Training in progress, step 120, checkpoint

checkpoint-120/model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9ac8a998e13d90f4c3f2c7e5fd26b6451ffa761a1498612b7fb15e35d3a7d61c
+oid sha256:508ee628ab4d6001f157613a92700a5eea00b0a38cee04ba241922505788866d
 size 4961251752

checkpoint-120/model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:210aa1b36169057d2a51c95e7654501879229b8a762d9cfe8b22dbdead3a2ef0
+oid sha256:eef615bff0f0ce15c4034027960eda4e8189e733a4d3cb84333a059e677c5888
 size 3639026128

checkpoint-120/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:25ddc0ce2b884dde2f529537240e6f2bca4381c267f77913d0c8b7e36104f0d1
+oid sha256:9808b73ad336f9770e0dc514b3abaa944d259e9efdfe989f7c6217f832d397d9
 size 17201097520

checkpoint-120/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:20544de5b6b8034ed1329fbd13f4ba2b0abccc58acd3f625d4f25f466b829254
+oid sha256:aaa59229f17456a3d03f4bfec1665afca0841cdd143273e8086fb67f6328f06d
 size 1465

checkpoint-120/trainer_state.json

@@ -2,7 +2,7 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.08301625735039779,
+  "epoch": 0.08210040194988455,
   "eval_steps": 10,
   "global_step": 120,
   "is_hyper_param_search": false,
@@ -10,260 +10,260 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "entropy": 2.010164797306061,
-      "epoch": 0.0069180214458664825,
-      "grad_norm": 213.0,
+      "entropy": 2.0267144709825518,
+      "epoch": 0.006841700162490379,
+      "grad_norm": 233.0,
       "learning_rate": 2.0454545454545457e-06,
-      "loss": 11.6494,
-      "mean_token_accuracy": 0.18611789774149656,
-      "num_tokens": 44798.0,
+      "loss": 11.475,
+      "mean_token_accuracy": 0.19283651001751423,
+      "num_tokens": 44918.0,
       "step": 10
     },
     {
-      "epoch": 0.0069180214458664825,
-      "eval_entropy": 2.2144161474704744,
-      "eval_loss": 9.646827697753906,
-      "eval_mean_token_accuracy": 0.2581794735789299,
-      "eval_num_tokens": 44798.0,
-      "eval_runtime": 30.0225,
-      "eval_samples_per_second": 6.662,
-      "eval_steps_per_second": 3.331,
+      "epoch": 0.006841700162490379,
+      "eval_entropy": 2.1978431594371797,
+      "eval_loss": 9.854133605957031,
+      "eval_mean_token_accuracy": 0.24782681614160537,
+      "eval_num_tokens": 44918.0,
+      "eval_runtime": 29.6063,
+      "eval_samples_per_second": 6.755,
+      "eval_steps_per_second": 3.378,
       "step": 10
     },
     {
-      "entropy": 2.8024502843618393,
-      "epoch": 0.013836042891732965,
-      "grad_norm": 87.0,
+      "entropy": 2.7814523041248322,
+      "epoch": 0.013683400324980758,
+      "grad_norm": 72.0,
       "learning_rate": 4.3181818181818185e-06,
-      "loss": 7.8303,
-      "mean_token_accuracy": 0.3611512377858162,
-      "num_tokens": 89551.0,
+      "loss": 7.8101,
+      "mean_token_accuracy": 0.3553792662918568,
+      "num_tokens": 89798.0,
       "step": 20
     },
     {
-      "epoch": 0.013836042891732965,
-      "eval_entropy": 4.162467215061188,
-      "eval_loss": 5.341795444488525,
-      "eval_mean_token_accuracy": 0.5161851558089257,
-      "eval_num_tokens": 89551.0,
-      "eval_runtime": 30.0237,
-      "eval_samples_per_second": 6.661,
-      "eval_steps_per_second": 3.331,
+      "epoch": 0.013683400324980758,
+      "eval_entropy": 4.193059599399566,
+      "eval_loss": 5.401592254638672,
+      "eval_mean_token_accuracy": 0.5131221789121628,
+      "eval_num_tokens": 89798.0,
+      "eval_runtime": 29.4863,
+      "eval_samples_per_second": 6.783,
+      "eval_steps_per_second": 3.391,
       "step": 20
     },
     {
-      "entropy": 5.205361837148667,
-      "epoch": 0.020754064337599448,
-      "grad_norm": 45.25,
+      "entropy": 5.28521448969841,
+      "epoch": 0.020525100487471137,
+      "grad_norm": 48.25,
       "learning_rate": 6.590909090909091e-06,
-      "loss": 3.8722,
-      "mean_token_accuracy": 0.5812089148908853,
-      "num_tokens": 134427.0,
+      "loss": 3.8489,
+      "mean_token_accuracy": 0.57691092453897,
+      "num_tokens": 134848.0,
       "step": 30
     },
     {
-      "epoch": 0.020754064337599448,
-      "eval_entropy": 5.786367483139038,
-      "eval_loss": 2.314438581466675,
-      "eval_mean_token_accuracy": 0.7237287980318069,
-      "eval_num_tokens": 134427.0,
-      "eval_runtime": 29.9714,
-      "eval_samples_per_second": 6.673,
-      "eval_steps_per_second": 3.337,
+      "epoch": 0.020525100487471137,
+      "eval_entropy": 6.013385376930237,
+      "eval_loss": 2.3701281547546387,
+      "eval_mean_token_accuracy": 0.7218476647138595,
+      "eval_num_tokens": 134848.0,
+      "eval_runtime": 30.3106,
+      "eval_samples_per_second": 6.598,
+      "eval_steps_per_second": 3.299,
       "step": 30
     },
     {
-      "entropy": 4.8913937240839,
-      "epoch": 0.02767208578346593,
-      "grad_norm": 47.0,
+      "entropy": 4.856430059671402,
+      "epoch": 0.027366800649961515,
+      "grad_norm": 35.75,
       "learning_rate": 8.863636363636365e-06,
-      "loss": 1.7587,
-      "mean_token_accuracy": 0.7653285600244999,
-      "num_tokens": 179334.0,
+      "loss": 1.7849,
+      "mean_token_accuracy": 0.7600100472569465,
+      "num_tokens": 179771.0,
       "step": 40
     },
     {
-      "epoch": 0.02767208578346593,
-      "eval_entropy": 4.2611740207672115,
-      "eval_loss": 1.2193154096603394,
-      "eval_mean_token_accuracy": 0.8546426892280579,
-      "eval_num_tokens": 179334.0,
-      "eval_runtime": 30.0206,
-      "eval_samples_per_second": 6.662,
-      "eval_steps_per_second": 3.331,
+      "epoch": 0.027366800649961515,
+      "eval_entropy": 3.8757058000564575,
+      "eval_loss": 1.2098770141601562,
+      "eval_mean_token_accuracy": 0.8486587703227997,
+      "eval_num_tokens": 179771.0,
+      "eval_runtime": 30.1855,
+      "eval_samples_per_second": 6.626,
+      "eval_steps_per_second": 3.313,
       "step": 40
     },
     {
-      "entropy": 2.854546858370304,
-      "epoch": 0.03459010722933241,
-      "grad_norm": 28.0,
-      "learning_rate": 9.964336661911556e-06,
-      "loss": 0.6938,
-      "mean_token_accuracy": 0.8879799745976925,
-      "num_tokens": 224191.0,
+      "entropy": 2.369047798216343,
+      "epoch": 0.03420850081245189,
+      "grad_norm": 39.5,
+      "learning_rate": 9.964739069111425e-06,
+      "loss": 0.6405,
+      "mean_token_accuracy": 0.8753940276801586,
+      "num_tokens": 224527.0,
       "step": 50
     },
     {
-      "epoch": 0.03459010722933241,
-      "eval_entropy": 2.2573838114738463,
-      "eval_loss": 0.31782782077789307,
-      "eval_mean_token_accuracy": 0.8828409242630005,
-      "eval_num_tokens": 224191.0,
-      "eval_runtime": 30.357,
-      "eval_samples_per_second": 6.588,
-      "eval_steps_per_second": 3.294,
+      "epoch": 0.03420850081245189,
+      "eval_entropy": 1.5494409036636352,
+      "eval_loss": 0.2679590880870819,
+      "eval_mean_token_accuracy": 0.8952399933338165,
+      "eval_num_tokens": 224527.0,
+      "eval_runtime": 31.2875,
+      "eval_samples_per_second": 6.392,
+      "eval_steps_per_second": 3.196,
       "step": 50
     },
     {
-      "entropy": 2.4331204503774644,
-      "epoch": 0.041508128675198895,
-      "grad_norm": 13.6875,
-      "learning_rate": 9.893009985734665e-06,
-      "loss": 0.3026,
-      "mean_token_accuracy": 0.8833079569041729,
-      "num_tokens": 269084.0,
+      "entropy": 1.61202352643013,
+      "epoch": 0.041050200974942275,
+      "grad_norm": 19.25,
+      "learning_rate": 9.894217207334274e-06,
+      "loss": 0.244,
+      "mean_token_accuracy": 0.8936950318515301,
+      "num_tokens": 269230.0,
       "step": 60
     },
     {
-      "epoch": 0.041508128675198895,
-      "eval_entropy": 2.460804523229599,
-      "eval_loss": 0.23251518607139587,
-      "eval_mean_token_accuracy": 0.8987711769342422,
-      "eval_num_tokens": 269084.0,
-      "eval_runtime": 29.9779,
-      "eval_samples_per_second": 6.672,
-      "eval_steps_per_second": 3.336,
+      "epoch": 0.041050200974942275,
+      "eval_entropy": 2.1578941643238068,
+      "eval_loss": 0.24808701872825623,
+      "eval_mean_token_accuracy": 0.8934747791290283,
+      "eval_num_tokens": 269230.0,
+      "eval_runtime": 30.4369,
+      "eval_samples_per_second": 6.571,
+      "eval_steps_per_second": 3.285,
       "step": 60
     },
     {
-      "entropy": 2.5201364308595657,
-      "epoch": 0.048426150121065374,
-      "grad_norm": 6.6875,
-      "learning_rate": 9.821683309557775e-06,
-      "loss": 0.2343,
-      "mean_token_accuracy": 0.8990107223391532,
-      "num_tokens": 313857.0,
+      "entropy": 2.500882962346077,
+      "epoch": 0.04789190113743265,
+      "grad_norm": 16.125,
+      "learning_rate": 9.823695345557123e-06,
+      "loss": 0.2284,
+      "mean_token_accuracy": 0.8963869497179985,
+      "num_tokens": 314084.0,
       "step": 70
     },
     {
-      "epoch": 0.048426150121065374,
-      "eval_entropy": 2.1981401789188384,
-      "eval_loss": 0.2192859798669815,
-      "eval_mean_token_accuracy": 0.8919178038835526,
-      "eval_num_tokens": 313857.0,
-      "eval_runtime": 29.9901,
-      "eval_samples_per_second": 6.669,
-      "eval_steps_per_second": 3.334,
+      "epoch": 0.04789190113743265,
+      "eval_entropy": 2.796394021511078,
+      "eval_loss": 0.21758534014225006,
+      "eval_mean_token_accuracy": 0.9052125871181488,
+      "eval_num_tokens": 314084.0,
+      "eval_runtime": 29.9526,
+      "eval_samples_per_second": 6.677,
+      "eval_steps_per_second": 3.339,
       "step": 70
     },
     {
-      "entropy": 2.003435145318508,
-      "epoch": 0.05534417156693186,
-      "grad_norm": 7.28125,
-      "learning_rate": 9.750356633380886e-06,
-      "loss": 0.2104,
-      "mean_token_accuracy": 0.8997310034930706,
-      "num_tokens": 358622.0,
+      "entropy": 2.7663454830646517,
+      "epoch": 0.05473360129992303,
+      "grad_norm": 10.0625,
+      "learning_rate": 9.753173483779972e-06,
+      "loss": 0.2122,
+      "mean_token_accuracy": 0.901957081258297,
+      "num_tokens": 359026.0,
       "step": 80
     },
     {
-      "epoch": 0.05534417156693186,
-      "eval_entropy": 1.9411760997772216,
-      "eval_loss": 0.19223187863826752,
-      "eval_mean_token_accuracy": 0.9055801618099213,
-      "eval_num_tokens": 358622.0,
-      "eval_runtime": 29.9158,
-      "eval_samples_per_second": 6.685,
-      "eval_steps_per_second": 3.343,
+      "epoch": 0.05473360129992303,
+      "eval_entropy": 2.448452844619751,
+      "eval_loss": 0.19542743265628815,
+      "eval_mean_token_accuracy": 0.9078808641433715,
+      "eval_num_tokens": 359026.0,
+      "eval_runtime": 29.8422,
+      "eval_samples_per_second": 6.702,
+      "eval_steps_per_second": 3.351,
       "step": 80
     },
     {
-      "entropy": 1.8479327753186225,
-      "epoch": 0.06226219301279834,
-      "grad_norm": 7.8125,
-      "learning_rate": 9.679029957203994e-06,
-      "loss": 0.1831,
-      "mean_token_accuracy": 0.9064941309392452,
-      "num_tokens": 403518.0,
+      "entropy": 1.9316013976931572,
+      "epoch": 0.06157530146241341,
+      "grad_norm": 9.125,
+      "learning_rate": 9.682651622002823e-06,
+      "loss": 0.1996,
+      "mean_token_accuracy": 0.9041122056543827,
+      "num_tokens": 403675.0,
       "step": 90
     },
     {
-      "epoch": 0.06226219301279834,
-      "eval_entropy": 1.7790459179878235,
-      "eval_loss": 0.1928865760564804,
-      "eval_mean_token_accuracy": 0.9028073137998581,
-      "eval_num_tokens": 403518.0,
-      "eval_runtime": 30.3798,
-      "eval_samples_per_second": 6.583,
-      "eval_steps_per_second": 3.292,
+      "epoch": 0.06157530146241341,
+      "eval_entropy": 1.6268229800462723,
+      "eval_loss": 0.18592144548892975,
+      "eval_mean_token_accuracy": 0.9096919119358062,
+      "eval_num_tokens": 403675.0,
+      "eval_runtime": 30.5761,
+      "eval_samples_per_second": 6.541,
+      "eval_steps_per_second": 3.271,
       "step": 90
     },
     {
-      "entropy": 1.7204390287399292,
-      "epoch": 0.06918021445866482,
-      "grad_norm": 7.96875,
-      "learning_rate": 9.607703281027105e-06,
-      "loss": 0.177,
-      "mean_token_accuracy": 0.9128127194941044,
-      "num_tokens": 448405.0,
+      "entropy": 1.6001751273870468,
+      "epoch": 0.06841700162490379,
+      "grad_norm": 12.875,
+      "learning_rate": 9.61212976022567e-06,
+      "loss": 0.1915,
+      "mean_token_accuracy": 0.9078994408249855,
+      "num_tokens": 448529.0,
       "step": 100
     },
     {
-      "epoch": 0.06918021445866482,
-      "eval_entropy": 1.5752793765068054,
-      "eval_loss": 0.18161609768867493,
-      "eval_mean_token_accuracy": 0.9143004041910171,
-      "eval_num_tokens": 448405.0,
-      "eval_runtime": 29.9547,
-      "eval_samples_per_second": 6.677,
-      "eval_steps_per_second": 3.338,
+      "epoch": 0.06841700162490379,
+      "eval_entropy": 1.66984717130661,
+      "eval_loss": 0.20213660597801208,
+      "eval_mean_token_accuracy": 0.8984966450929641,
+      "eval_num_tokens": 448529.0,
+      "eval_runtime": 29.5976,
+      "eval_samples_per_second": 6.757,
+      "eval_steps_per_second": 3.379,
       "step": 100
     },
     {
-      "entropy": 1.552296456694603,
-      "epoch": 0.0760982359045313,
-      "grad_norm": 6.3125,
-      "learning_rate": 9.536376604850215e-06,
-      "loss": 0.1842,
-      "mean_token_accuracy": 0.9075778618454933,
-      "num_tokens": 493303.0,
+      "entropy": 1.4694452270865441,
+      "epoch": 0.07525870178739416,
+      "grad_norm": 8.125,
+      "learning_rate": 9.541607898448519e-06,
+      "loss": 0.1825,
+      "mean_token_accuracy": 0.9069189541041851,
+      "num_tokens": 493396.0,
       "step": 110
     },
     {
-      "epoch": 0.0760982359045313,
-      "eval_entropy": 1.6148131275177002,
-      "eval_loss": 0.17123353481292725,
-      "eval_mean_token_accuracy": 0.9093608874082565,
-      "eval_num_tokens": 493303.0,
-      "eval_runtime": 29.9443,
-      "eval_samples_per_second": 6.679,
-      "eval_steps_per_second": 3.34,
+      "epoch": 0.07525870178739416,
+      "eval_entropy": 1.3290829557180404,
+      "eval_loss": 0.18572577834129333,
+      "eval_mean_token_accuracy": 0.9053398305177689,
+      "eval_num_tokens": 493396.0,
+      "eval_runtime": 29.8751,
+      "eval_samples_per_second": 6.695,
+      "eval_steps_per_second": 3.347,
       "step": 110
     },
     {
-      "entropy": 1.6193801417946816,
-      "epoch": 0.08301625735039779,
-      "grad_norm": 4.84375,
-      "learning_rate": 9.465049928673324e-06,
-      "loss": 0.1659,
-      "mean_token_accuracy": 0.9159238807857036,
-      "num_tokens": 538145.0,
+      "entropy": 1.4422638706862927,
+      "epoch": 0.08210040194988455,
+      "grad_norm": 9.5,
+      "learning_rate": 9.471086036671368e-06,
+      "loss": 0.1902,
+      "mean_token_accuracy": 0.9103416793048382,
+      "num_tokens": 538289.0,
       "step": 120
     },
     {
-      "epoch": 0.08301625735039779,
-      "eval_entropy": 1.6491443312168121,
-      "eval_loss": 0.17252139747142792,
-      "eval_mean_token_accuracy": 0.9111752665042877,
-      "eval_num_tokens": 538145.0,
-      "eval_runtime": 29.8721,
-      "eval_samples_per_second": 6.695,
-      "eval_steps_per_second": 3.348,
+      "epoch": 0.08210040194988455,
+      "eval_entropy": 1.5916276335716248,
+      "eval_loss": 0.17153747379779816,
+      "eval_mean_token_accuracy": 0.9156510829925537,
+      "eval_num_tokens": 538289.0,
+      "eval_runtime": 29.7995,
+      "eval_samples_per_second": 6.712,
+      "eval_steps_per_second": 3.356,
       "step": 120
     }
   ],
   "logging_steps": 10,
-  "max_steps": 1446,
+  "max_steps": 1462,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 40,
@@ -279,7 +279,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.18447024765872e+16,
+  "total_flos": 1.18616633038992e+16,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null

checkpoint-120/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:448c687cf1191e057c505ddb0360ab17fa030e53c92c4164b125a9718c9e235e
+oid sha256:0ce48f4ac927f92a60fe16e448e6e50e0778434470463d98a397d7e6523fe160
 size 6481