End of training

all_results.json

@@ -1,15 +1,15 @@
 {
     "epoch": 10.0,
-    "eval_accuracy": 0.4865023134633221,
-    "eval_loss": 2.557213068008423,
-    "eval_runtime": 3.9929,
+    "eval_accuracy": 0.5002726362611487,
+    "eval_loss": 2.4750711917877197,
+    "eval_runtime": 3.3673,
     "eval_samples": 4491,
-    "eval_samples_per_second": 1124.739,
-    "eval_steps_per_second": 2.254,
-    "perplexity": 12.8998162619748,
-    "train_loss": 2.8840711333061178,
-    "train_runtime": 613.5857,
+    "eval_samples_per_second": 1333.713,
+    "eval_steps_per_second": 10.691,
+    "perplexity": 11.882553023180511,
+    "train_loss": 2.566523836479812,
+    "train_runtime": 773.4482,
     "train_samples": 46845,
-    "train_samples_per_second": 763.463,
-    "train_steps_per_second": 11.93
+    "train_samples_per_second": 605.664,
+    "train_steps_per_second": 37.856
 }

eval_results.json

@@ -1,10 +1,10 @@
 {
     "epoch": 10.0,
-    "eval_accuracy": 0.4865023134633221,
-    "eval_loss": 2.557213068008423,
-    "eval_runtime": 3.9929,
+    "eval_accuracy": 0.5002726362611487,
+    "eval_loss": 2.4750711917877197,
+    "eval_runtime": 3.3673,
     "eval_samples": 4491,
-    "eval_samples_per_second": 1124.739,
-    "eval_steps_per_second": 2.254,
-    "perplexity": 12.8998162619748
+    "eval_samples_per_second": 1333.713,
+    "eval_steps_per_second": 10.691,
+    "perplexity": 11.882553023180511
 }

runs/Mar25_20-52-31_phyl-ling-p01.la.utexas.edu/events.out.tfevents.1711418749.phyl-ling-p01.la.utexas.edu.1053486.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73dd140c641bce6ab815923a38a518c39621912246183e774b2853f313ce659b
+size 417

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 10.0,
-    "train_loss": 2.8840711333061178,
-    "train_runtime": 613.5857,
+    "train_loss": 2.566523836479812,
+    "train_runtime": 773.4482,
     "train_samples": 46845,
-    "train_samples_per_second": 763.463,
-    "train_steps_per_second": 11.93
+    "train_samples_per_second": 605.664,
+    "train_steps_per_second": 37.856
 }

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": 1.1626055240631104,
       "learning_rate": 6.25e-05,
-      "loss": 6.0722,
+      "loss": 6.1545,
       "step": 500
     },
+    {
+      "epoch": 0.34,
+      "grad_norm": 1.0916112661361694,
+      "learning_rate": 0.000125,
+      "loss": 3.6268,
+      "step": 1000
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 1.0328254699707031,
+      "learning_rate": 0.0001875,
+      "loss": 3.3534,
+      "step": 1500
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 0.9192247986793518,
+      "learning_rate": 0.00025,
+      "loss": 3.1803,
+      "step": 2000
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 1.00106680393219,
+      "learning_rate": 0.0003125,
+      "loss": 3.0578,
+      "step": 2500
+    },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.3931940872120444,
-      "eval_loss": 3.5448415279388428,
-      "eval_runtime": 3.7398,
-      "eval_samples_per_second": 1200.87,
-      "eval_steps_per_second": 2.407,
-      "step": 732
+      "eval_accuracy": 0.43743304632011176,
+      "eval_loss": 3.019747495651245,
+      "eval_runtime": 3.374,
+      "eval_samples_per_second": 1331.065,
+      "eval_steps_per_second": 10.67,
+      "step": 2928
+    },
+    {
+      "epoch": 1.02,
+      "grad_norm": 0.9110632538795471,
+      "learning_rate": 0.000375,
+      "loss": 2.9579,
+      "step": 3000
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 0.87364661693573,
+      "learning_rate": 0.00043750000000000006,
+      "loss": 2.8674,
+      "step": 3500
     },
     {
       "epoch": 1.37,
-      "learning_rate": 0.000125,
-      "loss": 3.4869,
-      "step": 1000
+      "grad_norm": 0.9194164276123047,
+      "learning_rate": 0.0005,
+      "loss": 2.8216,
+      "step": 4000
+    },
+    {
+      "epoch": 1.54,
+      "grad_norm": 0.9358027577400208,
+      "learning_rate": 0.0005625000000000001,
+      "loss": 2.776,
+      "step": 4500
+    },
+    {
+      "epoch": 1.71,
+      "grad_norm": 0.8049463033676147,
+      "learning_rate": 0.000625,
+      "loss": 2.7402,
+      "step": 5000
+    },
+    {
+      "epoch": 1.88,
+      "grad_norm": 0.7903618812561035,
+      "learning_rate": 0.0006875,
+      "loss": 2.7165,
+      "step": 5500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.43195226849148866,
-      "eval_loss": 3.122854709625244,
-      "eval_runtime": 3.7445,
-      "eval_samples_per_second": 1199.365,
-      "eval_steps_per_second": 2.404,
-      "step": 1464
+      "eval_accuracy": 0.459875130839106,
+      "eval_loss": 2.780012845993042,
+      "eval_runtime": 3.4904,
+      "eval_samples_per_second": 1286.685,
+      "eval_steps_per_second": 10.314,
+      "step": 5856
     },
     {
       "epoch": 2.05,
-      "learning_rate": 0.0001875,
-      "loss": 3.1466,
-      "step": 1500
+      "grad_norm": 0.7955217361450195,
+      "learning_rate": 0.00075,
+      "loss": 2.6667,
+      "step": 6000
+    },
+    {
+      "epoch": 2.22,
+      "grad_norm": 0.7588896751403809,
+      "learning_rate": 0.0008125,
+      "loss": 2.621,
+      "step": 6500
+    },
+    {
+      "epoch": 2.39,
+      "grad_norm": 0.7160364985466003,
+      "learning_rate": 0.0008750000000000001,
+      "loss": 2.613,
+      "step": 7000
+    },
+    {
+      "epoch": 2.56,
+      "grad_norm": 0.7120616436004639,
+      "learning_rate": 0.0009375,
+      "loss": 2.6087,
+      "step": 7500
     },
     {
       "epoch": 2.73,
-      "learning_rate": 0.00025,
-      "loss": 2.9405,
-      "step": 2000
+      "grad_norm": 0.6713188290596008,
+      "learning_rate": 0.001,
+      "loss": 2.5961,
+      "step": 8000
+    },
+    {
+      "epoch": 2.9,
+      "grad_norm": 0.6472102999687195,
+      "learning_rate": 0.0010625,
+      "loss": 2.5811,
+      "step": 8500
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.449262479464616,
-      "eval_loss": 2.915255069732666,
-      "eval_runtime": 3.8042,
-      "eval_samples_per_second": 1180.526,
-      "eval_steps_per_second": 2.366,
-      "step": 2196
+      "eval_accuracy": 0.4701529042336642,
+      "eval_loss": 2.6874144077301025,
+      "eval_runtime": 3.4827,
+      "eval_samples_per_second": 1289.523,
+      "eval_steps_per_second": 10.337,
+      "step": 8784
+    },
+    {
+      "epoch": 3.07,
+      "grad_norm": 0.6498757004737854,
+      "learning_rate": 0.0011250000000000001,
+      "loss": 2.5409,
+      "step": 9000
+    },
+    {
+      "epoch": 3.24,
+      "grad_norm": 0.6110037565231323,
+      "learning_rate": 0.0011875,
+      "loss": 2.5227,
+      "step": 9500
     },
     {
       "epoch": 3.42,
-      "learning_rate": 0.0003125,
-      "loss": 2.8073,
-      "step": 2500
+      "grad_norm": 0.6215284466743469,
+      "learning_rate": 0.00125,
+      "loss": 2.5237,
+      "step": 10000
+    },
+    {
+      "epoch": 3.59,
+      "grad_norm": 0.5484191179275513,
+      "learning_rate": 0.0013125,
+      "loss": 2.5089,
+      "step": 10500
+    },
+    {
+      "epoch": 3.76,
+      "grad_norm": 0.5084063410758972,
+      "learning_rate": 0.001375,
+      "loss": 2.5077,
+      "step": 11000
+    },
+    {
+      "epoch": 3.93,
+      "grad_norm": 0.4629107713699341,
+      "learning_rate": 0.0014375000000000002,
+      "loss": 2.5062,
+      "step": 11500
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.46225083587998395,
-      "eval_loss": 2.783942937850952,
-      "eval_runtime": 3.8779,
-      "eval_samples_per_second": 1158.102,
-      "eval_steps_per_second": 2.321,
-      "step": 2928
+      "eval_accuracy": 0.47650331283739833,
+      "eval_loss": 2.6332497596740723,
+      "eval_runtime": 3.4691,
+      "eval_samples_per_second": 1294.569,
+      "eval_steps_per_second": 10.377,
+      "step": 11712
     },
     {
       "epoch": 4.1,
-      "learning_rate": 0.000375,
-      "loss": 2.7004,
-      "step": 3000
+      "grad_norm": 0.46648529171943665,
+      "learning_rate": 0.0015,
+      "loss": 2.4668,
+      "step": 12000
+    },
+    {
+      "epoch": 4.27,
+      "grad_norm": 0.46222758293151855,
+      "learning_rate": 0.0015625,
+      "loss": 2.448,
+      "step": 12500
+    },
+    {
+      "epoch": 4.44,
+      "grad_norm": 0.42898261547088623,
+      "learning_rate": 0.001625,
+      "loss": 2.4481,
+      "step": 13000
+    },
+    {
+      "epoch": 4.61,
+      "grad_norm": 0.4155634641647339,
+      "learning_rate": 0.0016875,
+      "loss": 2.4523,
+      "step": 13500
     },
     {
       "epoch": 4.78,
-      "learning_rate": 0.00043750000000000006,
-      "loss": 2.6116,
-      "step": 3500
+      "grad_norm": 0.4071389436721802,
+      "learning_rate": 0.0017500000000000003,
+      "loss": 2.4633,
+      "step": 14000
+    },
+    {
+      "epoch": 4.95,
+      "grad_norm": 0.4090569317340851,
+      "learning_rate": 0.0018124999999999999,
+      "loss": 2.469,
+      "step": 14500
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.47078408786076087,
-      "eval_loss": 2.6958837509155273,
-      "eval_runtime": 4.0364,
-      "eval_samples_per_second": 1112.617,
-      "eval_steps_per_second": 2.23,
-      "step": 3660
+      "eval_accuracy": 0.480833933834423,
+      "eval_loss": 2.6065454483032227,
+      "eval_runtime": 3.4106,
+      "eval_samples_per_second": 1316.773,
+      "eval_steps_per_second": 10.555,
+      "step": 14640
+    },
+    {
+      "epoch": 5.12,
+      "grad_norm": 0.36704427003860474,
+      "learning_rate": 0.001875,
+      "loss": 2.3982,
+      "step": 15000
+    },
+    {
+      "epoch": 5.29,
+      "grad_norm": 0.40789690613746643,
+      "learning_rate": 0.0019375000000000002,
+      "loss": 2.3963,
+      "step": 15500
     },
     {
       "epoch": 5.46,
-      "learning_rate": 0.0005,
-      "loss": 2.5433,
-      "step": 4000
+      "grad_norm": 0.3699958622455597,
+      "learning_rate": 0.002,
+      "loss": 2.4208,
+      "step": 16000
+    },
+    {
+      "epoch": 5.64,
+      "grad_norm": 0.3502542972564697,
+      "learning_rate": 0.0020625,
+      "loss": 2.4209,
+      "step": 16500
+    },
+    {
+      "epoch": 5.81,
+      "grad_norm": 0.3579181134700775,
+      "learning_rate": 0.002125,
+      "loss": 2.42,
+      "step": 17000
+    },
+    {
+      "epoch": 5.98,
+      "grad_norm": 0.3378722071647644,
+      "learning_rate": 0.0021874999999999998,
+      "loss": 2.4281,
+      "step": 17500
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.47723618715997174,
-      "eval_loss": 2.637035608291626,
-      "eval_runtime": 4.1265,
-      "eval_samples_per_second": 1088.339,
-      "eval_steps_per_second": 2.181,
-      "step": 4392
+      "eval_accuracy": 0.484081022938265,
+      "eval_loss": 2.5812158584594727,
+      "eval_runtime": 3.4105,
+      "eval_samples_per_second": 1316.809,
+      "eval_steps_per_second": 10.556,
+      "step": 17568
     },
     {
       "epoch": 6.15,
-      "learning_rate": 0.0005625000000000001,
-      "loss": 2.4916,
-      "step": 4500
+      "grad_norm": 0.35191810131073,
+      "learning_rate": 0.0022500000000000003,
+      "loss": 2.3517,
+      "step": 18000
+    },
+    {
+      "epoch": 6.32,
+      "grad_norm": 0.32431522011756897,
+      "learning_rate": 0.0023125000000000003,
+      "loss": 2.3664,
+      "step": 18500
+    },
+    {
+      "epoch": 6.49,
+      "grad_norm": 0.33494833111763,
+      "learning_rate": 0.002375,
+      "loss": 2.3751,
+      "step": 19000
+    },
+    {
+      "epoch": 6.66,
+      "grad_norm": 0.2809942066669464,
+      "learning_rate": 0.0024375,
+      "loss": 2.3887,
+      "step": 19500
     },
     {
       "epoch": 6.83,
-      "learning_rate": 0.000625,
-      "loss": 2.4481,
-      "step": 5000
+      "grad_norm": 0.36362403631210327,
+      "learning_rate": 0.0025,
+      "loss": 2.3831,
+      "step": 20000
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.48569229447521467,
+      "eval_loss": 2.5669186115264893,
+      "eval_runtime": 3.4109,
+      "eval_samples_per_second": 1316.642,
+      "eval_steps_per_second": 10.554,
+      "step": 20496
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.48059723997426174,
-      "eval_loss": 2.6047284603118896,
-      "eval_runtime": 3.9004,
-      "eval_samples_per_second": 1151.432,
-      "eval_steps_per_second": 2.307,
-      "step": 5124
+      "grad_norm": 0.3269413709640503,
+      "learning_rate": 0.0025625,
+      "loss": 2.3969,
+      "step": 20500
+    },
+    {
+      "epoch": 7.17,
+      "grad_norm": 0.32689368724823,
+      "learning_rate": 0.002625,
+      "loss": 2.3181,
+      "step": 21000
+    },
+    {
+      "epoch": 7.34,
+      "grad_norm": 0.3296111524105072,
+      "learning_rate": 0.0026875000000000002,
+      "loss": 2.3339,
+      "step": 21500
     },
     {
       "epoch": 7.51,
-      "learning_rate": 0.0006875,
-      "loss": 2.4044,
-      "step": 5500
+      "grad_norm": 0.3480948805809021,
+      "learning_rate": 0.00275,
+      "loss": 2.3513,
+      "step": 22000
+    },
+    {
+      "epoch": 7.68,
+      "grad_norm": 0.3491817116737366,
+      "learning_rate": 0.0028125,
+      "loss": 2.3593,
+      "step": 22500
+    },
+    {
+      "epoch": 7.86,
+      "grad_norm": 0.3150429129600525,
+      "learning_rate": 0.0028750000000000004,
+      "loss": 2.3556,
+      "step": 23000
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.48257319538464505,
-      "eval_loss": 2.581453800201416,
-      "eval_runtime": 4.0742,
-      "eval_samples_per_second": 1102.31,
-      "eval_steps_per_second": 2.209,
-      "step": 5856
+      "eval_accuracy": 0.4877524077025442,
+      "eval_loss": 2.554654836654663,
+      "eval_runtime": 3.4547,
+      "eval_samples_per_second": 1299.986,
+      "eval_steps_per_second": 10.421,
+      "step": 23424
+    },
+    {
+      "epoch": 8.03,
+      "grad_norm": 0.3214956820011139,
+      "learning_rate": 0.0029375,
+      "loss": 2.3454,
+      "step": 23500
     },
     {
       "epoch": 8.2,
-      "learning_rate": 0.00075,
-      "loss": 2.3857,
-      "step": 6000
+      "grad_norm": 0.3276998698711395,
+      "learning_rate": 0.003,
+      "loss": 2.289,
+      "step": 24000
+    },
+    {
+      "epoch": 8.37,
+      "grad_norm": 0.34141024947166443,
+      "learning_rate": 0.002715909090909091,
+      "loss": 2.3054,
+      "step": 24500
+    },
+    {
+      "epoch": 8.54,
+      "grad_norm": 0.3355658948421478,
+      "learning_rate": 0.0024318181818181817,
+      "loss": 2.3034,
+      "step": 25000
+    },
+    {
+      "epoch": 8.71,
+      "grad_norm": 0.3076556324958801,
+      "learning_rate": 0.002147727272727273,
+      "loss": 2.293,
+      "step": 25500
     },
     {
       "epoch": 8.88,
-      "learning_rate": 0.0008125,
-      "loss": 2.3619,
-      "step": 6500
+      "grad_norm": 0.3368534743785858,
+      "learning_rate": 0.0018636363636363638,
+      "loss": 2.2852,
+      "step": 26000
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.4853626763588419,
-      "eval_loss": 2.569573402404785,
-      "eval_runtime": 3.911,
-      "eval_samples_per_second": 1148.287,
-      "eval_steps_per_second": 2.301,
-      "step": 6588
+      "eval_accuracy": 0.4947883518568195,
+      "eval_loss": 2.5027360916137695,
+      "eval_runtime": 3.4682,
+      "eval_samples_per_second": 1294.896,
+      "eval_steps_per_second": 10.38,
+      "step": 26352
+    },
+    {
+      "epoch": 9.05,
+      "grad_norm": 0.31588613986968994,
+      "learning_rate": 0.0015795454545454546,
+      "loss": 2.2399,
+      "step": 26500
+    },
+    {
+      "epoch": 9.22,
+      "grad_norm": 0.3448103070259094,
+      "learning_rate": 0.0012954545454545456,
+      "loss": 2.1651,
+      "step": 27000
+    },
+    {
+      "epoch": 9.39,
+      "grad_norm": 0.346188485622406,
+      "learning_rate": 0.0010113636363636364,
+      "loss": 2.1708,
+      "step": 27500
     },
     {
       "epoch": 9.56,
-      "learning_rate": 0.0008750000000000001,
-      "loss": 2.322,
-      "step": 7000
+      "grad_norm": 0.3420686721801758,
+      "learning_rate": 0.0007272727272727273,
+      "loss": 2.1565,
+      "step": 28000
+    },
+    {
+      "epoch": 9.73,
+      "grad_norm": 0.2834959626197815,
+      "learning_rate": 0.0004431818181818182,
+      "loss": 2.1478,
+      "step": 28500
+    },
+    {
+      "epoch": 9.9,
+      "grad_norm": 0.30597221851348877,
+      "learning_rate": 0.0001590909090909091,
+      "loss": 2.1378,
+      "step": 29000
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.4865023134633221,
-      "eval_loss": 2.557213068008423,
-      "eval_runtime": 3.7775,
-      "eval_samples_per_second": 1188.876,
-      "eval_steps_per_second": 2.383,
-      "step": 7320
+      "eval_accuracy": 0.5002726362611487,
+      "eval_loss": 2.4750711917877197,
+      "eval_runtime": 3.4442,
+      "eval_samples_per_second": 1303.949,
+      "eval_steps_per_second": 10.452,
+      "step": 29280
     },
     {
       "epoch": 10.0,
-      "step": 7320,
+      "step": 29280,
       "total_flos": 2273237316403200.0,
-      "train_loss": 2.8840711333061178,
-      "train_runtime": 613.5857,
-      "train_samples_per_second": 763.463,
-      "train_steps_per_second": 11.93
+      "train_loss": 2.566523836479812,
+      "train_runtime": 773.4482,
+      "train_samples_per_second": 605.664,
+      "train_steps_per_second": 37.856
     }
   ],
   "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
 }