      "eval_loss": 0.8055066466331482,
+      "eval_runtime": 62.5718,
+      "eval_samples_per_second": 15.774,
+      "eval_steps_per_second": 1.982,
+      "step": 33500
+    },
+    {
+      "epoch": 3.6262798634812285,
+      "grad_norm": 924.8953857421875,
+      "learning_rate": 4.6715017064846414e-09,
+      "loss": 0.1873,
+      "step": 34000
+    },
+    {
+      "epoch": 3.6262798634812285,
+      "eval_loss": 0.8047504425048828,
+      "eval_runtime": 62.3396,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.989,
+      "step": 34000
+    },
+    {
+      "epoch": 3.679607508532423,
+      "grad_norm": 46.50166320800781,
+      "learning_rate": 4.00490614334471e-09,
+      "loss": 0.1432,
+      "step": 34500
+    },
+    {
+      "epoch": 3.679607508532423,
+      "eval_loss": 0.8042765259742737,
+      "eval_runtime": 62.8255,
+      "eval_samples_per_second": 15.71,
+      "eval_steps_per_second": 1.974,
+      "step": 34500
+    },
+    {
+      "epoch": 3.7329351535836177,
+      "grad_norm": 6.692138867947506e-06,
+      "learning_rate": 3.3383105802047776e-09,
+      "loss": 0.1485,
+      "step": 35000
+    },
+    {
+      "epoch": 3.7329351535836177,
+      "eval_loss": 0.8037394881248474,
+      "eval_runtime": 62.9403,
+      "eval_samples_per_second": 15.682,
+      "eval_steps_per_second": 1.97,
+      "step": 35000
+    },
+    {
+      "epoch": 3.7862627986348123,
+      "grad_norm": 31.358840942382812,
+      "learning_rate": 2.671715017064846e-09,
+      "loss": 0.1286,
+      "step": 35500
+    },
+    {
+      "epoch": 3.7862627986348123,
+      "eval_loss": 0.8033393025398254,
+      "eval_runtime": 62.6525,
+      "eval_samples_per_second": 15.754,
+      "eval_steps_per_second": 1.979,
+      "step": 35500
+    },
+    {
+      "epoch": 3.839590443686007,
+      "grad_norm": 6.9743475914001465,
+      "learning_rate": 2.0051194539249145e-09,
+      "loss": 0.1469,
+      "step": 36000
+    },
+    {
+      "epoch": 3.839590443686007,
+      "eval_loss": 0.8030404448509216,
+      "eval_runtime": 63.0643,
+      "eval_samples_per_second": 15.651,
+      "eval_steps_per_second": 1.966,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8929180887372015,
+      "grad_norm": 0.7286427021026611,
+      "learning_rate": 1.3385238907849828e-09,
+      "loss": 0.1708,
+      "step": 36500
+    },
+    {
+      "epoch": 3.8929180887372015,
+      "eval_loss": 0.8029284477233887,
+      "eval_runtime": 63.0742,
+      "eval_samples_per_second": 15.648,
+      "eval_steps_per_second": 1.966,
+      "step": 36500
+    },
+    {
+      "epoch": 3.946245733788396,
+      "grad_norm": 29.70534896850586,
+      "learning_rate": 6.719283276450512e-10,
+      "loss": 0.1226,
+      "step": 37000
+    },
+    {
+      "epoch": 3.946245733788396,
+      "eval_loss": 0.802849292755127,
+      "eval_runtime": 62.5897,
+      "eval_samples_per_second": 15.769,
+      "eval_steps_per_second": 1.981,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9995733788395906,
+      "grad_norm": 196.9478759765625,
+      "learning_rate": 5.332764505119454e-12,
+      "loss": 0.1549,
+      "step": 37500
+    },
+    {
+      "epoch": 3.9995733788395906,
+      "eval_loss": 0.8028334379196167,
+      "eval_runtime": 63.0184,
+      "eval_samples_per_second": 15.662,
+      "eval_steps_per_second": 1.968,
+      "step": 37500
+    },
     {
       "epoch": 4.0,
-      "step": 21712,
-      "total_flos": 7811792839259640.0,
-      "train_loss": 0.2173631976611841,
-      "train_runtime": 16487.0261,
-      "train_samples_per_second": 2.634,
-      "train_steps_per_second": 1.317
+      "step": 37504,
+      "total_flos": 1.349309672235756e+16,
+      "train_loss": 0.20338401917281557,
+      "train_runtime": 27801.3632,
+      "train_samples_per_second": 2.698,
+      "train_steps_per_second": 1.349
     }
   ],
   "logging_steps": 500,
-  "max_steps": 21712,
+  "max_steps": 37504,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 4,
   "save_steps": 500,
@@ -680,7 +1160,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 7811792839259640.0,
+  "total_flos": 1.349309672235756e+16,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null