Training in progress, epoch 3, checkpoint

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4b03b11193656c1a3d536bab7a0679a12f7b4ef3b480229b7a017067ad82448d
+oid sha256:fdd44fc4623b0c6e26def0652e563c082035561f560ea4295f56472e47c88c6b
 size 14244

last-checkpoint/trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.12813186645507812,
-  "best_model_checkpoint": "./vit5_qqp/checkpoint-6570",
+  "best_metric": 0.12257253378629684,
+  "best_model_checkpoint": "./vit5_qqp/checkpoint-9855",
   "epoch": 3.0,
   "eval_steps": 500,
   "global_step": 9855,
@@ -156,6 +156,14 @@
       "learning_rate": 1.8011161846778284e-06,
       "loss": 0.1069,
       "step": 9500
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 0.12257253378629684,
+      "eval_runtime": 67.7838,
+      "eval_samples_per_second": 193.084,
+      "eval_steps_per_second": 5.37,
+      "step": 9855
     }
   ],
   "logging_steps": 500,