Upload folder using huggingface_hub

training_checkpoints/checkpoint-643/adapter_config.json

@@ -21,8 +21,8 @@
   "revision": null,
   "target_modules": [
     "key",
-    "dense",
     "value",
+    "dense",
     "query"
   ],
   "task_type": "SEQ_CLS",

training_checkpoints/checkpoint-643/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c53182b2c2c69c0d0d5eced0d73bd90fb6351291f5abd2a3f553054ac412349b
+oid sha256:807626fbe31da1676bfb47c4e711c05c686ee9ed330a1aa1123f64f769be412f
 size 104549068

training_checkpoints/checkpoint-643/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f922e3622a510677ec7a22a5cfc432084a4335086bbca12acbf57a0b4e684c29
+oid sha256:eb13d7468a59df546c16ce0cf0e090673b2cafc0fcfcca2c1ccc7ac3de7586e6
 size 21646778

training_checkpoints/checkpoint-643/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6438ed6cfc5010cc62ed940c59c896be41c9578565a6a8d984b6ec712fc0303a
+oid sha256:74888fb227ac28db8dee18137a1df7903e74e06fc48019f644883eaf139f2072
 size 1064

training_checkpoints/checkpoint-643/trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.4262450946907797,
-  "best_model_checkpoint": "legal-bert-lora/checkpoint-482",
+  "best_metric": 0.46141164864389705,
+  "best_model_checkpoint": "legal-bert-lora/checkpoint-643",
   "epoch": 4.0,
   "eval_steps": 500,
   "global_step": 643,
@@ -10,102 +10,102 @@
   "log_history": [
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.6297443841982959,
-      "eval_f1_macro": 0.274206359888179,
-      "eval_f1_micro": 0.6297443841982959,
-      "eval_f1_weighted": 0.5693924444438366,
-      "eval_loss": 1.3252344131469727,
-      "eval_macro_fpr": 0.04170202326759921,
-      "eval_macro_sensitivity": 0.31095312319919677,
-      "eval_macro_specificity": 0.9716859606487191,
-      "eval_precision": 0.5643117258906628,
-      "eval_precision_macro": 0.2864620870649182,
-      "eval_recall": 0.6297443841982959,
-      "eval_recall_macro": 0.31095312319919677,
-      "eval_runtime": 29.2455,
-      "eval_samples_per_second": 44.143,
-      "eval_steps_per_second": 5.539,
-      "eval_weighted_fpr": 0.0403035413153457,
-      "eval_weighted_sensitivity": 0.6297443841982959,
-      "eval_weighted_specificity": 0.945545025532491,
+      "eval_accuracy": 0.6374903175832688,
+      "eval_f1_macro": 0.3394539856434376,
+      "eval_f1_micro": 0.6374903175832688,
+      "eval_f1_weighted": 0.5765523433851403,
+      "eval_loss": 1.3127473592758179,
+      "eval_macro_fpr": 0.040137717078130394,
+      "eval_macro_sensitivity": 0.4017115530536667,
+      "eval_macro_specificity": 0.9725650393599338,
+      "eval_precision": 0.5466227876047401,
+      "eval_precision_macro": 0.31141306924471374,
+      "eval_recall": 0.6374903175832688,
+      "eval_recall_macro": 0.4017115530536667,
+      "eval_runtime": 44.4128,
+      "eval_samples_per_second": 29.068,
+      "eval_steps_per_second": 3.648,
+      "eval_weighted_fpr": 0.0390325271059216,
+      "eval_weighted_sensitivity": 0.6374903175832688,
+      "eval_weighted_specificity": 0.950985272815738,
       "step": 160
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.7312161115414407,
-      "eval_f1_macro": 0.39668888106795264,
-      "eval_f1_micro": 0.7312161115414407,
-      "eval_f1_weighted": 0.6996238601563893,
-      "eval_loss": 0.8870222568511963,
-      "eval_macro_fpr": 0.025654492164758257,
-      "eval_macro_sensitivity": 0.4525442260434048,
-      "eval_macro_specificity": 0.9800113592544849,
-      "eval_precision": 0.687310022738013,
-      "eval_precision_macro": 0.37417463878901125,
-      "eval_recall": 0.7312161115414407,
-      "eval_recall_macro": 0.4525442260434048,
-      "eval_runtime": 31.4769,
-      "eval_samples_per_second": 41.014,
-      "eval_steps_per_second": 5.147,
-      "eval_weighted_fpr": 0.025584310255843103,
-      "eval_weighted_sensitivity": 0.7312161115414407,
-      "eval_weighted_specificity": 0.9689542772758329,
+      "eval_accuracy": 0.7273431448489543,
+      "eval_f1_macro": 0.3969800602789428,
+      "eval_f1_micro": 0.7273431448489543,
+      "eval_f1_weighted": 0.6960627383550935,
+      "eval_loss": 0.8160030841827393,
+      "eval_macro_fpr": 0.026417996388603746,
+      "eval_macro_sensitivity": 0.450135625445356,
+      "eval_macro_specificity": 0.9796396430402219,
+      "eval_precision": 0.6997929823446861,
+      "eval_precision_macro": 0.39453164470926966,
+      "eval_recall": 0.7273431448489543,
+      "eval_recall_macro": 0.450135625445356,
+      "eval_runtime": 34.3148,
+      "eval_samples_per_second": 37.622,
+      "eval_steps_per_second": 4.721,
+      "eval_weighted_fpr": 0.026077937472218107,
+      "eval_weighted_sensitivity": 0.7273431448489543,
+      "eval_weighted_specificity": 0.9672515007543734,
       "step": 321
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.7482571649883811,
-      "eval_f1_macro": 0.4262450946907797,
-      "eval_f1_micro": 0.7482571649883811,
-      "eval_f1_weighted": 0.7282078768513273,
-      "eval_loss": 0.7794041037559509,
-      "eval_macro_fpr": 0.023941075707085962,
-      "eval_macro_sensitivity": 0.46803400264857875,
-      "eval_macro_specificity": 0.9812882861560729,
-      "eval_precision": 0.716881212213352,
-      "eval_precision_macro": 0.40594566191566384,
-      "eval_recall": 0.7482571649883811,
-      "eval_recall_macro": 0.46803400264857875,
-      "eval_runtime": 31.9176,
-      "eval_samples_per_second": 40.448,
-      "eval_steps_per_second": 5.076,
-      "eval_weighted_fpr": 0.02346739836811322,
-      "eval_weighted_sensitivity": 0.7482571649883811,
-      "eval_weighted_specificity": 0.9710671273527125,
+      "eval_accuracy": 0.7443841982958946,
+      "eval_f1_macro": 0.4210575074216338,
+      "eval_f1_micro": 0.7443841982958946,
+      "eval_f1_weighted": 0.7223603623418946,
+      "eval_loss": 0.7322217226028442,
+      "eval_macro_fpr": 0.024451015698283406,
+      "eval_macro_sensitivity": 0.46251578703813245,
+      "eval_macro_specificity": 0.9809603839363172,
+      "eval_precision": 0.711002444877192,
+      "eval_precision_macro": 0.4030436676815596,
+      "eval_recall": 0.7443841982958946,
+      "eval_recall_macro": 0.46251578703813245,
+      "eval_runtime": 32.8708,
+      "eval_samples_per_second": 39.275,
+      "eval_steps_per_second": 4.928,
+      "eval_weighted_fpr": 0.023940800928612886,
+      "eval_weighted_sensitivity": 0.7443841982958946,
+      "eval_weighted_specificity": 0.9700215607488639,
       "step": 482
     },
     {
       "epoch": 3.11,
-      "learning_rate": 3.4375e-05,
-      "loss": 1.2835,
+      "learning_rate": 3.958333333333333e-05,
+      "loss": 1.2229,
       "step": 500
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.7443841982958946,
-      "eval_f1_macro": 0.4099967964216515,
-      "eval_f1_micro": 0.7443841982958946,
-      "eval_f1_weighted": 0.7145941103131069,
-      "eval_loss": 0.7480739951133728,
-      "eval_macro_fpr": 0.024347730036769315,
-      "eval_macro_sensitivity": 0.45879715060247334,
-      "eval_macro_specificity": 0.9809610383366641,
-      "eval_precision": 0.7084810945380595,
-      "eval_precision_macro": 0.39966194371515146,
-      "eval_recall": 0.7443841982958946,
-      "eval_recall_macro": 0.45879715060247334,
-      "eval_runtime": 31.4007,
-      "eval_samples_per_second": 41.114,
-      "eval_steps_per_second": 5.159,
-      "eval_weighted_fpr": 0.023940800928612886,
-      "eval_weighted_sensitivity": 0.7443841982958946,
-      "eval_weighted_specificity": 0.9700313767540658,
+      "eval_accuracy": 0.7536793183578622,
+      "eval_f1_macro": 0.46141164864389705,
+      "eval_f1_micro": 0.7536793183578622,
+      "eval_f1_weighted": 0.7337325056665881,
+      "eval_loss": 0.7034952640533447,
+      "eval_macro_fpr": 0.023326126599014015,
+      "eval_macro_sensitivity": 0.4865735295815666,
+      "eval_macro_specificity": 0.9815978320880049,
+      "eval_precision": 0.7340611054176746,
+      "eval_precision_macro": 0.4982390488816621,
+      "eval_recall": 0.7536793183578622,
+      "eval_recall_macro": 0.4865735295815666,
+      "eval_runtime": 33.282,
+      "eval_samples_per_second": 38.79,
+      "eval_steps_per_second": 4.868,
+      "eval_weighted_fpr": 0.022812051649928264,
+      "eval_weighted_sensitivity": 0.7536793183578622,
+      "eval_weighted_specificity": 0.9702881629622103,
       "step": 643
     }
   ],
   "logging_steps": 500,
-  "max_steps": 1600,
-  "num_train_epochs": 10,
+  "max_steps": 2400,
+  "num_train_epochs": 15,
   "save_steps": 500,
   "total_flos": 5577944869453824.0,
   "trial_name": null,

training_checkpoints/checkpoint-643/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1be05b6bb510587fe353991f6f2836b34a5909cbc5adc58389938bd163228a1f
+oid sha256:36a33fe5941ace0a96599f0a2229d1ce6d7002fe79c5372925862ab481915414
 size 4600