Upload trainer_state.json with huggingface_hub

trainer_state.json

@@ -2,9 +2,9 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.22598870056497175,
+  "epoch": 0.4519774011299435,
   "eval_steps": 10,
-  "global_step": 20,
+  "global_step": 40,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -50,6 +50,48 @@
       "eval_samples_per_second": 4.38,
       "eval_steps_per_second": 1.095,
       "step": 20
+    },
+    {
+      "entropy": 7.635345196723938,
+      "epoch": 0.3389830508474576,
+      "grad_norm": 8.9375,
+      "learning_rate": 0.00017325581395348838,
+      "loss": 9.085212707519531,
+      "mean_token_accuracy": 0.11288385493680834,
+      "num_tokens": 505921.0,
+      "step": 30
+    },
+    {
+      "epoch": 0.3389830508474576,
+      "eval_entropy": 7.880531340837479,
+      "eval_loss": 7.876997470855713,
+      "eval_mean_token_accuracy": 0.09879284957423806,
+      "eval_num_tokens": 505921.0,
+      "eval_runtime": 43.9757,
+      "eval_samples_per_second": 4.366,
+      "eval_steps_per_second": 1.092,
+      "step": 30
+    },
+    {
+      "entropy": 7.543534195423126,
+      "epoch": 0.4519774011299435,
+      "grad_norm": 1.40625,
+      "learning_rate": 0.00016162790697674419,
+      "loss": 7.40704116821289,
+      "mean_token_accuracy": 0.11707657705992461,
+      "num_tokens": 677163.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.4519774011299435,
+      "eval_entropy": 7.573027561108272,
+      "eval_loss": 7.244246006011963,
+      "eval_mean_token_accuracy": 0.10892315845315655,
+      "eval_num_tokens": 677163.0,
+      "eval_runtime": 44.408,
+      "eval_samples_per_second": 4.324,
+      "eval_steps_per_second": 1.081,
+      "step": 40
     }
   ],
   "logging_steps": 10,
@@ -69,7 +111,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 2.397293116495872e+16,
+  "total_flos": 4.854116392922726e+16,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null