Training in progress, step 8800

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:56290d577bb399219d12b202be37ca101e6a6f0a9668d1d0cfa2d9aad00953f3
+oid sha256:8c7978bf8054801da6f4c756ac3ee82d92c0f486af86b40dc0bf28323cf03296
 size 997349515

last-checkpoint/pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fd83f389d0c97375b7f818c8e82d986e4ae56a618b8b2c299c81b885ff7a1e3f
+oid sha256:19f88d4ac00fa7feb29c8dd174036ce80b27f550f2e7dcaa40ee80769e3c83a2
 size 498663405

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d35e892693875b95b0138807496870cd0cc89c55b72b1b6dd1430d39771a0ee
+oid sha256:fc559e8f3be7931753acd7e14135ef3c355c45429768068a18a2cdbd924ae84e
 size 14503

last-checkpoint/scaler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aab32762904ae4476cadd93f071777e19190a68ba8e7e9fe09145ebc857602b7
+oid sha256:0ac5f79294afe5b58702f8cfd595843c97cf8ecf3eb18e0b1f699ed162ad0894
 size 559

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f5835770cef613970a169521bc333c3ce82fb4729d6ec5ebe78749ca3671853b
+oid sha256:7ac05aa2ac2c93c43b69f88d2a334721a2027f6a3102b80f21e4e6745984e25d
 size 623

last-checkpoint/trainer_state.json

@@ -1,8 +1,8 @@
 {
-  "best_metric": 0.907479569921263,
-  "best_model_checkpoint": "/home2/s5431786/nlp-final-project/results/roberta-base-e-snli-classification-nli-base/checkpoint-7200",
-  "epoch": 0.8387698042870456,
-  "global_step": 7200,
+  "best_metric": 0.9120778013038088,
+  "best_model_checkpoint": "/home2/s5431786/nlp-final-project/results/roberta-base-e-snli-classification-nli-base/checkpoint-8000",
+  "epoch": 1.0251630941286114,
+  "global_step": 8800,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -294,11 +294,75 @@
       "eval_samples_per_second": 2361.347,
       "eval_steps_per_second": 36.949,
       "step": 7200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.421517331589274e-06,
+      "loss": 0.3309,
+      "step": 7600
+    },
+    {
+      "epoch": 0.89,
+      "eval_accuracy": 0.9061166429587483,
+      "eval_f1": 0.9058061283874342,
+      "eval_loss": 0.2577354609966278,
+      "eval_runtime": 4.4559,
+      "eval_samples_per_second": 2208.773,
+      "eval_steps_per_second": 34.561,
+      "step": 7600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.258011772400262e-06,
+      "loss": 0.3236,
+      "step": 8000
+    },
+    {
+      "epoch": 0.93,
+      "eval_accuracy": 0.9121113594797805,
+      "eval_f1": 0.9120778013038088,
+      "eval_loss": 0.25613412261009216,
+      "eval_runtime": 4.1922,
+      "eval_samples_per_second": 2347.706,
+      "eval_steps_per_second": 36.735,
+      "step": 8000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.09450621321125e-06,
+      "loss": 0.3183,
+      "step": 8400
+    },
+    {
+      "epoch": 0.98,
+      "eval_accuracy": 0.908758382442593,
+      "eval_f1": 0.9083914158392551,
+      "eval_loss": 0.2555808424949646,
+      "eval_runtime": 4.2479,
+      "eval_samples_per_second": 2316.917,
+      "eval_steps_per_second": 36.253,
+      "step": 8400
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 6.931000654022237e-06,
+      "loss": 0.3022,
+      "step": 8800
+    },
+    {
+      "epoch": 1.03,
+      "eval_accuracy": 0.906421459053038,
+      "eval_f1": 0.9056135727779063,
+      "eval_loss": 0.2667880356311798,
+      "eval_runtime": 4.4383,
+      "eval_samples_per_second": 2217.519,
+      "eval_steps_per_second": 34.698,
+      "step": 8800
     }
   ],
   "max_steps": 25752,
   "num_train_epochs": 3,
-  "total_flos": 1.2570506645372928e+16,
+  "total_flos": 1.536997071500478e+16,
   "trial_name": null,
   "trial_params": null
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fd83f389d0c97375b7f818c8e82d986e4ae56a618b8b2c299c81b885ff7a1e3f
+oid sha256:19f88d4ac00fa7feb29c8dd174036ce80b27f550f2e7dcaa40ee80769e3c83a2
 size 498663405