add more optimization for the model achieve accuracy 96.42

optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:47e33277c95e6cbe30a488f7d174fc4983a0c3df1805e46371673ca51e478fea
+oid sha256:84ebd6a965ada73992ddc5d156df7f076add8a3cd9961d6d19b6abd4b9f7d42f
 size 2693488029

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:89bbbbd6f722c9a0de62650c24f46ea8ed1b71ee078058f7c2ca629dbe84ee9e
+oid sha256:7a896430556f8a49a84207742f7ab56fb2ed7fd003512c4d74d6693cf2330707
 size 1346791341

rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:76a963db23e1ffd2f6f4f4df078cdc6ee636c7665596bfab5d059dc3baa67dca
+oid sha256:d8bd75f8d4290e14f52b50f9b9f28e231b4c912e2fc632f615ea313424af2402
 size 14503

scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:19cfba3471e0fe55718aff3c8ffb4d09ef0ba480c0e672e291ea757073d3e490
+oid sha256:b69b4c315a8a6ee315e3092dd746705b93da4e03bf145a6ce0e6fb151e2af5f1
 size 623

trainer_state.json

@@ -1,88 +1,70 @@
 {
-  "best_metric": 0.11405563354492188,
-  "best_model_checkpoint": "BERT-HARD-balanced/checkpoint-1500",
-  "epoch": 1.680672268907563,
-  "global_step": 2000,
+  "best_metric": 0.11092716455459595,
+  "best_model_checkpoint": "BERT-HARD-balanced-2/checkpoint-500",
+  "epoch": 2.5210084033613445,
+  "global_step": 1500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.42,
-      "learning_rate": 4.2997198879551826e-05,
-      "loss": 0.1682,
+      "epoch": 0.84,
+      "learning_rate": 7.198879551820729e-06,
+      "loss": 0.1421,
       "step": 500
     },
     {
-      "epoch": 0.42,
-      "eval_accuracy": 0.950212866603595,
-      "eval_f1": 0.9498032669607726,
-      "eval_loss": 0.13225148618221283,
-      "eval_precision": 0.9576821351286366,
-      "eval_recall": 0.9420529801324503,
-      "eval_runtime": 6.9677,
-      "eval_samples_per_second": 1213.595,
-      "eval_steps_per_second": 19.088,
+      "epoch": 0.84,
+      "eval_accuracy": 0.9602649006622517,
+      "eval_f1": 0.9607751575998131,
+      "eval_loss": 0.11092716455459595,
+      "eval_precision": 0.9485938220378054,
+      "eval_recall": 0.9732734153263954,
+      "eval_runtime": 6.7747,
+      "eval_samples_per_second": 1248.164,
+      "eval_steps_per_second": 9.89,
       "step": 500
     },
     {
-      "epoch": 0.84,
-      "learning_rate": 3.5994397759103643e-05,
-      "loss": 0.1337,
+      "epoch": 1.68,
+      "learning_rate": 4.397759103641457e-06,
+      "loss": 0.0981,
       "step": 1000
     },
     {
-      "epoch": 0.84,
-      "eval_accuracy": 0.9571901608325449,
-      "eval_f1": 0.9578481602235678,
-      "eval_loss": 0.12283609807491302,
-      "eval_precision": 0.943348623853211,
-      "eval_recall": 0.9728003784295175,
-      "eval_runtime": 6.9706,
-      "eval_samples_per_second": 1213.091,
-      "eval_steps_per_second": 19.08,
+      "epoch": 1.68,
+      "eval_accuracy": 0.962038789025544,
+      "eval_f1": 0.9620881067674502,
+      "eval_loss": 0.11493375152349472,
+      "eval_precision": 0.9608398207124321,
+      "eval_recall": 0.9633396404919584,
+      "eval_runtime": 6.7211,
+      "eval_samples_per_second": 1258.119,
+      "eval_steps_per_second": 9.969,
       "step": 1000
     },
     {
-      "epoch": 1.26,
-      "learning_rate": 2.8991596638655467e-05,
-      "loss": 0.1132,
+      "epoch": 2.52,
+      "learning_rate": 1.5966386554621848e-06,
+      "loss": 0.0794,
       "step": 1500
     },
     {
-      "epoch": 1.26,
-      "eval_accuracy": 0.9605014191106906,
-      "eval_f1": 0.9605294256676908,
-      "eval_loss": 0.11405563354492188,
-      "eval_precision": 0.9598488427019367,
-      "eval_recall": 0.9612109744560076,
-      "eval_runtime": 7.0035,
-      "eval_samples_per_second": 1207.4,
-      "eval_steps_per_second": 18.991,
+      "epoch": 2.52,
+      "eval_accuracy": 0.962038789025544,
+      "eval_f1": 0.9624077760861928,
+      "eval_loss": 0.12313080579042435,
+      "eval_precision": 0.9531431222454188,
+      "eval_recall": 0.9718543046357616,
+      "eval_runtime": 6.7353,
+      "eval_samples_per_second": 1255.468,
+      "eval_steps_per_second": 9.948,
       "step": 1500
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 2.1988795518207285e-05,
-      "loss": 0.0999,
-      "step": 2000
-    },
-    {
-      "epoch": 1.68,
-      "eval_accuracy": 0.9609744560075686,
-      "eval_f1": 0.9610940815845319,
-      "eval_loss": 0.1178417056798935,
-      "eval_precision": 0.9581570286788904,
-      "eval_recall": 0.9640491958372753,
-      "eval_runtime": 6.9707,
-      "eval_samples_per_second": 1213.078,
-      "eval_steps_per_second": 19.08,
-      "step": 2000
     }
   ],
-  "max_steps": 3570,
+  "max_steps": 1785,
   "num_train_epochs": 3,
-  "total_flos": 1.4904145310863872e+16,
+  "total_flos": 2.2352839715103744e+16,
   "trial_name": null,
   "trial_params": null
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5a8a5638fed315463a869a97c283cb7ba57b1e142b1bbb12e94d66133ac47bd2
+oid sha256:4a05b36be39b1bf0bf08b95a78f70802e97e153d8839e252ac9a320c083092cc
 size 2991