Training complete

all_results.json

@@ -1,12 +1,12 @@
 {
     "epoch": 3.0,
-    "eval_loss": 0.7299028038978577,
-    "eval_runtime": 101.4278,
-    "eval_samples_per_second": 1.094,
-    "eval_steps_per_second": 0.069,
-    "total_flos": 3.744392573136077e+16,
-    "train_loss": 0.5652910582129917,
-    "train_runtime": 2830.2027,
-    "train_samples_per_second": 0.353,
-    "train_steps_per_second": 0.089
+    "eval_loss": 0.6883889436721802,
+    "eval_runtime": 103.9672,
+    "eval_samples_per_second": 1.068,
+    "eval_steps_per_second": 0.135,
+    "total_flos": 4.081496890291814e+16,
+    "train_loss": 0.6921328268353901,
+    "train_runtime": 3110.5155,
+    "train_samples_per_second": 0.321,
+    "train_steps_per_second": 0.02
 }

eval_results.json

@@ -1,7 +1,7 @@
 {
     "epoch": 3.0,
-    "eval_loss": 0.7299028038978577,
-    "eval_runtime": 101.4278,
-    "eval_samples_per_second": 1.094,
-    "eval_steps_per_second": 0.069
+    "eval_loss": 0.6883889436721802,
+    "eval_runtime": 103.9672,
+    "eval_samples_per_second": 1.068,
+    "eval_steps_per_second": 0.135
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 3.0,
-    "total_flos": 3.744392573136077e+16,
-    "train_loss": 0.5652910582129917,
-    "train_runtime": 2830.2027,
-    "train_samples_per_second": 0.353,
-    "train_steps_per_second": 0.089
+    "total_flos": 4.081496890291814e+16,
+    "train_loss": 0.6921328268353901,
+    "train_runtime": 3110.5155,
+    "train_samples_per_second": 0.321,
+    "train_steps_per_second": 0.02
 }

trainer_state.json

@@ -3,203 +3,94 @@
   "best_model_checkpoint": null,
   "epoch": 3.0,
   "eval_steps": 500,
-  "global_step": 252,
+  "global_step": 63,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
-    {
-      "epoch": 0.12,
-      "grad_norm": 350422.84375,
-      "learning_rate": 0.0002988358809900258,
-      "loss": 1.2276,
-      "step": 10
-    },
-    {
-      "epoch": 0.24,
-      "grad_norm": 127592.3203125,
-      "learning_rate": 0.00029536159293436166,
-      "loss": 0.7731,
-      "step": 20
-    },
-    {
-      "epoch": 0.36,
-      "grad_norm": 113769.28125,
-      "learning_rate": 0.00028963106229663063,
-      "loss": 0.7033,
-      "step": 30
-    },
     {
       "epoch": 0.48,
-      "grad_norm": 109609.40625,
-      "learning_rate": 0.0002817332360055343,
-      "loss": 0.7912,
-      "step": 40
-    },
-    {
-      "epoch": 0.6,
-      "grad_norm": 130476.6171875,
-      "learning_rate": 0.0002717907008573785,
-      "loss": 0.7413,
-      "step": 50
-    },
-    {
-      "epoch": 0.71,
-      "grad_norm": 86021.2265625,
-      "learning_rate": 0.0002599577807744739,
-      "loss": 0.7233,
-      "step": 60
-    },
-    {
-      "epoch": 0.83,
-      "grad_norm": 87099.53125,
-      "learning_rate": 0.0002464181414529809,
-      "loss": 0.759,
-      "step": 70
+      "grad_norm": 78838.859375,
+      "learning_rate": 0.00018782215733702286,
+      "loss": 1.2119,
+      "step": 10
     },
     {
       "epoch": 0.95,
-      "grad_norm": 79031.671875,
-      "learning_rate": 0.0002313819395798639,
-      "loss": 0.7152,
-      "step": 80
-    },
-    {
-      "epoch": 1.07,
-      "grad_norm": 96164.9609375,
-      "learning_rate": 0.00021508256086763368,
-      "loss": 0.6194,
-      "step": 90
-    },
-    {
-      "epoch": 1.19,
-      "grad_norm": 82587.0859375,
-      "learning_rate": 0.00019777299753775265,
-      "loss": 0.5308,
-      "step": 100
+      "grad_norm": 79446.625,
+      "learning_rate": 0.00015425462638657595,
+      "loss": 0.7581,
+      "step": 20
     },
     {
-      "epoch": 1.31,
-      "grad_norm": 83371.2578125,
-      "learning_rate": 0.0001797219214799096,
-      "loss": 0.5415,
-      "step": 110
+      "epoch": 1.0,
+      "eval_loss": 0.7133347392082214,
+      "eval_runtime": 104.4914,
+      "eval_samples_per_second": 1.062,
+      "eval_steps_per_second": 0.134,
+      "step": 21
     },
     {
       "epoch": 1.43,
-      "grad_norm": 107494.8984375,
-      "learning_rate": 0.00016120951403796364,
-      "loss": 0.5953,
-      "step": 120
-    },
-    {
-      "epoch": 1.55,
-      "grad_norm": 110181.2578125,
-      "learning_rate": 0.0001425231171508954,
-      "loss": 0.591,
-      "step": 130
-    },
-    {
-      "epoch": 1.67,
-      "grad_norm": 84472.59375,
-      "learning_rate": 0.00012395277334996044,
-      "loss": 0.5071,
-      "step": 140
-    },
-    {
-      "epoch": 1.79,
-      "grad_norm": 72922.546875,
-      "learning_rate": 0.00010578672383836435,
-      "loss": 0.5188,
-      "step": 150
+      "grad_norm": 60001.390625,
+      "learning_rate": 0.00010747300935864243,
+      "loss": 0.6176,
+      "step": 30
     },
     {
       "epoch": 1.9,
-      "grad_norm": 84903.71875,
-      "learning_rate": 8.830693453040829e-05,
-      "loss": 0.5508,
-      "step": 160
-    },
-    {
-      "epoch": 2.02,
-      "grad_norm": 71174.5546875,
-      "learning_rate": 7.17847194930753e-05,
-      "loss": 0.4979,
-      "step": 170
-    },
-    {
-      "epoch": 2.14,
-      "grad_norm": 190946.0625,
-      "learning_rate": 5.6476529721189974e-05,
-      "loss": 0.3451,
-      "step": 180
+      "grad_norm": 59219.640625,
+      "learning_rate": 5.887128968693887e-05,
+      "loss": 0.612,
+      "step": 40
     },
     {
-      "epoch": 2.26,
-      "grad_norm": 73154.4609375,
-      "learning_rate": 4.261997261104223e-05,
-      "loss": 0.3543,
-      "step": 190
+      "epoch": 2.0,
+      "eval_loss": 0.680975079536438,
+      "eval_runtime": 104.4154,
+      "eval_samples_per_second": 1.063,
+      "eval_steps_per_second": 0.134,
+      "step": 42
     },
     {
       "epoch": 2.38,
-      "grad_norm": 97452.4453125,
-      "learning_rate": 3.0430123916561672e-05,
-      "loss": 0.3384,
-      "step": 200
-    },
-    {
-      "epoch": 2.5,
-      "grad_norm": 75840.453125,
-      "learning_rate": 2.009618943233419e-05,
-      "loss": 0.3708,
-      "step": 210
-    },
-    {
-      "epoch": 2.62,
-      "grad_norm": 101251.2578125,
-      "learning_rate": 1.1778568219438839e-05,
-      "loss": 0.354,
-      "step": 220
-    },
-    {
-      "epoch": 2.74,
-      "grad_norm": 77960.1015625,
-      "learning_rate": 5.606362957498195e-06,
-      "loss": 0.3275,
-      "step": 230
+      "grad_norm": 56114.09765625,
+      "learning_rate": 2.0286749277707782e-05,
+      "loss": 0.5092,
+      "step": 50
     },
     {
       "epoch": 2.86,
-      "grad_norm": 92760.578125,
-      "learning_rate": 1.6753760662307215e-06,
-      "loss": 0.3273,
-      "step": 240
+      "grad_norm": 58096.5703125,
+      "learning_rate": 1.1169173774871478e-06,
+      "loss": 0.4944,
+      "step": 60
     },
     {
-      "epoch": 2.98,
-      "grad_norm": 78068.015625,
-      "learning_rate": 4.662269987756317e-08,
-      "loss": 0.3638,
-      "step": 250
+      "epoch": 3.0,
+      "eval_loss": 0.6883889436721802,
+      "eval_runtime": 104.4068,
+      "eval_samples_per_second": 1.063,
+      "eval_steps_per_second": 0.134,
+      "step": 63
     },
     {
       "epoch": 3.0,
-      "step": 252,
-      "total_flos": 3.744392573136077e+16,
-      "train_loss": 0.5652910582129917,
-      "train_runtime": 2830.2027,
-      "train_samples_per_second": 0.353,
-      "train_steps_per_second": 0.089
+      "step": 63,
+      "total_flos": 4.081496890291814e+16,
+      "train_loss": 0.6921328268353901,
+      "train_runtime": 3110.5155,
+      "train_samples_per_second": 0.321,
+      "train_steps_per_second": 0.02
     }
   ],
   "logging_steps": 10,
-  "max_steps": 252,
+  "max_steps": 63,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 3,
   "save_steps": 10,
-  "total_flos": 3.744392573136077e+16,
-  "train_batch_size": 4,
+  "total_flos": 4.081496890291814e+16,
+  "train_batch_size": 8,
   "trial_name": null,
   "trial_params": null
 }