Upload 12 files

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7f8f509219b929f3addf6eef83576b1688eb0a1497dfb0647bbc12353db31306
+oid sha256:63341a8be72e1aaf8a935da18537db767e623966c2e183e44140e4e5764e777c
 size 901339568

optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:21db5a3775c1bccc4edb88c345b076b31e12d3016f9287bdd7bce760470ee43d
+oid sha256:a49a0d3f0c94d68183a3813996c0843b0adc852b73546dcbdb5fde4510308d65
 size 227854586

rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8ca98b8084893ea4bfdd9710f4315a84edb01c91507a364764caf1650ea4774d
+oid sha256:29bc513d119655394ebc9d360d76eca366abc3bc988dcfafceea6e869b84307c
 size 14180

scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:58147daee2921e5017eb63cf8d33546022b1bb37b22c55766c3cb5363ebb9fe1
+oid sha256:1ba6863dbb07ef7d843820e79752e6f6253dc743ad8e6f958a2039d8817b85b9
 size 1064

trainer_state.json

@@ -1,9 +1,9 @@
 {
-  "best_metric": 0.8944271909999159,
-  "best_model_checkpoint": "/home01/k157a01/Nguyen-Workdir/bitter/BitT5/output/BTP720_ckpt/sequence_selfies_biot5_plus_base/dropout_0.05/fold_10/checkpoint-1224",
-  "epoch": 34.0,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 50.0,
   "eval_steps": 500,
-  "global_step": 1224,
+  "global_step": 1800,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -22,10 +22,10 @@
       "eval_precision": 0.691358024691358,
       "eval_recall": 0.7777777777777778,
       "eval_roc_auc": 0.7881944444444445,
-      "eval_runtime": 1.8808,
-      "eval_samples_per_second": 76.564,
+      "eval_runtime": 1.7448,
+      "eval_samples_per_second": 82.531,
       "eval_specificity": 0.6527777777777778,
-      "eval_steps_per_second": 0.532,
+      "eval_steps_per_second": 0.573,
       "eval_tn": 47,
       "eval_tp": 56,
       "step": 36
@@ -44,10 +44,10 @@
       "eval_precision": 1.0,
       "eval_recall": 0.4583333333333333,
       "eval_roc_auc": 0.8784722222222222,
-      "eval_runtime": 1.582,
-      "eval_samples_per_second": 91.022,
+      "eval_runtime": 1.564,
+      "eval_samples_per_second": 92.072,
       "eval_specificity": 1.0,
-      "eval_steps_per_second": 0.632,
+      "eval_steps_per_second": 0.639,
       "eval_tn": 72,
       "eval_tp": 33,
       "step": 72
@@ -66,10 +66,10 @@
       "eval_precision": 0.94,
       "eval_recall": 0.6527777777777778,
       "eval_roc_auc": 0.8809799382716048,
-      "eval_runtime": 1.6146,
-      "eval_samples_per_second": 89.184,
+      "eval_runtime": 1.5621,
+      "eval_samples_per_second": 92.185,
       "eval_specificity": 0.9583333333333334,
-      "eval_steps_per_second": 0.619,
+      "eval_steps_per_second": 0.64,
       "eval_tn": 69,
       "eval_tp": 47,
       "step": 108
@@ -88,8 +88,8 @@
       "eval_precision": 0.9090909090909091,
       "eval_recall": 0.8333333333333334,
       "eval_roc_auc": 0.9261188271604939,
-      "eval_runtime": 1.5646,
-      "eval_samples_per_second": 92.038,
+      "eval_runtime": 1.5641,
+      "eval_samples_per_second": 92.065,
       "eval_specificity": 0.9166666666666666,
       "eval_steps_per_second": 0.639,
       "eval_tn": 66,
@@ -110,10 +110,10 @@
       "eval_precision": 0.90625,
       "eval_recall": 0.8055555555555556,
       "eval_roc_auc": 0.91820987654321,
-      "eval_runtime": 1.6154,
-      "eval_samples_per_second": 89.143,
+      "eval_runtime": 1.6068,
+      "eval_samples_per_second": 89.618,
       "eval_specificity": 0.9166666666666666,
-      "eval_steps_per_second": 0.619,
+      "eval_steps_per_second": 0.622,
       "eval_tn": 66,
       "eval_tp": 58,
       "step": 180
@@ -132,10 +132,10 @@
       "eval_precision": 0.9473684210526315,
       "eval_recall": 0.75,
       "eval_roc_auc": 0.9330632716049384,
-      "eval_runtime": 1.6201,
-      "eval_samples_per_second": 88.883,
+      "eval_runtime": 1.5703,
+      "eval_samples_per_second": 91.705,
       "eval_specificity": 0.9583333333333334,
-      "eval_steps_per_second": 0.617,
+      "eval_steps_per_second": 0.637,
       "eval_tn": 69,
       "eval_tp": 54,
       "step": 216
@@ -154,10 +154,10 @@
       "eval_precision": 0.8714285714285714,
       "eval_recall": 0.8472222222222222,
       "eval_roc_auc": 0.9386574074074074,
-      "eval_runtime": 1.6193,
-      "eval_samples_per_second": 88.925,
+      "eval_runtime": 1.5692,
+      "eval_samples_per_second": 91.766,
       "eval_specificity": 0.875,
-      "eval_steps_per_second": 0.618,
+      "eval_steps_per_second": 0.637,
       "eval_tn": 63,
       "eval_tp": 61,
       "step": 252
@@ -176,10 +176,10 @@
       "eval_precision": 0.9649122807017544,
       "eval_recall": 0.7638888888888888,
       "eval_roc_auc": 0.9415509259259259,
-      "eval_runtime": 1.5675,
-      "eval_samples_per_second": 91.866,
+      "eval_runtime": 1.564,
+      "eval_samples_per_second": 92.069,
       "eval_specificity": 0.9722222222222222,
-      "eval_steps_per_second": 0.638,
+      "eval_steps_per_second": 0.639,
       "eval_tn": 70,
       "eval_tp": 55,
       "step": 288
@@ -198,10 +198,10 @@
       "eval_precision": 1.0,
       "eval_recall": 0.7777777777777778,
       "eval_roc_auc": 0.9461805555555556,
-      "eval_runtime": 1.5673,
-      "eval_samples_per_second": 91.879,
+      "eval_runtime": 1.5731,
+      "eval_samples_per_second": 91.538,
       "eval_specificity": 1.0,
-      "eval_steps_per_second": 0.638,
+      "eval_steps_per_second": 0.636,
       "eval_tn": 72,
       "eval_tp": 56,
       "step": 324
@@ -220,10 +220,10 @@
       "eval_precision": 0.9666666666666667,
       "eval_recall": 0.8055555555555556,
       "eval_roc_auc": 0.9157021604938271,
-      "eval_runtime": 1.5842,
-      "eval_samples_per_second": 90.896,
+      "eval_runtime": 1.5699,
+      "eval_samples_per_second": 91.725,
       "eval_specificity": 0.9722222222222222,
-      "eval_steps_per_second": 0.631,
+      "eval_steps_per_second": 0.637,
       "eval_tn": 70,
       "eval_tp": 58,
       "step": 360
@@ -242,10 +242,10 @@
       "eval_precision": 0.9,
       "eval_recall": 0.875,
       "eval_roc_auc": 0.929398148148148,
-      "eval_runtime": 1.5719,
-      "eval_samples_per_second": 91.612,
+      "eval_runtime": 1.5762,
+      "eval_samples_per_second": 91.359,
       "eval_specificity": 0.9027777777777778,
-      "eval_steps_per_second": 0.636,
+      "eval_steps_per_second": 0.634,
       "eval_tn": 65,
       "eval_tp": 63,
       "step": 396
@@ -264,10 +264,10 @@
       "eval_precision": 0.921875,
       "eval_recall": 0.8194444444444444,
       "eval_roc_auc": 0.9430941358024691,
-      "eval_runtime": 1.6148,
-      "eval_samples_per_second": 89.175,
+      "eval_runtime": 1.5674,
+      "eval_samples_per_second": 91.874,
       "eval_specificity": 0.9305555555555556,
-      "eval_steps_per_second": 0.619,
+      "eval_steps_per_second": 0.638,
       "eval_tn": 67,
       "eval_tp": 59,
       "step": 432
@@ -286,10 +286,10 @@
       "eval_precision": 0.8857142857142857,
       "eval_recall": 0.8611111111111112,
       "eval_roc_auc": 0.9290123456790124,
-      "eval_runtime": 1.6227,
-      "eval_samples_per_second": 88.742,
+      "eval_runtime": 1.5785,
+      "eval_samples_per_second": 91.228,
       "eval_specificity": 0.8888888888888888,
-      "eval_steps_per_second": 0.616,
+      "eval_steps_per_second": 0.634,
       "eval_tn": 64,
       "eval_tp": 62,
       "step": 468
@@ -315,10 +315,10 @@
       "eval_precision": 0.8985507246376812,
       "eval_recall": 0.8611111111111112,
       "eval_roc_auc": 0.9479166666666667,
-      "eval_runtime": 1.5643,
-      "eval_samples_per_second": 92.057,
+      "eval_runtime": 1.5778,
+      "eval_samples_per_second": 91.264,
       "eval_specificity": 0.9027777777777778,
-      "eval_steps_per_second": 0.639,
+      "eval_steps_per_second": 0.634,
       "eval_tn": 65,
       "eval_tp": 62,
       "step": 504
@@ -337,10 +337,10 @@
       "eval_precision": 0.9538461538461539,
       "eval_recall": 0.8611111111111112,
       "eval_roc_auc": 0.9479166666666666,
-      "eval_runtime": 1.5675,
-      "eval_samples_per_second": 91.867,
+      "eval_runtime": 1.5734,
+      "eval_samples_per_second": 91.52,
       "eval_specificity": 0.9583333333333334,
-      "eval_steps_per_second": 0.638,
+      "eval_steps_per_second": 0.636,
       "eval_tn": 69,
       "eval_tp": 62,
       "step": 540
@@ -359,10 +359,10 @@
       "eval_precision": 1.0,
       "eval_recall": 0.7777777777777778,
       "eval_roc_auc": 0.9598765432098766,
-      "eval_runtime": 1.5706,
-      "eval_samples_per_second": 91.682,
+      "eval_runtime": 1.5683,
+      "eval_samples_per_second": 91.821,
       "eval_specificity": 1.0,
-      "eval_steps_per_second": 0.637,
+      "eval_steps_per_second": 0.638,
       "eval_tn": 72,
       "eval_tp": 56,
       "step": 576
@@ -381,10 +381,10 @@
       "eval_precision": 0.9833333333333333,
       "eval_recall": 0.8194444444444444,
       "eval_roc_auc": 0.9525462962962963,
-      "eval_runtime": 1.6087,
-      "eval_samples_per_second": 89.513,
+      "eval_runtime": 1.5821,
+      "eval_samples_per_second": 91.016,
       "eval_specificity": 0.9861111111111112,
-      "eval_steps_per_second": 0.622,
+      "eval_steps_per_second": 0.632,
       "eval_tn": 71,
       "eval_tp": 59,
       "step": 612
@@ -403,10 +403,10 @@
       "eval_precision": 0.9142857142857143,
       "eval_recall": 0.8888888888888888,
       "eval_roc_auc": 0.957175925925926,
-      "eval_runtime": 1.6236,
-      "eval_samples_per_second": 88.694,
+      "eval_runtime": 1.6038,
+      "eval_samples_per_second": 89.787,
       "eval_specificity": 0.9166666666666666,
-      "eval_steps_per_second": 0.616,
+      "eval_steps_per_second": 0.624,
       "eval_tn": 66,
       "eval_tp": 64,
       "step": 648
@@ -425,10 +425,10 @@
       "eval_precision": 0.9836065573770492,
       "eval_recall": 0.8333333333333334,
       "eval_roc_auc": 0.961033950617284,
-      "eval_runtime": 1.5708,
-      "eval_samples_per_second": 91.673,
+      "eval_runtime": 1.5679,
+      "eval_samples_per_second": 91.842,
       "eval_specificity": 0.9861111111111112,
-      "eval_steps_per_second": 0.637,
+      "eval_steps_per_second": 0.638,
       "eval_tn": 71,
       "eval_tp": 60,
       "step": 684
@@ -447,10 +447,10 @@
       "eval_precision": 0.9523809523809523,
       "eval_recall": 0.8333333333333334,
       "eval_roc_auc": 0.964891975308642,
-      "eval_runtime": 1.573,
-      "eval_samples_per_second": 91.546,
+      "eval_runtime": 1.5824,
+      "eval_samples_per_second": 91.003,
       "eval_specificity": 0.9583333333333334,
-      "eval_steps_per_second": 0.636,
+      "eval_steps_per_second": 0.632,
       "eval_tn": 69,
       "eval_tp": 60,
       "step": 720
@@ -469,10 +469,10 @@
       "eval_precision": 0.8481012658227848,
       "eval_recall": 0.9305555555555556,
       "eval_roc_auc": 0.9456018518518517,
-      "eval_runtime": 1.5869,
-      "eval_samples_per_second": 90.743,
+      "eval_runtime": 1.5692,
+      "eval_samples_per_second": 91.769,
       "eval_specificity": 0.8333333333333334,
-      "eval_steps_per_second": 0.63,
+      "eval_steps_per_second": 0.637,
       "eval_tn": 60,
       "eval_tp": 67,
       "step": 756
@@ -491,10 +491,10 @@
       "eval_precision": 0.9,
       "eval_recall": 0.875,
       "eval_roc_auc": 0.9533179012345679,
-      "eval_runtime": 1.5922,
-      "eval_samples_per_second": 90.442,
+      "eval_runtime": 1.5845,
+      "eval_samples_per_second": 90.883,
       "eval_specificity": 0.9027777777777778,
-      "eval_steps_per_second": 0.628,
+      "eval_steps_per_second": 0.631,
       "eval_tn": 65,
       "eval_tp": 63,
       "step": 792
@@ -513,10 +513,10 @@
       "eval_precision": 0.9264705882352942,
       "eval_recall": 0.875,
       "eval_roc_auc": 0.9672067901234568,
-      "eval_runtime": 1.5827,
-      "eval_samples_per_second": 90.986,
+      "eval_runtime": 1.5931,
+      "eval_samples_per_second": 90.391,
       "eval_specificity": 0.9305555555555556,
-      "eval_steps_per_second": 0.632,
+      "eval_steps_per_second": 0.628,
       "eval_tn": 67,
       "eval_tp": 63,
       "step": 828
@@ -535,10 +535,10 @@
       "eval_precision": 0.96875,
       "eval_recall": 0.8611111111111112,
       "eval_roc_auc": 0.9457947530864197,
-      "eval_runtime": 1.6426,
-      "eval_samples_per_second": 87.666,
+      "eval_runtime": 1.5739,
+      "eval_samples_per_second": 91.49,
       "eval_specificity": 0.9722222222222222,
-      "eval_steps_per_second": 0.609,
+      "eval_steps_per_second": 0.635,
       "eval_tn": 70,
       "eval_tp": 62,
       "step": 864
@@ -557,10 +557,10 @@
       "eval_precision": 0.9538461538461539,
       "eval_recall": 0.8611111111111112,
       "eval_roc_auc": 0.9637345679012346,
-      "eval_runtime": 1.5896,
-      "eval_samples_per_second": 90.591,
+      "eval_runtime": 1.5981,
+      "eval_samples_per_second": 90.107,
       "eval_specificity": 0.9583333333333334,
-      "eval_steps_per_second": 0.629,
+      "eval_steps_per_second": 0.626,
       "eval_tn": 69,
       "eval_tp": 62,
       "step": 900
@@ -579,10 +579,10 @@
       "eval_precision": 0.9545454545454546,
       "eval_recall": 0.875,
       "eval_roc_auc": 0.9591049382716049,
-      "eval_runtime": 1.5858,
-      "eval_samples_per_second": 90.805,
+      "eval_runtime": 1.5707,
+      "eval_samples_per_second": 91.676,
       "eval_specificity": 0.9583333333333334,
-      "eval_steps_per_second": 0.631,
+      "eval_steps_per_second": 0.637,
       "eval_tn": 69,
       "eval_tp": 63,
       "step": 936
@@ -601,10 +601,10 @@
       "eval_precision": 0.9682539682539683,
       "eval_recall": 0.8472222222222222,
       "eval_roc_auc": 0.961033950617284,
-      "eval_runtime": 1.6198,
-      "eval_samples_per_second": 88.898,
+      "eval_runtime": 1.5701,
+      "eval_samples_per_second": 91.714,
       "eval_specificity": 0.9722222222222222,
-      "eval_steps_per_second": 0.617,
+      "eval_steps_per_second": 0.637,
       "eval_tn": 70,
       "eval_tp": 61,
       "step": 972
@@ -630,10 +630,10 @@
       "eval_precision": 0.9841269841269841,
       "eval_recall": 0.8611111111111112,
       "eval_roc_auc": 0.9716435185185186,
-      "eval_runtime": 1.585,
-      "eval_samples_per_second": 90.851,
+      "eval_runtime": 1.5758,
+      "eval_samples_per_second": 91.382,
       "eval_specificity": 0.9861111111111112,
-      "eval_steps_per_second": 0.631,
+      "eval_steps_per_second": 0.635,
       "eval_tn": 71,
       "eval_tp": 62,
       "step": 1008
@@ -652,10 +652,10 @@
       "eval_precision": 0.9295774647887324,
       "eval_recall": 0.9166666666666666,
       "eval_roc_auc": 0.9670138888888891,
-      "eval_runtime": 1.6358,
-      "eval_samples_per_second": 88.03,
+      "eval_runtime": 1.609,
+      "eval_samples_per_second": 89.498,
       "eval_specificity": 0.9305555555555556,
-      "eval_steps_per_second": 0.611,
+      "eval_steps_per_second": 0.622,
       "eval_tn": 67,
       "eval_tp": 66,
       "step": 1044
@@ -674,8 +674,8 @@
       "eval_precision": 0.9552238805970149,
       "eval_recall": 0.8888888888888888,
       "eval_roc_auc": 0.9658564814814816,
-      "eval_runtime": 1.5788,
-      "eval_samples_per_second": 91.206,
+      "eval_runtime": 1.5802,
+      "eval_samples_per_second": 91.126,
       "eval_specificity": 0.9583333333333334,
       "eval_steps_per_second": 0.633,
       "eval_tn": 69,
@@ -696,10 +696,10 @@
       "eval_precision": 0.9538461538461539,
       "eval_recall": 0.8611111111111112,
       "eval_roc_auc": 0.9635416666666667,
-      "eval_runtime": 1.6079,
-      "eval_samples_per_second": 89.56,
+      "eval_runtime": 1.606,
+      "eval_samples_per_second": 89.664,
       "eval_specificity": 0.9583333333333334,
-      "eval_steps_per_second": 0.622,
+      "eval_steps_per_second": 0.623,
       "eval_tn": 69,
       "eval_tp": 62,
       "step": 1116
@@ -718,10 +718,10 @@
       "eval_precision": 1.0,
       "eval_recall": 0.8472222222222222,
       "eval_roc_auc": 0.9363425925925926,
-      "eval_runtime": 1.5899,
-      "eval_samples_per_second": 90.574,
+      "eval_runtime": 1.6096,
+      "eval_samples_per_second": 89.462,
       "eval_specificity": 1.0,
-      "eval_steps_per_second": 0.629,
+      "eval_steps_per_second": 0.621,
       "eval_tn": 72,
       "eval_tp": 61,
       "step": 1152
@@ -740,10 +740,10 @@
       "eval_precision": 0.9696969696969697,
       "eval_recall": 0.8888888888888888,
       "eval_roc_auc": 0.9567901234567903,
-      "eval_runtime": 1.584,
-      "eval_samples_per_second": 90.91,
+      "eval_runtime": 1.6127,
+      "eval_samples_per_second": 89.292,
       "eval_specificity": 0.9722222222222222,
-      "eval_steps_per_second": 0.631,
+      "eval_steps_per_second": 0.62,
       "eval_tn": 70,
       "eval_tp": 64,
       "step": 1188
@@ -762,13 +762,372 @@
       "eval_precision": 1.0,
       "eval_recall": 0.8888888888888888,
       "eval_roc_auc": 0.9633487654320988,
-      "eval_runtime": 1.59,
-      "eval_samples_per_second": 90.567,
+      "eval_runtime": 1.574,
+      "eval_samples_per_second": 91.488,
       "eval_specificity": 1.0,
-      "eval_steps_per_second": 0.629,
+      "eval_steps_per_second": 0.635,
       "eval_tn": 72,
       "eval_tp": 64,
       "step": 1224
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.9166666666666666,
+      "eval_auc": 0.9575617283950617,
+      "eval_balanced_acc": 0.9166666666666667,
+      "eval_f1": 0.9090909090909091,
+      "eval_fn": 12,
+      "eval_fp": 0,
+      "eval_gmean": 0.9128709291752769,
+      "eval_loss": 0.23990610241889954,
+      "eval_matthews_correlation": 0.8451542547285166,
+      "eval_precision": 1.0,
+      "eval_recall": 0.8333333333333334,
+      "eval_roc_auc": 0.9575617283950617,
+      "eval_runtime": 1.5748,
+      "eval_samples_per_second": 91.441,
+      "eval_specificity": 1.0,
+      "eval_steps_per_second": 0.635,
+      "eval_tn": 72,
+      "eval_tp": 60,
+      "step": 1260
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.9236111111111112,
+      "eval_auc": 0.9538966049382716,
+      "eval_balanced_acc": 0.9236111111111112,
+      "eval_f1": 0.9185185185185185,
+      "eval_fn": 10,
+      "eval_fp": 1,
+      "eval_gmean": 0.9214940230776874,
+      "eval_loss": 0.23596133291721344,
+      "eval_matthews_correlation": 0.8539197353171429,
+      "eval_precision": 0.9841269841269841,
+      "eval_recall": 0.8611111111111112,
+      "eval_roc_auc": 0.9538966049382716,
+      "eval_runtime": 1.5725,
+      "eval_samples_per_second": 91.573,
+      "eval_specificity": 0.9861111111111112,
+      "eval_steps_per_second": 0.636,
+      "eval_tn": 71,
+      "eval_tp": 62,
+      "step": 1296
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy": 0.9305555555555556,
+      "eval_auc": 0.9683641975308642,
+      "eval_balanced_acc": 0.9305555555555556,
+      "eval_f1": 0.927536231884058,
+      "eval_fn": 8,
+      "eval_fp": 2,
+      "eval_gmean": 0.9296222517045284,
+      "eval_loss": 0.24645011126995087,
+      "eval_matthews_correlation": 0.8641167436739938,
+      "eval_precision": 0.9696969696969697,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9683641975308642,
+      "eval_runtime": 1.5741,
+      "eval_samples_per_second": 91.483,
+      "eval_specificity": 0.9722222222222222,
+      "eval_steps_per_second": 0.635,
+      "eval_tn": 70,
+      "eval_tp": 64,
+      "step": 1332
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.9375,
+      "eval_auc": 0.9710648148148149,
+      "eval_balanced_acc": 0.9375,
+      "eval_f1": 0.9343065693430657,
+      "eval_fn": 8,
+      "eval_fp": 1,
+      "eval_gmean": 0.9362388636862621,
+      "eval_loss": 0.20392011106014252,
+      "eval_matthews_correlation": 0.8791648688396009,
+      "eval_precision": 0.9846153846153847,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9710648148148149,
+      "eval_runtime": 1.5784,
+      "eval_samples_per_second": 91.233,
+      "eval_specificity": 0.9861111111111112,
+      "eval_steps_per_second": 0.634,
+      "eval_tn": 71,
+      "eval_tp": 64,
+      "step": 1368
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.9375,
+      "eval_auc": 0.9737654320987655,
+      "eval_balanced_acc": 0.9375,
+      "eval_f1": 0.9343065693430657,
+      "eval_fn": 8,
+      "eval_fp": 1,
+      "eval_gmean": 0.9362388636862621,
+      "eval_loss": 0.1912021040916443,
+      "eval_matthews_correlation": 0.8791648688396009,
+      "eval_precision": 0.9846153846153847,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9737654320987655,
+      "eval_runtime": 1.5696,
+      "eval_samples_per_second": 91.745,
+      "eval_specificity": 0.9861111111111112,
+      "eval_steps_per_second": 0.637,
+      "eval_tn": 71,
+      "eval_tp": 64,
+      "step": 1404
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.9444444444444444,
+      "eval_auc": 0.970679012345679,
+      "eval_balanced_acc": 0.9444444444444444,
+      "eval_f1": 0.9411764705882353,
+      "eval_fn": 8,
+      "eval_fp": 0,
+      "eval_gmean": 0.9428090415820634,
+      "eval_loss": 0.20200136303901672,
+      "eval_matthews_correlation": 0.8944271909999159,
+      "eval_precision": 1.0,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.970679012345679,
+      "eval_runtime": 1.6039,
+      "eval_samples_per_second": 89.78,
+      "eval_specificity": 1.0,
+      "eval_steps_per_second": 0.623,
+      "eval_tn": 72,
+      "eval_tp": 64,
+      "step": 1440
+    },
+    {
+      "epoch": 41.0,
+      "eval_accuracy": 0.9444444444444444,
+      "eval_auc": 0.9758873456790124,
+      "eval_balanced_acc": 0.9444444444444444,
+      "eval_f1": 0.9411764705882353,
+      "eval_fn": 8,
+      "eval_fp": 0,
+      "eval_gmean": 0.9428090415820634,
+      "eval_loss": 0.1961589753627777,
+      "eval_matthews_correlation": 0.8944271909999159,
+      "eval_precision": 1.0,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9758873456790124,
+      "eval_runtime": 1.5711,
+      "eval_samples_per_second": 91.653,
+      "eval_specificity": 1.0,
+      "eval_steps_per_second": 0.636,
+      "eval_tn": 72,
+      "eval_tp": 64,
+      "step": 1476
+    },
+    {
+      "epoch": 41.666666666666664,
+      "grad_norm": 0.13324905931949615,
+      "learning_rate": 8.771929824561403e-05,
+      "loss": 0.1136,
+      "step": 1500
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy": 0.9444444444444444,
+      "eval_auc": 0.9751157407407407,
+      "eval_balanced_acc": 0.9444444444444444,
+      "eval_f1": 0.9411764705882353,
+      "eval_fn": 8,
+      "eval_fp": 0,
+      "eval_gmean": 0.9428090415820634,
+      "eval_loss": 0.1993894726037979,
+      "eval_matthews_correlation": 0.8944271909999159,
+      "eval_precision": 1.0,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9751157407407407,
+      "eval_runtime": 1.5781,
+      "eval_samples_per_second": 91.25,
+      "eval_specificity": 1.0,
+      "eval_steps_per_second": 0.634,
+      "eval_tn": 72,
+      "eval_tp": 64,
+      "step": 1512
+    },
+    {
+      "epoch": 43.0,
+      "eval_accuracy": 0.9375,
+      "eval_auc": 0.9810956790123457,
+      "eval_balanced_acc": 0.9375,
+      "eval_f1": 0.9343065693430657,
+      "eval_fn": 8,
+      "eval_fp": 1,
+      "eval_gmean": 0.9362388636862621,
+      "eval_loss": 0.1959426999092102,
+      "eval_matthews_correlation": 0.8791648688396009,
+      "eval_precision": 0.9846153846153847,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9810956790123457,
+      "eval_runtime": 1.5846,
+      "eval_samples_per_second": 90.874,
+      "eval_specificity": 0.9861111111111112,
+      "eval_steps_per_second": 0.631,
+      "eval_tn": 71,
+      "eval_tp": 64,
+      "step": 1548
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.9375,
+      "eval_auc": 0.9824459876543209,
+      "eval_balanced_acc": 0.9375,
+      "eval_f1": 0.9343065693430657,
+      "eval_fn": 8,
+      "eval_fp": 1,
+      "eval_gmean": 0.9362388636862621,
+      "eval_loss": 0.19525253772735596,
+      "eval_matthews_correlation": 0.8791648688396009,
+      "eval_precision": 0.9846153846153847,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9824459876543209,
+      "eval_runtime": 1.5729,
+      "eval_samples_per_second": 91.553,
+      "eval_specificity": 0.9861111111111112,
+      "eval_steps_per_second": 0.636,
+      "eval_tn": 71,
+      "eval_tp": 64,
+      "step": 1584
+    },
+    {
+      "epoch": 45.0,
+      "eval_accuracy": 0.9444444444444444,
+      "eval_auc": 0.9836033950617284,
+      "eval_balanced_acc": 0.9444444444444444,
+      "eval_f1": 0.9411764705882353,
+      "eval_fn": 8,
+      "eval_fp": 0,
+      "eval_gmean": 0.9428090415820634,
+      "eval_loss": 0.19979047775268555,
+      "eval_matthews_correlation": 0.8944271909999159,
+      "eval_precision": 1.0,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9836033950617284,
+      "eval_runtime": 1.5771,
+      "eval_samples_per_second": 91.309,
+      "eval_specificity": 1.0,
+      "eval_steps_per_second": 0.634,
+      "eval_tn": 72,
+      "eval_tp": 64,
+      "step": 1620
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy": 0.9375,
+      "eval_auc": 0.9818672839506173,
+      "eval_balanced_acc": 0.9375,
+      "eval_f1": 0.9343065693430657,
+      "eval_fn": 8,
+      "eval_fp": 1,
+      "eval_gmean": 0.9362388636862621,
+      "eval_loss": 0.19916905462741852,
+      "eval_matthews_correlation": 0.8791648688396009,
+      "eval_precision": 0.9846153846153847,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9818672839506173,
+      "eval_runtime": 1.5731,
+      "eval_samples_per_second": 91.542,
+      "eval_specificity": 0.9861111111111112,
+      "eval_steps_per_second": 0.636,
+      "eval_tn": 71,
+      "eval_tp": 64,
+      "step": 1656
+    },
+    {
+      "epoch": 47.0,
+      "eval_accuracy": 0.9444444444444444,
+      "eval_auc": 0.9820601851851852,
+      "eval_balanced_acc": 0.9444444444444444,
+      "eval_f1": 0.9411764705882353,
+      "eval_fn": 8,
+      "eval_fp": 0,
+      "eval_gmean": 0.9428090415820634,
+      "eval_loss": 0.20321370661258698,
+      "eval_matthews_correlation": 0.8944271909999159,
+      "eval_precision": 1.0,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9820601851851852,
+      "eval_runtime": 1.5764,
+      "eval_samples_per_second": 91.349,
+      "eval_specificity": 1.0,
+      "eval_steps_per_second": 0.634,
+      "eval_tn": 72,
+      "eval_tp": 64,
+      "step": 1692
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.9444444444444444,
+      "eval_auc": 0.9809027777777778,
+      "eval_balanced_acc": 0.9444444444444444,
+      "eval_f1": 0.9411764705882353,
+      "eval_fn": 8,
+      "eval_fp": 0,
+      "eval_gmean": 0.9428090415820634,
+      "eval_loss": 0.2023700326681137,
+      "eval_matthews_correlation": 0.8944271909999159,
+      "eval_precision": 1.0,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9809027777777778,
+      "eval_runtime": 1.5713,
+      "eval_samples_per_second": 91.644,
+      "eval_specificity": 1.0,
+      "eval_steps_per_second": 0.636,
+      "eval_tn": 72,
+      "eval_tp": 64,
+      "step": 1728
+    },
+    {
+      "epoch": 49.0,
+      "eval_accuracy": 0.9444444444444444,
+      "eval_auc": 0.9803240740740742,
+      "eval_balanced_acc": 0.9444444444444444,
+      "eval_f1": 0.9411764705882353,
+      "eval_fn": 8,
+      "eval_fp": 0,
+      "eval_gmean": 0.9428090415820634,
+      "eval_loss": 0.20228758454322815,
+      "eval_matthews_correlation": 0.8944271909999159,
+      "eval_precision": 1.0,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9803240740740742,
+      "eval_runtime": 1.5796,
+      "eval_samples_per_second": 91.16,
+      "eval_specificity": 1.0,
+      "eval_steps_per_second": 0.633,
+      "eval_tn": 72,
+      "eval_tp": 64,
+      "step": 1764
+    },
+    {
+      "epoch": 50.0,
+      "eval_accuracy": 0.9375,
+      "eval_auc": 0.9797453703703703,
+      "eval_balanced_acc": 0.9375,
+      "eval_f1": 0.9343065693430657,
+      "eval_fn": 8,
+      "eval_fp": 1,
+      "eval_gmean": 0.9362388636862621,
+      "eval_loss": 0.20149865746498108,
+      "eval_matthews_correlation": 0.8791648688396009,
+      "eval_precision": 0.9846153846153847,
+      "eval_recall": 0.8888888888888888,
+      "eval_roc_auc": 0.9797453703703703,
+      "eval_runtime": 1.5697,
+      "eval_samples_per_second": 91.74,
+      "eval_specificity": 0.9861111111111112,
+      "eval_steps_per_second": 0.637,
+      "eval_tn": 71,
+      "eval_tp": 64,
+      "step": 1800
     }
   ],
   "logging_steps": 500,
@@ -783,12 +1142,12 @@
         "should_evaluate": false,
         "should_log": false,
         "should_save": true,
-        "should_training_stop": false
+        "should_training_stop": true
       },
       "attributes": {}
     }
   },
-  "total_flos": 5149923155232576.0,
+  "total_flos": 7568362736557056.0,
   "train_batch_size": 16,
   "trial_name": null,
   "trial_params": null

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:32ad303d16c8436f4fdd81f66e5e9d19243ff49251c50c8244a2933921f828a2
-size 5560
+oid sha256:16c25ba32b3d7a2dc23d0eb48efdc875efa4bbc703183fc9f9055e824f0bd152
+size 5496