Upload 11 files

config.json

@@ -10,126 +10,108 @@
   "hidden_dropout_prob": 0.1,
   "hidden_size": 768,
   "id2label": {
-    "0": "\u7d42\u8eab",
-    "1": "\u8eca\u9ad4\u96aa",
-    "2": "\u9577\u671f\u7167\u8b77\u96aa",
-    "3": "\u6c7d\u8eca\u96aa",
-    "4": "\u7279\u5b9a\u50b7\u75c5\u96aa",
-    "5": "\u526f\u672c\u7406\u8ce0",
-    "6": "\u5931\u80fd\u96aa/\u5931\u80fd\u6276\u52a9\u96aa",
-    "7": "\u58fd\u96aa",
-    "8": "\u4fdd\u8b49\u7d66\u4ed8",
-    "9": "\u5be6\u652f\u578b\u610f\u5916\u96aa",
-    "10": "\u5931\u80fd\u589e\u984d",
-    "11": "\u6b98\u5ee2\u96aa",
-    "12": "\u5e74\u91d1\u4fdd\u96aa",
-    "13": "\u91cd\u5927\u75be\u75c5\u96aa",
-    "14": "\u6295\u4fdd",
-    "15": "\u7522\u96aa",
-    "16": "\u6a5f\u8eca\u96aa",
-    "17": "\u512a\u9ad4\u58fd\u96aa",
-    "18": "\u52de\u5065\u4fdd",
-    "19": "\u91cd\u5927\u71d2\u71d9\u50b7",
-    "20": "\u6295\u8cc7\u7406\u8ca1",
-    "21": "\u4e3b\u7d04",
-    "22": "\u65c5\u904a\u4e0d\u4fbf\u96aa",
-    "23": "\u4e00\u6b21\u7d66\u4ed8\u578b\u764c\u75c7\u96aa",
-    "24": "\u764c\u75c7\u96aa",
-    "25": "\u7642\u7a0b\u578b\u764c\u75c7\u96aa",
-    "26": "\u5b9a\u671f",
-    "27": "\u65c5\u904a\u5e73\u5b89\u96aa",
-    "28": "\u8d85\u984d\u8cac\u4efb\u96aa",
-    "29": "\u65e5\u984d\u578b\u610f\u5916\u96aa",
-    "30": "\u5eb7\u5065\u4eba\u58fd",
-    "31": "\u9644\u52a0\u4e58\u5ba2\u96aa",
-    "32": "\u7406\u8ce0",
-    "33": "\u7121\u7b49\u5f85\u671f",
-    "34": "\u91cd\u5927\u50b7\u75c5\u96aa/\u91cd\u5927\u75be\u75c5\u96aa",
-    "35": "\u6b98\u6276\u96aa",
-    "36": "\u7b2c\u4e09\u4eba\u8cac\u4efb\u96aa",
-    "37": "\u5916\u5e63\u5546\u54c1",
-    "38": "\u8c41\u514d\u4fdd\u8cbb",
-    "39": "\u5132\u84c4\u96aa",
-    "40": "\u9580\u8a3a\u624b\u8853",
-    "41": "\u5931\u80fd\u6276\u52a9\u96aa",
-    "42": "\u4fdd\u8b49\u7e8c\u4fdd",
-    "43": "\u99d5\u99db\u4eba\u50b7\u5bb3\u96aa",
-    "44": "\u8f49\u63db\u4f4f\u9662\u65e5\u984d",
-    "45": "\u5931\u80fd\u96aa",
-    "46": "\u5be6\u652f\u578b\u91ab\u7642\u96aa",
-    "47": "\u65e5\u984d\u578b\u91ab\u7642\u96aa",
-    "48": "\u91ab\u7642\u96aa",
-    "49": "\u9ad8CP\u503c",
-    "50": "\u610f\u5916\u96aa",
-    "51": "\u5718\u9ad4\u4fdd\u96aa",
-    "52": "\u89e3\u7d04",
-    "53": "\u5f37\u5236\u96aa",
-    "54": "\u9084\u672c\u578b",
-    "55": "\u9644\u7d04",
-    "56": "\u91cd\u5927\u50b7\u75c5\u96aa",
-    "57": "\u7aca\u76dc\u96aa"
+    "0": "\u91cd\u5927\u75be\u75c5\u96aa",
+    "1": "\u7522\u96aa",
+    "2": "\u6b98\u5ee2\u96aa",
+    "3": "\u7279\u5b9a\u50b7\u75c5\u96aa",
+    "4": "\u5132\u84c4\u96aa",
+    "5": "\u6b98\u6276\u96aa",
+    "6": "\u764c\u75c7\u96aa",
+    "7": "\u5718\u9ad4\u4fdd\u96aa",
+    "8": "\u99d5\u99db\u4eba\u50b7\u5bb3\u96aa",
+    "9": "\u4e00\u6b21\u7d66\u4ed8\u578b\u764c\u75c7\u96aa",
+    "10": "\u4e3b\u7d04",
+    "11": "\u4fdd\u8b49\u7d66\u4ed8",
+    "12": "\u5be6\u652f\u578b",
+    "13": "\u6c7d\u8eca\u96aa",
+    "14": "\u9084\u672c\u578b",
+    "15": "\u9644\u52a0\u4e58\u5ba2\u96aa",
+    "16": "\u7aca\u76dc\u96aa",
+    "17": "\u7d42\u8eab",
+    "18": "\u91ab\u7642\u96aa",
+    "19": "\u610f\u5916\u96aa",
+    "20": "\u58fd\u96aa",
+    "21": "\u9580\u8a3a\u624b\u8853",
+    "22": "\u91cd\u5927\u71d2\u71d9\u50b7",
+    "23": "\u65c5\u904a\u5e73\u5b89\u96aa",
+    "24": "\u5931\u80fd\u96aa",
+    "25": "\u7b2c\u4e09\u4eba\u8cac\u4efb\u96aa",
+    "26": "\u4fdd\u8b49\u7e8c\u4fdd",
+    "27": "\u5f37\u5236\u96aa",
+    "28": "\u7642\u7a0b\u578b\u764c\u75c7\u96aa",
+    "29": "\u5931\u80fd\u589e\u984d",
+    "30": "\u7121\u7b49\u5f85\u671f",
+    "31": "\u8c41\u514d\u4fdd\u8cbb",
+    "32": "\u5e74\u91d1\u4fdd\u96aa",
+    "33": "\u526f\u672c\u7406\u8ce0",
+    "34": "\u5b9a\u671f",
+    "35": "\u52de\u5065\u4fdd",
+    "36": "\u8d85\u984d\u8cac\u4efb\u96aa",
+    "37": "\u9644\u7d04",
+    "38": "\u5916\u5e63\u5546\u54c1",
+    "39": "\u8eca\u9ad4\u96aa",
+    "40": "\u91cd\u5927\u50b7\u75c5\u96aa",
+    "41": "\u8f49\u63db\u4f4f\u9662\u65e5\u984d",
+    "42": "\u9577\u671f\u7167\u8b77\u96aa",
+    "43": "\u65c5\u904a\u4e0d\u4fbf\u96aa",
+    "44": "\u6295\u8cc7\u7406\u8ca1",
+    "45": "\u6a5f\u8eca\u96aa",
+    "46": "\u5931\u80fd\u6276\u52a9\u96aa",
+    "47": "\u512a\u9ad4\u58fd\u96aa",
+    "48": "\u65e5\u984d\u578b"
   },
   "initializer_range": 0.02,
   "intermediate_size": 3072,
   "label2id": {
-    "\u4e00\u6b21\u7d66\u4ed8\u578b\u764c\u75c7\u96aa": 23,
-    "\u4e3b\u7d04": 21,
-    "\u4fdd\u8b49\u7d66\u4ed8": 8,
-    "\u4fdd\u8b49\u7e8c\u4fdd": 42,
-    "\u512a\u9ad4\u58fd\u96aa": 17,
-    "\u5132\u84c4\u96aa": 39,
-    "\u526f\u672c\u7406\u8ce0": 5,
-    "\u52de\u5065\u4fdd": 18,
-    "\u5718\u9ad4\u4fdd\u96aa": 51,
-    "\u58fd\u96aa": 7,
-    "\u5916\u5e63\u5546\u54c1": 37,
-    "\u5931\u80fd\u589e\u984d": 10,
-    "\u5931\u80fd\u6276\u52a9\u96aa": 41,
-    "\u5931\u80fd\u96aa": 45,
-    "\u5931\u80fd\u96aa/\u5931\u80fd\u6276\u52a9\u96aa": 6,
-    "\u5b9a\u671f": 26,
-    "\u5be6\u652f\u578b\u610f\u5916\u96aa": 9,
-    "\u5be6\u652f\u578b\u91ab\u7642\u96aa": 46,
-    "\u5e74\u91d1\u4fdd\u96aa": 12,
-    "\u5eb7\u5065\u4eba\u58fd": 30,
-    "\u5f37\u5236\u96aa": 53,
-    "\u610f\u5916\u96aa": 50,
-    "\u6295\u4fdd": 14,
-    "\u6295\u8cc7\u7406\u8ca1": 20,
-    "\u65c5\u904a\u4e0d\u4fbf\u96aa": 22,
-    "\u65c5\u904a\u5e73\u5b89\u96aa": 27,
-    "\u65e5\u984d\u578b\u610f\u5916\u96aa": 29,
-    "\u65e5\u984d\u578b\u91ab\u7642\u96aa": 47,
-    "\u6a5f\u8eca\u96aa": 16,
-    "\u6b98\u5ee2\u96aa": 11,
-    "\u6b98\u6276\u96aa": 35,
-    "\u6c7d\u8eca\u96aa": 3,
-    "\u7121\u7b49\u5f85\u671f": 33,
-    "\u7279\u5b9a\u50b7\u75c5\u96aa": 4,
-    "\u7406\u8ce0": 32,
-    "\u7522\u96aa": 15,
-    "\u7642\u7a0b\u578b\u764c\u75c7\u96aa": 25,
-    "\u764c\u75c7\u96aa": 24,
-    "\u7aca\u76dc\u96aa": 57,
-    "\u7b2c\u4e09\u4eba\u8cac\u4efb\u96aa": 36,
-    "\u7d42\u8eab": 0,
-    "\u89e3\u7d04": 52,
-    "\u8c41\u514d\u4fdd\u8cbb": 38,
-    "\u8d85\u984d\u8cac\u4efb\u96aa": 28,
-    "\u8eca\u9ad4\u96aa": 1,
-    "\u8f49\u63db\u4f4f\u9662\u65e5\u984d": 44,
-    "\u9084\u672c\u578b": 54,
-    "\u91ab\u7642\u96aa": 48,
-    "\u91cd\u5927\u50b7\u75c5\u96aa": 56,
-    "\u91cd\u5927\u50b7\u75c5\u96aa/\u91cd\u5927\u75be\u75c5\u96aa": 34,
-    "\u91cd\u5927\u71d2\u71d9\u50b7": 19,
-    "\u91cd\u5927\u75be\u75c5\u96aa": 13,
-    "\u9577\u671f\u7167\u8b77\u96aa": 2,
-    "\u9580\u8a3a\u624b\u8853": 40,
-    "\u9644\u52a0\u4e58\u5ba2\u96aa": 31,
-    "\u9644\u7d04": 55,
-    "\u99d5\u99db\u4eba\u50b7\u5bb3\u96aa": 43,
-    "\u9ad8CP\u503c": 49
+    "\u4e00\u6b21\u7d66\u4ed8\u578b\u764c\u75c7\u96aa": 9,
+    "\u4e3b\u7d04": 10,
+    "\u4fdd\u8b49\u7d66\u4ed8": 11,
+    "\u4fdd\u8b49\u7e8c\u4fdd": 26,
+    "\u512a\u9ad4\u58fd\u96aa": 47,
+    "\u5132\u84c4\u96aa": 4,
+    "\u526f\u672c\u7406\u8ce0": 33,
+    "\u52de\u5065\u4fdd": 35,
+    "\u5718\u9ad4\u4fdd\u96aa": 7,
+    "\u58fd\u96aa": 20,
+    "\u5916\u5e63\u5546\u54c1": 38,
+    "\u5931\u80fd\u589e\u984d": 29,
+    "\u5931\u80fd\u6276\u52a9\u96aa": 46,
+    "\u5931\u80fd\u96aa": 24,
+    "\u5b9a\u671f": 34,
+    "\u5be6\u652f\u578b": 12,
+    "\u5e74\u91d1\u4fdd\u96aa": 32,
+    "\u5f37\u5236\u96aa": 27,
+    "\u610f\u5916\u96aa": 19,
+    "\u6295\u8cc7\u7406\u8ca1": 44,
+    "\u65c5\u904a\u4e0d\u4fbf\u96aa": 43,
+    "\u65c5\u904a\u5e73\u5b89\u96aa": 23,
+    "\u65e5\u984d\u578b": 48,
+    "\u6a5f\u8eca\u96aa": 45,
+    "\u6b98\u5ee2\u96aa": 2,
+    "\u6b98\u6276\u96aa": 5,
+    "\u6c7d\u8eca\u96aa": 13,
+    "\u7121\u7b49\u5f85\u671f": 30,
+    "\u7279\u5b9a\u50b7\u75c5\u96aa": 3,
+    "\u7522\u96aa": 1,
+    "\u7642\u7a0b\u578b\u764c\u75c7\u96aa": 28,
+    "\u764c\u75c7\u96aa": 6,
+    "\u7aca\u76dc\u96aa": 16,
+    "\u7b2c\u4e09\u4eba\u8cac\u4efb\u96aa": 25,
+    "\u7d42\u8eab": 17,
+    "\u8c41\u514d\u4fdd\u8cbb": 31,
+    "\u8d85\u984d\u8cac\u4efb\u96aa": 36,
+    "\u8eca\u9ad4\u96aa": 39,
+    "\u8f49\u63db\u4f4f\u9662\u65e5\u984d": 41,
+    "\u9084\u672c\u578b": 14,
+    "\u91ab\u7642\u96aa": 18,
+    "\u91cd\u5927\u50b7\u75c5\u96aa": 40,
+    "\u91cd\u5927\u71d2\u71d9\u50b7": 22,
+    "\u91cd\u5927\u75be\u75c5\u96aa": 0,
+    "\u9577\u671f\u7167\u8b77\u96aa": 42,
+    "\u9580\u8a3a\u624b\u8853": 21,
+    "\u9644\u52a0\u4e58\u5ba2\u96aa": 15,
+    "\u9644\u7d04": 37,
+    "\u99d5\u99db\u4eba\u50b7\u5bb3\u96aa": 8
   },
   "layer_norm_eps": 1e-12,
   "max_position_embeddings": 512,
@@ -145,7 +127,7 @@
   "position_embedding_type": "absolute",
   "problem_type": "multi_label_classification",
   "torch_dtype": "float32",
-  "transformers_version": "4.30.2",
+  "transformers_version": "4.31.0",
   "type_vocab_size": 2,
   "use_cache": true,
   "vocab_size": 21128

optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5a76601140035efb6c21063a432df1a3db1623bf4ff1458b627bba8e2c13726e
-size 818615877
+oid sha256:f9642edc572d7b2e23173b6054c49f58211a2429fd9b09fb3bb2c18b636a77ba
+size 818560581

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a70cba2d4b660e85c8df0ffce5f11fe9ec2f2ccd02d8d59b303bd50d2f33a441
-size 409321781
+oid sha256:2ed670329e7c3c1d17b57e491064e921b4646ff79c9b2fe900a588c8cc4f92a7
+size 409289713

rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cb270d3ae860f11cf3c460aaa8e4c5def620c1c0810226037f53557bc70107bb
+oid sha256:9af5634aa9361a6940f3d1c32c45e51f9ffdc722fd2baafa8ebd03f20b659a4a
 size 14575

scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:80636f86a16567ef0f8d224d7dc105da5b294f11a4aa0269a9c057ac9ebd35ac
+oid sha256:fa6c7a04dbc7c613fe79004c534f37c1cebe30a25814c99e56027fb1c86fd918
 size 627

trainer_state.json

@@ -1,284 +1,278 @@
 {
-  "best_metric": 0.564568262274896,
-  "best_model_checkpoint": "bert-finetuned-insurence/checkpoint-3852",
+  "best_metric": 0.5812987875768144,
+  "best_model_checkpoint": "bert-finetuned-insurence/checkpoint-1504",
   "epoch": 20.0,
-  "global_step": 4280,
+  "global_step": 3760,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.0,
-      "eval_f1": 0.2651817608741332,
-      "eval_loss": 0.2465316206216812,
-      "eval_roc_auc": 0.5780009947992923,
-      "eval_runtime": 4.1824,
-      "eval_samples_per_second": 143.457,
-      "eval_steps_per_second": 2.391,
-      "step": 214
+      "eval_accuracy": 0.016666666666666666,
+      "eval_f1": 0.1856089499110094,
+      "eval_loss": 0.2537216246128082,
+      "eval_roc_auc": 0.5502226265229397,
+      "eval_runtime": 4.2613,
+      "eval_samples_per_second": 140.801,
+      "eval_steps_per_second": 2.347,
+      "step": 188
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.0033333333333333335,
-      "eval_f1": 0.4286243974786801,
-      "eval_loss": 0.2168939709663391,
-      "eval_roc_auc": 0.6460988356101586,
-      "eval_runtime": 4.1125,
-      "eval_samples_per_second": 145.895,
-      "eval_steps_per_second": 2.432,
-      "step": 428
+      "eval_accuracy": 0.02666666666666667,
+      "eval_f1": 0.4750710900473934,
+      "eval_loss": 0.2218438684940338,
+      "eval_roc_auc": 0.6751318414846118,
+      "eval_runtime": 4.2236,
+      "eval_samples_per_second": 142.059,
+      "eval_steps_per_second": 2.368,
+      "step": 376
     },
     {
-      "epoch": 2.34,
-      "learning_rate": 1.766355140186916e-05,
-      "loss": 0.2607,
+      "epoch": 2.66,
+      "learning_rate": 1.7340425531914894e-05,
+      "loss": 0.2657,
       "step": 500
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.01,
-      "eval_f1": 0.46524160229926353,
-      "eval_loss": 0.20735658705234528,
-      "eval_roc_auc": 0.6640500103615281,
-      "eval_runtime": 4.1378,
-      "eval_samples_per_second": 145.003,
-      "eval_steps_per_second": 2.417,
-      "step": 642
+      "eval_accuracy": 0.041666666666666664,
+      "eval_f1": 0.5161290322580645,
+      "eval_loss": 0.20993947982788086,
+      "eval_roc_auc": 0.6941419674142428,
+      "eval_runtime": 4.2298,
+      "eval_samples_per_second": 141.852,
+      "eval_steps_per_second": 2.364,
+      "step": 564
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.006666666666666667,
-      "eval_f1": 0.49712192569335417,
-      "eval_loss": 0.2018413543701172,
-      "eval_roc_auc": 0.6807712075579161,
-      "eval_runtime": 4.1419,
-      "eval_samples_per_second": 144.86,
-      "eval_steps_per_second": 2.414,
-      "step": 856
+      "eval_accuracy": 0.041666666666666664,
+      "eval_f1": 0.537253469685902,
+      "eval_loss": 0.20495453476905823,
+      "eval_roc_auc": 0.7082594844493445,
+      "eval_runtime": 4.2212,
+      "eval_samples_per_second": 142.139,
+      "eval_steps_per_second": 2.369,
+      "step": 752
     },
     {
-      "epoch": 4.67,
-      "learning_rate": 1.5327102803738318e-05,
-      "loss": 0.1957,
-      "step": 1000
+      "epoch": 5.0,
+      "eval_accuracy": 0.055,
+      "eval_f1": 0.5567817728729085,
+      "eval_loss": 0.20180459320545197,
+      "eval_roc_auc": 0.7213133386735834,
+      "eval_runtime": 4.2593,
+      "eval_samples_per_second": 140.868,
+      "eval_steps_per_second": 2.348,
+      "step": 940
     },
     {
-      "epoch": 5.0,
-      "eval_accuracy": 0.013333333333333334,
-      "eval_f1": 0.531224523071906,
-      "eval_loss": 0.1989612579345703,
-      "eval_roc_auc": 0.7047636174438967,
-      "eval_runtime": 4.1224,
-      "eval_samples_per_second": 145.547,
-      "eval_steps_per_second": 2.426,
-      "step": 1070
+      "epoch": 5.32,
+      "learning_rate": 1.4680851063829789e-05,
+      "loss": 0.1948,
+      "step": 1000
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.028333333333333332,
-      "eval_f1": 0.552503531627688,
-      "eval_loss": 0.19641733169555664,
-      "eval_roc_auc": 0.7204743245073841,
-      "eval_runtime": 4.1162,
-      "eval_samples_per_second": 145.765,
-      "eval_steps_per_second": 2.429,
-      "step": 1284
+      "eval_accuracy": 0.04666666666666667,
+      "eval_f1": 0.5600428877769836,
+      "eval_loss": 0.20023241639137268,
+      "eval_roc_auc": 0.7222440876299687,
+      "eval_runtime": 4.2842,
+      "eval_samples_per_second": 140.05,
+      "eval_steps_per_second": 2.334,
+      "step": 1128
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.021666666666666667,
-      "eval_f1": 0.5480784313725491,
-      "eval_loss": 0.19820699095726013,
-      "eval_roc_auc": 0.7184704069589273,
-      "eval_runtime": 4.1253,
-      "eval_samples_per_second": 145.442,
-      "eval_steps_per_second": 2.424,
-      "step": 1498
+      "eval_accuracy": 0.05,
+      "eval_f1": 0.5645614035087719,
+      "eval_loss": 0.2023993879556656,
+      "eval_roc_auc": 0.72737489130751,
+      "eval_runtime": 4.2454,
+      "eval_samples_per_second": 141.329,
+      "eval_steps_per_second": 2.355,
+      "step": 1316
     },
     {
-      "epoch": 7.01,
-      "learning_rate": 1.2990654205607478e-05,
-      "loss": 0.1749,
+      "epoch": 7.98,
+      "learning_rate": 1.2021276595744681e-05,
+      "loss": 0.1685,
       "step": 1500
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.025,
-      "eval_f1": 0.541625534715367,
-      "eval_loss": 0.19906778633594513,
-      "eval_roc_auc": 0.7081857426146021,
-      "eval_runtime": 4.1299,
-      "eval_samples_per_second": 145.284,
-      "eval_steps_per_second": 2.421,
-      "step": 1712
+      "eval_accuracy": 0.05,
+      "eval_f1": 0.5812987875768144,
+      "eval_loss": 0.20405741035938263,
+      "eval_roc_auc": 0.7451393429382976,
+      "eval_runtime": 4.2486,
+      "eval_samples_per_second": 141.223,
+      "eval_steps_per_second": 2.354,
+      "step": 1504
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.023333333333333334,
-      "eval_f1": 0.5581826611033842,
-      "eval_loss": 0.20057238638401031,
-      "eval_roc_auc": 0.7257264689844256,
-      "eval_runtime": 4.123,
-      "eval_samples_per_second": 145.527,
-      "eval_steps_per_second": 2.425,
-      "step": 1926
-    },
-    {
-      "epoch": 9.35,
-      "learning_rate": 1.0654205607476635e-05,
-      "loss": 0.1552,
-      "step": 2000
+      "eval_accuracy": 0.051666666666666666,
+      "eval_f1": 0.5782645324347093,
+      "eval_loss": 0.2039116472005844,
+      "eval_roc_auc": 0.7410205309812413,
+      "eval_runtime": 4.2832,
+      "eval_samples_per_second": 140.082,
+      "eval_steps_per_second": 2.335,
+      "step": 1692
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.023333333333333334,
-      "eval_f1": 0.5491239048811013,
-      "eval_loss": 0.20264221727848053,
-      "eval_roc_auc": 0.7193901260078247,
-      "eval_runtime": 4.1359,
-      "eval_samples_per_second": 145.073,
-      "eval_steps_per_second": 2.418,
-      "step": 2140
+      "eval_accuracy": 0.05333333333333334,
+      "eval_f1": 0.5611961057023644,
+      "eval_loss": 0.20895147323608398,
+      "eval_roc_auc": 0.7272260462547598,
+      "eval_runtime": 4.2515,
+      "eval_samples_per_second": 141.128,
+      "eval_steps_per_second": 2.352,
+      "step": 1880
     },
     {
-      "epoch": 11.0,
-      "eval_accuracy": 0.023333333333333334,
-      "eval_f1": 0.5452513966480447,
-      "eval_loss": 0.20413073897361755,
-      "eval_roc_auc": 0.7144236705813142,
-      "eval_runtime": 4.1888,
-      "eval_samples_per_second": 143.238,
-      "eval_steps_per_second": 2.387,
-      "step": 2354
+      "epoch": 10.64,
+      "learning_rate": 9.361702127659576e-06,
+      "loss": 0.1481,
+      "step": 2000
     },
     {
-      "epoch": 11.68,
-      "learning_rate": 8.317757009345795e-06,
-      "loss": 0.142,
-      "step": 2500
+      "epoch": 11.0,
+      "eval_accuracy": 0.05,
+      "eval_f1": 0.5766101694915254,
+      "eval_loss": 0.20634876191616058,
+      "eval_roc_auc": 0.7391470139862534,
+      "eval_runtime": 4.2535,
+      "eval_samples_per_second": 141.06,
+      "eval_steps_per_second": 2.351,
+      "step": 2068
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.03,
-      "eval_f1": 0.5433158480368782,
-      "eval_loss": 0.2068767249584198,
-      "eval_roc_auc": 0.7141540882215025,
-      "eval_runtime": 4.1429,
-      "eval_samples_per_second": 144.826,
-      "eval_steps_per_second": 2.414,
-      "step": 2568
+      "eval_accuracy": 0.05333333333333334,
+      "eval_f1": 0.5601417183348095,
+      "eval_loss": 0.2106322944164276,
+      "eval_roc_auc": 0.7236795284079115,
+      "eval_runtime": 4.2654,
+      "eval_samples_per_second": 140.666,
+      "eval_steps_per_second": 2.344,
+      "step": 2256
     },
     {
       "epoch": 13.0,
-      "eval_accuracy": 0.02666666666666667,
-      "eval_f1": 0.5504587155963303,
-      "eval_loss": 0.20743216574192047,
-      "eval_roc_auc": 0.7182797203748162,
-      "eval_runtime": 4.1314,
-      "eval_samples_per_second": 145.228,
-      "eval_steps_per_second": 2.42,
-      "step": 2782
+      "eval_accuracy": 0.05,
+      "eval_f1": 0.5558127114117857,
+      "eval_loss": 0.21359555423259735,
+      "eval_roc_auc": 0.7208235913661607,
+      "eval_runtime": 4.2657,
+      "eval_samples_per_second": 140.656,
+      "eval_steps_per_second": 2.344,
+      "step": 2444
     },
     {
-      "epoch": 14.0,
-      "eval_accuracy": 0.02666666666666667,
-      "eval_f1": 0.5574488041269346,
-      "eval_loss": 0.20869675278663635,
-      "eval_roc_auc": 0.7234869573231687,
-      "eval_runtime": 4.1354,
-      "eval_samples_per_second": 145.09,
-      "eval_steps_per_second": 2.418,
-      "step": 2996
+      "epoch": 13.3,
+      "learning_rate": 6.702127659574469e-06,
+      "loss": 0.1331,
+      "step": 2500
     },
     {
-      "epoch": 14.02,
-      "learning_rate": 5.981308411214953e-06,
-      "loss": 0.1314,
-      "step": 3000
+      "epoch": 14.0,
+      "eval_accuracy": 0.04,
+      "eval_f1": 0.5686879068174031,
+      "eval_loss": 0.2141793817281723,
+      "eval_roc_auc": 0.7333805433804568,
+      "eval_runtime": 4.2245,
+      "eval_samples_per_second": 142.028,
+      "eval_steps_per_second": 2.367,
+      "step": 2632
     },
     {
       "epoch": 15.0,
-      "eval_accuracy": 0.028333333333333332,
-      "eval_f1": 0.5433563071004968,
-      "eval_loss": 0.2121913880109787,
-      "eval_roc_auc": 0.712902953754437,
-      "eval_runtime": 4.165,
-      "eval_samples_per_second": 144.058,
-      "eval_steps_per_second": 2.401,
-      "step": 3210
+      "eval_accuracy": 0.058333333333333334,
+      "eval_f1": 0.5659075224292616,
+      "eval_loss": 0.21738983690738678,
+      "eval_roc_auc": 0.730793042125031,
+      "eval_runtime": 4.2924,
+      "eval_samples_per_second": 139.781,
+      "eval_steps_per_second": 2.33,
+      "step": 2820
     },
     {
-      "epoch": 16.0,
-      "eval_accuracy": 0.03166666666666667,
-      "eval_f1": 0.5553453169347209,
-      "eval_loss": 0.21151989698410034,
-      "eval_roc_auc": 0.7210906101371354,
-      "eval_runtime": 4.1328,
-      "eval_samples_per_second": 145.182,
-      "eval_steps_per_second": 2.42,
-      "step": 3424
+      "epoch": 15.96,
+      "learning_rate": 4.042553191489362e-06,
+      "loss": 0.1222,
+      "step": 3000
     },
     {
-      "epoch": 16.36,
-      "learning_rate": 3.6448598130841123e-06,
-      "loss": 0.1242,
-      "step": 3500
+      "epoch": 16.0,
+      "eval_accuracy": 0.04,
+      "eval_f1": 0.5679096818337325,
+      "eval_loss": 0.21772129833698273,
+      "eval_roc_auc": 0.7332271515025567,
+      "eval_runtime": 4.2711,
+      "eval_samples_per_second": 140.478,
+      "eval_steps_per_second": 2.341,
+      "step": 3008
     },
     {
       "epoch": 17.0,
-      "eval_accuracy": 0.03166666666666667,
-      "eval_f1": 0.5528838598145529,
-      "eval_loss": 0.21267028152942657,
-      "eval_roc_auc": 0.720453986434248,
-      "eval_runtime": 4.1485,
-      "eval_samples_per_second": 144.63,
-      "eval_steps_per_second": 2.41,
-      "step": 3638
+      "eval_accuracy": 0.041666666666666664,
+      "eval_f1": 0.5633460603956959,
+      "eval_loss": 0.21912114322185516,
+      "eval_roc_auc": 0.7285610702838653,
+      "eval_runtime": 4.246,
+      "eval_samples_per_second": 141.309,
+      "eval_steps_per_second": 2.355,
+      "step": 3196
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.028333333333333332,
-      "eval_f1": 0.564568262274896,
-      "eval_loss": 0.21246589720249176,
-      "eval_roc_auc": 0.7294850597946638,
-      "eval_runtime": 4.1681,
-      "eval_samples_per_second": 143.949,
-      "eval_steps_per_second": 2.399,
-      "step": 3852
+      "eval_accuracy": 0.045,
+      "eval_f1": 0.5691861457089235,
+      "eval_loss": 0.21997326612472534,
+      "eval_roc_auc": 0.7342966651772544,
+      "eval_runtime": 4.2727,
+      "eval_samples_per_second": 140.425,
+      "eval_steps_per_second": 2.34,
+      "step": 3384
     },
     {
-      "epoch": 18.69,
-      "learning_rate": 1.308411214953271e-06,
-      "loss": 0.1194,
-      "step": 4000
+      "epoch": 18.62,
+      "learning_rate": 1.3829787234042555e-06,
+      "loss": 0.1156,
+      "step": 3500
     },
     {
       "epoch": 19.0,
-      "eval_accuracy": 0.03333333333333333,
-      "eval_f1": 0.5572077007356393,
-      "eval_loss": 0.21363653242588043,
-      "eval_roc_auc": 0.7231682360522511,
-      "eval_runtime": 4.124,
-      "eval_samples_per_second": 145.491,
-      "eval_steps_per_second": 2.425,
-      "step": 4066
+      "eval_accuracy": 0.05,
+      "eval_f1": 0.5648404993065187,
+      "eval_loss": 0.2216050773859024,
+      "eval_roc_auc": 0.7294638689597607,
+      "eval_runtime": 4.2818,
+      "eval_samples_per_second": 140.127,
+      "eval_steps_per_second": 2.335,
+      "step": 3572
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.03166666666666667,
-      "eval_f1": 0.5593989669744873,
-      "eval_loss": 0.2135751098394394,
-      "eval_roc_auc": 0.7242125772444862,
-      "eval_runtime": 4.1303,
-      "eval_samples_per_second": 145.268,
-      "eval_steps_per_second": 2.421,
-      "step": 4280
+      "eval_accuracy": 0.04666666666666667,
+      "eval_f1": 0.5661549077108851,
+      "eval_loss": 0.22130462527275085,
+      "eval_roc_auc": 0.7309435085710135,
+      "eval_runtime": 4.2738,
+      "eval_samples_per_second": 140.39,
+      "eval_steps_per_second": 2.34,
+      "step": 3760
     }
   ],
-  "max_steps": 4280,
+  "max_steps": 3760,
   "num_train_epochs": 20,
-  "total_flos": 1.79926828336128e+16,
+  "total_flos": 1.579595742022656e+16,
   "trial_name": null,
   "trial_params": null
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:958ccbf5e147c249c07aac2a8acbe9a19c5128f7ab046f00ae2bebff4c73ad57
+oid sha256:8a70ff22bb0dc7f2725e461f5d604d3e51d14bf1c277ff029b006171115f80e3
 size 3963