End of training

all_results.json

@@ -1,15 +1,15 @@
 {
     "epoch": 10.0,
-    "eval_accuracy": 0.4868845302152862,
-    "eval_loss": 2.5603630542755127,
-    "eval_runtime": 4.1602,
+    "eval_accuracy": 0.4995467750899875,
+    "eval_loss": 2.4751720428466797,
+    "eval_runtime": 3.3732,
     "eval_samples": 4491,
-    "eval_samples_per_second": 1079.517,
-    "eval_steps_per_second": 2.163,
-    "perplexity": 12.940514571954653,
-    "train_loss": 2.8865941365559897,
-    "train_runtime": 626.0526,
+    "eval_samples_per_second": 1331.385,
+    "eval_steps_per_second": 10.672,
+    "perplexity": 11.883751451666422,
+    "train_loss": 2.5668444753344595,
+    "train_runtime": 741.451,
     "train_samples": 46845,
-    "train_samples_per_second": 748.26,
-    "train_steps_per_second": 11.692
+    "train_samples_per_second": 631.802,
+    "train_steps_per_second": 39.49
 }

eval_results.json

@@ -1,10 +1,10 @@
 {
     "epoch": 10.0,
-    "eval_accuracy": 0.4868845302152862,
-    "eval_loss": 2.5603630542755127,
-    "eval_runtime": 4.1602,
+    "eval_accuracy": 0.4995467750899875,
+    "eval_loss": 2.4751720428466797,
+    "eval_runtime": 3.3732,
     "eval_samples": 4491,
-    "eval_samples_per_second": 1079.517,
-    "eval_steps_per_second": 2.163,
-    "perplexity": 12.940514571954653
+    "eval_samples_per_second": 1331.385,
+    "eval_steps_per_second": 10.672,
+    "perplexity": 11.883751451666422
 }

runs/Mar25_20-39-20_phyl-ling-p01.la.utexas.edu/events.out.tfevents.1711417940.phyl-ling-p01.la.utexas.edu.1052752.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0897e0f08c85774b756c2326e2ede8a47838e86c02d8a7efa335e8d4a2ae6571
+size 417

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 10.0,
-    "train_loss": 2.8865941365559897,
-    "train_runtime": 626.0526,
+    "train_loss": 2.5668444753344595,
+    "train_runtime": 741.451,
     "train_samples": 46845,
-    "train_samples_per_second": 748.26,
-    "train_steps_per_second": 11.692
+    "train_samples_per_second": 631.802,
+    "train_steps_per_second": 39.49
 }

trainer_state.json

@@ -3,200 +3,524 @@
   "best_model_checkpoint": null,
   "epoch": 10.0,
   "eval_steps": 500,
-  "global_step": 7320,
+  "global_step": 29280,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.68,
+      "epoch": 0.17,
+      "grad_norm": 1.3195292949676514,
       "learning_rate": 6.25e-05,
-      "loss": 6.1051,
+      "loss": 6.1984,
       "step": 500
     },
+    {
+      "epoch": 0.34,
+      "grad_norm": 0.8932374119758606,
+      "learning_rate": 0.000125,
+      "loss": 3.6328,
+      "step": 1000
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 1.1820354461669922,
+      "learning_rate": 0.0001875,
+      "loss": 3.332,
+      "step": 1500
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 1.031574010848999,
+      "learning_rate": 0.00025,
+      "loss": 3.1653,
+      "step": 2000
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 0.942737340927124,
+      "learning_rate": 0.0003125,
+      "loss": 3.0639,
+      "step": 2500
+    },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.39493334876226643,
-      "eval_loss": 3.5367648601531982,
-      "eval_runtime": 3.7763,
-      "eval_samples_per_second": 1189.255,
-      "eval_steps_per_second": 2.383,
-      "step": 732
+      "eval_accuracy": 0.4379818254181153,
+      "eval_loss": 3.0180583000183105,
+      "eval_runtime": 3.3254,
+      "eval_samples_per_second": 1350.527,
+      "eval_steps_per_second": 10.826,
+      "step": 2928
+    },
+    {
+      "epoch": 1.02,
+      "grad_norm": 0.9794430732727051,
+      "learning_rate": 0.000375,
+      "loss": 2.941,
+      "step": 3000
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 0.9245285987854004,
+      "learning_rate": 0.00043750000000000006,
+      "loss": 2.8816,
+      "step": 3500
     },
     {
       "epoch": 1.37,
-      "learning_rate": 0.000125,
-      "loss": 3.4882,
-      "step": 1000
+      "grad_norm": 0.8375250697135925,
+      "learning_rate": 0.0005,
+      "loss": 2.8172,
+      "step": 4000
+    },
+    {
+      "epoch": 1.54,
+      "grad_norm": 0.8608168959617615,
+      "learning_rate": 0.0005625000000000001,
+      "loss": 2.7785,
+      "step": 4500
+    },
+    {
+      "epoch": 1.71,
+      "grad_norm": 0.875029981136322,
+      "learning_rate": 0.000625,
+      "loss": 2.7348,
+      "step": 5000
+    },
+    {
+      "epoch": 1.88,
+      "grad_norm": 0.8714506030082703,
+      "learning_rate": 0.0006875,
+      "loss": 2.7091,
+      "step": 5500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.4336529577089437,
-      "eval_loss": 3.115656852722168,
-      "eval_runtime": 3.9524,
-      "eval_samples_per_second": 1136.258,
-      "eval_steps_per_second": 2.277,
-      "step": 1464
+      "eval_accuracy": 0.45751871196461164,
+      "eval_loss": 2.788527011871338,
+      "eval_runtime": 3.3539,
+      "eval_samples_per_second": 1339.048,
+      "eval_steps_per_second": 10.734,
+      "step": 5856
     },
     {
       "epoch": 2.05,
-      "learning_rate": 0.0001875,
-      "loss": 3.1377,
-      "step": 1500
+      "grad_norm": 0.8536949753761292,
+      "learning_rate": 0.00075,
+      "loss": 2.6749,
+      "step": 6000
+    },
+    {
+      "epoch": 2.22,
+      "grad_norm": 0.7690100073814392,
+      "learning_rate": 0.0008125,
+      "loss": 2.622,
+      "step": 6500
+    },
+    {
+      "epoch": 2.39,
+      "grad_norm": 0.7544898986816406,
+      "learning_rate": 0.0008750000000000001,
+      "loss": 2.6122,
+      "step": 7000
+    },
+    {
+      "epoch": 2.56,
+      "grad_norm": 0.7384330034255981,
+      "learning_rate": 0.0009375,
+      "loss": 2.6138,
+      "step": 7500
     },
     {
       "epoch": 2.73,
-      "learning_rate": 0.00025,
-      "loss": 2.9431,
-      "step": 2000
+      "grad_norm": 0.760734498500824,
+      "learning_rate": 0.001,
+      "loss": 2.5969,
+      "step": 8000
+    },
+    {
+      "epoch": 2.9,
+      "grad_norm": 0.6459075808525085,
+      "learning_rate": 0.0010625,
+      "loss": 2.5802,
+      "step": 8500
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.45023730751090985,
-      "eval_loss": 2.916651725769043,
-      "eval_runtime": 3.9895,
-      "eval_samples_per_second": 1125.698,
-      "eval_steps_per_second": 2.256,
-      "step": 2196
+      "eval_accuracy": 0.471204876945492,
+      "eval_loss": 2.6925525665283203,
+      "eval_runtime": 3.4091,
+      "eval_samples_per_second": 1317.365,
+      "eval_steps_per_second": 10.56,
+      "step": 8784
+    },
+    {
+      "epoch": 3.07,
+      "grad_norm": 0.6531331539154053,
+      "learning_rate": 0.0011250000000000001,
+      "loss": 2.552,
+      "step": 9000
+    },
+    {
+      "epoch": 3.24,
+      "grad_norm": 0.5972659587860107,
+      "learning_rate": 0.0011875,
+      "loss": 2.5069,
+      "step": 9500
     },
     {
       "epoch": 3.42,
-      "learning_rate": 0.0003125,
-      "loss": 2.7991,
-      "step": 2500
+      "grad_norm": 0.5470172762870789,
+      "learning_rate": 0.00125,
+      "loss": 2.5182,
+      "step": 10000
+    },
+    {
+      "epoch": 3.59,
+      "grad_norm": 0.5466737747192383,
+      "learning_rate": 0.0013125,
+      "loss": 2.5224,
+      "step": 10500
+    },
+    {
+      "epoch": 3.76,
+      "grad_norm": 0.511480987071991,
+      "learning_rate": 0.001375,
+      "loss": 2.515,
+      "step": 11000
+    },
+    {
+      "epoch": 3.93,
+      "grad_norm": 0.5196664929389954,
+      "learning_rate": 0.0014375000000000002,
+      "loss": 2.5132,
+      "step": 11500
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.461877385567285,
-      "eval_loss": 2.7866480350494385,
-      "eval_runtime": 4.1737,
-      "eval_samples_per_second": 1076.013,
-      "eval_steps_per_second": 2.156,
-      "step": 2928
+      "eval_accuracy": 0.4760264185413697,
+      "eval_loss": 2.6402881145477295,
+      "eval_runtime": 3.418,
+      "eval_samples_per_second": 1313.932,
+      "eval_steps_per_second": 10.533,
+      "step": 11712
     },
     {
       "epoch": 4.1,
-      "learning_rate": 0.000375,
-      "loss": 2.706,
-      "step": 3000
+      "grad_norm": 0.4554392695426941,
+      "learning_rate": 0.0015,
+      "loss": 2.4592,
+      "step": 12000
+    },
+    {
+      "epoch": 4.27,
+      "grad_norm": 0.4852846562862396,
+      "learning_rate": 0.0015625,
+      "loss": 2.4524,
+      "step": 12500
+    },
+    {
+      "epoch": 4.44,
+      "grad_norm": 0.445295125246048,
+      "learning_rate": 0.001625,
+      "loss": 2.4504,
+      "step": 13000
+    },
+    {
+      "epoch": 4.61,
+      "grad_norm": 0.47399041056632996,
+      "learning_rate": 0.0016875,
+      "loss": 2.4624,
+      "step": 13500
     },
     {
       "epoch": 4.78,
-      "learning_rate": 0.00043750000000000006,
-      "loss": 2.6152,
-      "step": 3500
+      "grad_norm": 0.3997687101364136,
+      "learning_rate": 0.0017500000000000003,
+      "loss": 2.4596,
+      "step": 14000
+    },
+    {
+      "epoch": 4.95,
+      "grad_norm": 0.3656143248081207,
+      "learning_rate": 0.0018124999999999999,
+      "loss": 2.4661,
+      "step": 14500
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.47078584114861394,
-      "eval_loss": 2.697930097579956,
-      "eval_runtime": 4.1031,
-      "eval_samples_per_second": 1094.526,
-      "eval_steps_per_second": 2.193,
-      "step": 3660
+      "eval_accuracy": 0.4783635512494806,
+      "eval_loss": 2.6108150482177734,
+      "eval_runtime": 3.3988,
+      "eval_samples_per_second": 1321.358,
+      "eval_steps_per_second": 10.592,
+      "step": 14640
+    },
+    {
+      "epoch": 5.12,
+      "grad_norm": 0.4261285364627838,
+      "learning_rate": 0.001875,
+      "loss": 2.4039,
+      "step": 15000
+    },
+    {
+      "epoch": 5.29,
+      "grad_norm": 0.3942112624645233,
+      "learning_rate": 0.0019375000000000002,
+      "loss": 2.4014,
+      "step": 15500
     },
     {
       "epoch": 5.46,
-      "learning_rate": 0.0005,
-      "loss": 2.5487,
-      "step": 4000
+      "grad_norm": 0.4000963270664215,
+      "learning_rate": 0.002,
+      "loss": 2.4071,
+      "step": 16000
+    },
+    {
+      "epoch": 5.64,
+      "grad_norm": 0.3776912987232208,
+      "learning_rate": 0.0020625,
+      "loss": 2.4142,
+      "step": 16500
+    },
+    {
+      "epoch": 5.81,
+      "grad_norm": 0.3140333294868469,
+      "learning_rate": 0.002125,
+      "loss": 2.4253,
+      "step": 17000
+    },
+    {
+      "epoch": 5.98,
+      "grad_norm": 0.3446044325828552,
+      "learning_rate": 0.0021874999999999998,
+      "loss": 2.4316,
+      "step": 17500
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.4759054416795095,
-      "eval_loss": 2.6426050662994385,
-      "eval_runtime": 4.0254,
-      "eval_samples_per_second": 1115.658,
-      "eval_steps_per_second": 2.236,
-      "step": 4392
+      "eval_accuracy": 0.48305885611993893,
+      "eval_loss": 2.5828359127044678,
+      "eval_runtime": 3.4134,
+      "eval_samples_per_second": 1315.703,
+      "eval_steps_per_second": 10.547,
+      "step": 17568
     },
     {
       "epoch": 6.15,
-      "learning_rate": 0.0005625000000000001,
-      "loss": 2.4959,
-      "step": 4500
+      "grad_norm": 0.327090322971344,
+      "learning_rate": 0.0022500000000000003,
+      "loss": 2.3465,
+      "step": 18000
+    },
+    {
+      "epoch": 6.32,
+      "grad_norm": 0.3468206822872162,
+      "learning_rate": 0.0023125000000000003,
+      "loss": 2.3651,
+      "step": 18500
+    },
+    {
+      "epoch": 6.49,
+      "grad_norm": 0.3663696050643921,
+      "learning_rate": 0.002375,
+      "loss": 2.3794,
+      "step": 19000
+    },
+    {
+      "epoch": 6.66,
+      "grad_norm": 0.2861100137233734,
+      "learning_rate": 0.0024375,
+      "loss": 2.3947,
+      "step": 19500
     },
     {
       "epoch": 6.83,
-      "learning_rate": 0.000625,
-      "loss": 2.4491,
-      "step": 5000
+      "grad_norm": 0.3619392514228821,
+      "learning_rate": 0.0025,
+      "loss": 2.3891,
+      "step": 20000
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.48190869928834046,
-      "eval_loss": 2.6028361320495605,
-      "eval_runtime": 3.9975,
-      "eval_samples_per_second": 1123.445,
-      "eval_steps_per_second": 2.251,
-      "step": 5124
+      "eval_accuracy": 0.48500149906111434,
+      "eval_loss": 2.5699071884155273,
+      "eval_runtime": 3.4157,
+      "eval_samples_per_second": 1314.807,
+      "eval_steps_per_second": 10.54,
+      "step": 20496
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 0.3399389386177063,
+      "learning_rate": 0.0025625,
+      "loss": 2.387,
+      "step": 20500
+    },
+    {
+      "epoch": 7.17,
+      "grad_norm": 0.29669496417045593,
+      "learning_rate": 0.002625,
+      "loss": 2.3106,
+      "step": 21000
+    },
+    {
+      "epoch": 7.34,
+      "grad_norm": 0.3277042508125305,
+      "learning_rate": 0.0026875000000000002,
+      "loss": 2.3308,
+      "step": 21500
     },
     {
       "epoch": 7.51,
-      "learning_rate": 0.0006875,
-      "loss": 2.4034,
-      "step": 5500
+      "grad_norm": 0.2957574427127838,
+      "learning_rate": 0.00275,
+      "loss": 2.3548,
+      "step": 22000
+    },
+    {
+      "epoch": 7.68,
+      "grad_norm": 0.29501375555992126,
+      "learning_rate": 0.0028125,
+      "loss": 2.3511,
+      "step": 22500
+    },
+    {
+      "epoch": 7.86,
+      "grad_norm": 0.29575708508491516,
+      "learning_rate": 0.0028750000000000004,
+      "loss": 2.3599,
+      "step": 23000
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.48347263205325786,
-      "eval_loss": 2.5807623863220215,
-      "eval_runtime": 4.0838,
-      "eval_samples_per_second": 1099.699,
-      "eval_steps_per_second": 2.204,
-      "step": 5856
+      "eval_accuracy": 0.48726324039154423,
+      "eval_loss": 2.5551772117614746,
+      "eval_runtime": 3.3869,
+      "eval_samples_per_second": 1325.979,
+      "eval_steps_per_second": 10.629,
+      "step": 23424
+    },
+    {
+      "epoch": 8.03,
+      "grad_norm": 0.2991016209125519,
+      "learning_rate": 0.0029375,
+      "loss": 2.3489,
+      "step": 23500
     },
     {
       "epoch": 8.2,
-      "learning_rate": 0.00075,
-      "loss": 2.3808,
-      "step": 6000
+      "grad_norm": 0.332133412361145,
+      "learning_rate": 0.003,
+      "loss": 2.2938,
+      "step": 24000
+    },
+    {
+      "epoch": 8.37,
+      "grad_norm": 0.3354604244232178,
+      "learning_rate": 0.002715909090909091,
+      "loss": 2.3065,
+      "step": 24500
+    },
+    {
+      "epoch": 8.54,
+      "grad_norm": 0.32910189032554626,
+      "learning_rate": 0.0024318181818181817,
+      "loss": 2.2954,
+      "step": 25000
+    },
+    {
+      "epoch": 8.71,
+      "grad_norm": 0.3498801589012146,
+      "learning_rate": 0.002147727272727273,
+      "loss": 2.2956,
+      "step": 25500
     },
     {
       "epoch": 8.88,
-      "learning_rate": 0.0008125,
-      "loss": 2.364,
-      "step": 6500
+      "grad_norm": 0.3250719904899597,
+      "learning_rate": 0.0018636363636363638,
+      "loss": 2.2821,
+      "step": 26000
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.48510669633229714,
-      "eval_loss": 2.565694570541382,
-      "eval_runtime": 4.2034,
-      "eval_samples_per_second": 1068.427,
-      "eval_steps_per_second": 2.141,
-      "step": 6588
+      "eval_accuracy": 0.494129115624074,
+      "eval_loss": 2.501997470855713,
+      "eval_runtime": 3.3639,
+      "eval_samples_per_second": 1335.064,
+      "eval_steps_per_second": 10.702,
+      "step": 26352
+    },
+    {
+      "epoch": 9.05,
+      "grad_norm": 0.3215320408344269,
+      "learning_rate": 0.0015795454545454546,
+      "loss": 2.244,
+      "step": 26500
+    },
+    {
+      "epoch": 9.22,
+      "grad_norm": 0.275278240442276,
+      "learning_rate": 0.0012954545454545456,
+      "loss": 2.1635,
+      "step": 27000
+    },
+    {
+      "epoch": 9.39,
+      "grad_norm": 0.33266499638557434,
+      "learning_rate": 0.0010113636363636364,
+      "loss": 2.1561,
+      "step": 27500
     },
     {
       "epoch": 9.56,
-      "learning_rate": 0.0008750000000000001,
-      "loss": 2.3269,
-      "step": 7000
+      "grad_norm": 0.3034181594848633,
+      "learning_rate": 0.0007272727272727273,
+      "loss": 2.1569,
+      "step": 28000
+    },
+    {
+      "epoch": 9.73,
+      "grad_norm": 0.32139885425567627,
+      "learning_rate": 0.0004431818181818182,
+      "loss": 2.1493,
+      "step": 28500
+    },
+    {
+      "epoch": 9.9,
+      "grad_norm": 0.2935737073421478,
+      "learning_rate": 0.0001590909090909091,
+      "loss": 2.1425,
+      "step": 29000
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.4868845302152862,
-      "eval_loss": 2.5603630542755127,
-      "eval_runtime": 4.0953,
-      "eval_samples_per_second": 1096.619,
-      "eval_steps_per_second": 2.198,
-      "step": 7320
+      "eval_accuracy": 0.4995467750899875,
+      "eval_loss": 2.4751720428466797,
+      "eval_runtime": 3.364,
+      "eval_samples_per_second": 1335.021,
+      "eval_steps_per_second": 10.702,
+      "step": 29280
     },
     {
       "epoch": 10.0,
-      "step": 7320,
+      "step": 29280,
       "total_flos": 2273237316403200.0,
-      "train_loss": 2.8865941365559897,
-      "train_runtime": 626.0526,
-      "train_samples_per_second": 748.26,
-      "train_steps_per_second": 11.692
+      "train_loss": 2.5668444753344595,
+      "train_runtime": 741.451,
+      "train_samples_per_second": 631.802,
+      "train_steps_per_second": 39.49
     }
   ],
   "logging_steps": 500,
-  "max_steps": 7320,
+  "max_steps": 29280,
+  "num_input_tokens_seen": 0,
   "num_train_epochs": 10,
   "save_steps": 2000,
   "total_flos": 2273237316403200.0,
+  "train_batch_size": 16,
   "trial_name": null,
   "trial_params": null
 }