Training in progress, step 24, checkpoint

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3b1620b372b059d1acf1d8d8f8e7ed129fd3d08d3555671a805512a5f946e36c
+oid sha256:0dfea9e91529638011e2b2348c421b593487e3c8511b501b2034e4bfcd2358ee
 size 335922386

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6f70c56bbf1df2389440cc9d92d46fa3ba9afcb52998f573ea87fd1fd3118270
+oid sha256:a3e1982295476e84d089e86abce749a0a269a426f32611cee726d775e4c4d567
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:23b9d3742ef16544005722539fe9914bac40273bcf4801b3134df389839d2871
+oid sha256:9be43866b7a112efbf8125d7bbc11610819a4fb8c7f205bdfffb33dc32734ab8
 size 1064

last-checkpoint/trainer_state.json

@@ -1,9 +1,9 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.0006887860927826176,
+  "epoch": 0.0007514030103083101,
   "eval_steps": 2,
-  "global_step": 22,
+  "global_step": 24,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -257,6 +257,28 @@
       "eval_samples_per_second": 4.056,
       "eval_steps_per_second": 4.056,
       "step": 22
+    },
+    {
+      "epoch": 0.0007200945515454638,
+      "grad_norm": NaN,
+      "learning_rate": 8.645454235739903e-06,
+      "loss": 0.0,
+      "step": 23
+    },
+    {
+      "epoch": 0.0007514030103083101,
+      "grad_norm": NaN,
+      "learning_rate": 2.1852399266194314e-06,
+      "loss": 0.0,
+      "step": 24
+    },
+    {
+      "epoch": 0.0007514030103083101,
+      "eval_loss": NaN,
+      "eval_runtime": 3210.49,
+      "eval_samples_per_second": 4.189,
+      "eval_steps_per_second": 4.189,
+      "step": 24
     }
   ],
   "logging_steps": 1,
@@ -276,7 +298,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.5468848970989568e+16,
+  "total_flos": 1.6875107968352256e+16,
   "train_batch_size": 1,
   "trial_name": null,
   "trial_params": null