Upload 12 files
Browse files- model.safetensors +1 -1
- optimizer.pt +1 -1
- rng_state.pth +1 -1
- scheduler.pt +1 -1
- trainer_state.json +353 -126
- training_args.bin +2 -2
model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 901339568
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:131da3876aaca1897703a33f04d142f48c945f2450e189fec4a0316b6fce6329
|
3 |
size 901339568
|
optimizer.pt
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 227854586
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0fe10f5b77323a7a170ea2031248a93765886fb7c376f51845ee7278381151c1
|
3 |
size 227854586
|
rng_state.pth
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 14180
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4f8e8e8f9d8a69223288f91acadfe178658f954371c9c5e5acf05816f8c72e51
|
3 |
size 14180
|
scheduler.pt
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1064
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e773a85e1bc36c2275deb763e1f7c3e5cd88423162895f71855030dbf616e0bb
|
3 |
size 1064
|
trainer_state.json
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
{
|
2 |
-
"best_metric":
|
3 |
-
"best_model_checkpoint":
|
4 |
-
"epoch":
|
5 |
"eval_steps": 500,
|
6 |
-
"global_step":
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
@@ -22,10 +22,10 @@
|
|
22 |
"eval_precision": 0.0,
|
23 |
"eval_recall": 0.0,
|
24 |
"eval_roc_auc": 0.630615234375,
|
25 |
-
"eval_runtime":
|
26 |
-
"eval_samples_per_second":
|
27 |
"eval_specificity": 1.0,
|
28 |
-
"eval_steps_per_second": 0.
|
29 |
"eval_tn": 64,
|
30 |
"eval_tp": 0,
|
31 |
"step": 64
|
@@ -44,10 +44,10 @@
|
|
44 |
"eval_precision": 0.6276595744680851,
|
45 |
"eval_recall": 0.921875,
|
46 |
"eval_roc_auc": 0.797119140625,
|
47 |
-
"eval_runtime":
|
48 |
-
"eval_samples_per_second":
|
49 |
"eval_specificity": 0.453125,
|
50 |
-
"eval_steps_per_second": 0.
|
51 |
"eval_tn": 29,
|
52 |
"eval_tp": 59,
|
53 |
"step": 128
|
@@ -66,10 +66,10 @@
|
|
66 |
"eval_precision": 0.8518518518518519,
|
67 |
"eval_recall": 0.71875,
|
68 |
"eval_roc_auc": 0.891845703125,
|
69 |
-
"eval_runtime":
|
70 |
-
"eval_samples_per_second":
|
71 |
"eval_specificity": 0.875,
|
72 |
-
"eval_steps_per_second": 0.
|
73 |
"eval_tn": 56,
|
74 |
"eval_tp": 46,
|
75 |
"step": 192
|
@@ -88,10 +88,10 @@
|
|
88 |
"eval_precision": 0.9090909090909091,
|
89 |
"eval_recall": 0.78125,
|
90 |
"eval_roc_auc": 0.934814453125,
|
91 |
-
"eval_runtime":
|
92 |
-
"eval_samples_per_second":
|
93 |
"eval_specificity": 0.921875,
|
94 |
-
"eval_steps_per_second": 0.
|
95 |
"eval_tn": 59,
|
96 |
"eval_tp": 50,
|
97 |
"step": 256
|
@@ -110,10 +110,10 @@
|
|
110 |
"eval_precision": 0.8507462686567164,
|
111 |
"eval_recall": 0.890625,
|
112 |
"eval_roc_auc": 0.938232421875,
|
113 |
-
"eval_runtime":
|
114 |
-
"eval_samples_per_second":
|
115 |
"eval_specificity": 0.84375,
|
116 |
-
"eval_steps_per_second": 0.
|
117 |
"eval_tn": 54,
|
118 |
"eval_tp": 57,
|
119 |
"step": 320
|
@@ -132,10 +132,10 @@
|
|
132 |
"eval_precision": 0.8378378378378378,
|
133 |
"eval_recall": 0.96875,
|
134 |
"eval_roc_auc": 0.951171875,
|
135 |
-
"eval_runtime":
|
136 |
-
"eval_samples_per_second":
|
137 |
"eval_specificity": 0.8125,
|
138 |
-
"eval_steps_per_second": 0.
|
139 |
"eval_tn": 52,
|
140 |
"eval_tp": 62,
|
141 |
"step": 384
|
@@ -154,10 +154,10 @@
|
|
154 |
"eval_precision": 0.921875,
|
155 |
"eval_recall": 0.921875,
|
156 |
"eval_roc_auc": 0.96142578125,
|
157 |
-
"eval_runtime":
|
158 |
-
"eval_samples_per_second":
|
159 |
"eval_specificity": 0.921875,
|
160 |
-
"eval_steps_per_second": 0.
|
161 |
"eval_tn": 59,
|
162 |
"eval_tp": 59,
|
163 |
"step": 448
|
@@ -183,10 +183,10 @@
|
|
183 |
"eval_precision": 0.8076923076923077,
|
184 |
"eval_recall": 0.984375,
|
185 |
"eval_roc_auc": 0.953369140625,
|
186 |
-
"eval_runtime":
|
187 |
-
"eval_samples_per_second":
|
188 |
"eval_specificity": 0.765625,
|
189 |
-
"eval_steps_per_second": 0.
|
190 |
"eval_tn": 49,
|
191 |
"eval_tp": 63,
|
192 |
"step": 512
|
@@ -205,10 +205,10 @@
|
|
205 |
"eval_precision": 0.8181818181818182,
|
206 |
"eval_recall": 0.984375,
|
207 |
"eval_roc_auc": 0.95751953125,
|
208 |
-
"eval_runtime":
|
209 |
-
"eval_samples_per_second":
|
210 |
"eval_specificity": 0.78125,
|
211 |
-
"eval_steps_per_second": 0.
|
212 |
"eval_tn": 50,
|
213 |
"eval_tp": 63,
|
214 |
"step": 576
|
@@ -227,10 +227,10 @@
|
|
227 |
"eval_precision": 0.9322033898305084,
|
228 |
"eval_recall": 0.859375,
|
229 |
"eval_roc_auc": 0.9658203125,
|
230 |
-
"eval_runtime":
|
231 |
-
"eval_samples_per_second":
|
232 |
"eval_specificity": 0.9375,
|
233 |
-
"eval_steps_per_second": 0.
|
234 |
"eval_tn": 60,
|
235 |
"eval_tp": 55,
|
236 |
"step": 640
|
@@ -249,10 +249,10 @@
|
|
249 |
"eval_precision": 0.8923076923076924,
|
250 |
"eval_recall": 0.90625,
|
251 |
"eval_roc_auc": 0.959716796875,
|
252 |
-
"eval_runtime":
|
253 |
-
"eval_samples_per_second":
|
254 |
"eval_specificity": 0.890625,
|
255 |
-
"eval_steps_per_second": 0.
|
256 |
"eval_tn": 57,
|
257 |
"eval_tp": 58,
|
258 |
"step": 704
|
@@ -271,10 +271,10 @@
|
|
271 |
"eval_precision": 0.9333333333333333,
|
272 |
"eval_recall": 0.875,
|
273 |
"eval_roc_auc": 0.96240234375,
|
274 |
-
"eval_runtime":
|
275 |
-
"eval_samples_per_second":
|
276 |
"eval_specificity": 0.9375,
|
277 |
-
"eval_steps_per_second": 0.
|
278 |
"eval_tn": 60,
|
279 |
"eval_tp": 56,
|
280 |
"step": 768
|
@@ -293,10 +293,10 @@
|
|
293 |
"eval_precision": 0.9655172413793104,
|
294 |
"eval_recall": 0.875,
|
295 |
"eval_roc_auc": 0.963134765625,
|
296 |
-
"eval_runtime":
|
297 |
-
"eval_samples_per_second":
|
298 |
"eval_specificity": 0.96875,
|
299 |
-
"eval_steps_per_second": 0.
|
300 |
"eval_tn": 62,
|
301 |
"eval_tp": 56,
|
302 |
"step": 832
|
@@ -315,10 +315,10 @@
|
|
315 |
"eval_precision": 0.9180327868852459,
|
316 |
"eval_recall": 0.875,
|
317 |
"eval_roc_auc": 0.962158203125,
|
318 |
-
"eval_runtime":
|
319 |
-
"eval_samples_per_second":
|
320 |
"eval_specificity": 0.921875,
|
321 |
-
"eval_steps_per_second": 0.
|
322 |
"eval_tn": 59,
|
323 |
"eval_tp": 56,
|
324 |
"step": 896
|
@@ -337,10 +337,10 @@
|
|
337 |
"eval_precision": 0.9180327868852459,
|
338 |
"eval_recall": 0.875,
|
339 |
"eval_roc_auc": 0.947509765625,
|
340 |
-
"eval_runtime":
|
341 |
-
"eval_samples_per_second":
|
342 |
"eval_specificity": 0.921875,
|
343 |
-
"eval_steps_per_second": 0.
|
344 |
"eval_tn": 59,
|
345 |
"eval_tp": 56,
|
346 |
"step": 960
|
@@ -366,10 +366,10 @@
|
|
366 |
"eval_precision": 0.9206349206349206,
|
367 |
"eval_recall": 0.90625,
|
368 |
"eval_roc_auc": 0.94677734375,
|
369 |
-
"eval_runtime":
|
370 |
-
"eval_samples_per_second":
|
371 |
"eval_specificity": 0.921875,
|
372 |
-
"eval_steps_per_second": 0.
|
373 |
"eval_tn": 59,
|
374 |
"eval_tp": 58,
|
375 |
"step": 1024
|
@@ -388,10 +388,10 @@
|
|
388 |
"eval_precision": 0.9322033898305084,
|
389 |
"eval_recall": 0.859375,
|
390 |
"eval_roc_auc": 0.96875,
|
391 |
-
"eval_runtime":
|
392 |
-
"eval_samples_per_second":
|
393 |
"eval_specificity": 0.9375,
|
394 |
-
"eval_steps_per_second": 0.
|
395 |
"eval_tn": 60,
|
396 |
"eval_tp": 55,
|
397 |
"step": 1088
|
@@ -410,10 +410,10 @@
|
|
410 |
"eval_precision": 0.9180327868852459,
|
411 |
"eval_recall": 0.875,
|
412 |
"eval_roc_auc": 0.95947265625,
|
413 |
-
"eval_runtime":
|
414 |
-
"eval_samples_per_second":
|
415 |
"eval_specificity": 0.921875,
|
416 |
-
"eval_steps_per_second": 0.
|
417 |
"eval_tn": 59,
|
418 |
"eval_tp": 56,
|
419 |
"step": 1152
|
@@ -432,10 +432,10 @@
|
|
432 |
"eval_precision": 0.9230769230769231,
|
433 |
"eval_recall": 0.9375,
|
434 |
"eval_roc_auc": 0.958740234375,
|
435 |
-
"eval_runtime":
|
436 |
-
"eval_samples_per_second":
|
437 |
"eval_specificity": 0.921875,
|
438 |
-
"eval_steps_per_second": 0.
|
439 |
"eval_tn": 59,
|
440 |
"eval_tp": 60,
|
441 |
"step": 1216
|
@@ -454,10 +454,10 @@
|
|
454 |
"eval_precision": 0.9104477611940298,
|
455 |
"eval_recall": 0.953125,
|
456 |
"eval_roc_auc": 0.951416015625,
|
457 |
-
"eval_runtime":
|
458 |
-
"eval_samples_per_second":
|
459 |
"eval_specificity": 0.90625,
|
460 |
-
"eval_steps_per_second": 0.
|
461 |
"eval_tn": 58,
|
462 |
"eval_tp": 61,
|
463 |
"step": 1280
|
@@ -476,10 +476,10 @@
|
|
476 |
"eval_precision": 0.9491525423728814,
|
477 |
"eval_recall": 0.875,
|
478 |
"eval_roc_auc": 0.95654296875,
|
479 |
-
"eval_runtime":
|
480 |
-
"eval_samples_per_second":
|
481 |
"eval_specificity": 0.953125,
|
482 |
-
"eval_steps_per_second": 0.
|
483 |
"eval_tn": 61,
|
484 |
"eval_tp": 56,
|
485 |
"step": 1344
|
@@ -498,10 +498,10 @@
|
|
498 |
"eval_precision": 0.9104477611940298,
|
499 |
"eval_recall": 0.953125,
|
500 |
"eval_roc_auc": 0.95703125,
|
501 |
-
"eval_runtime":
|
502 |
-
"eval_samples_per_second":
|
503 |
"eval_specificity": 0.90625,
|
504 |
-
"eval_steps_per_second": 0.
|
505 |
"eval_tn": 58,
|
506 |
"eval_tp": 61,
|
507 |
"step": 1408
|
@@ -520,10 +520,10 @@
|
|
520 |
"eval_precision": 0.9253731343283582,
|
521 |
"eval_recall": 0.96875,
|
522 |
"eval_roc_auc": 0.966796875,
|
523 |
-
"eval_runtime":
|
524 |
-
"eval_samples_per_second":
|
525 |
"eval_specificity": 0.921875,
|
526 |
-
"eval_steps_per_second": 0.
|
527 |
"eval_tn": 59,
|
528 |
"eval_tp": 62,
|
529 |
"step": 1472
|
@@ -549,10 +549,10 @@
|
|
549 |
"eval_precision": 0.9242424242424242,
|
550 |
"eval_recall": 0.953125,
|
551 |
"eval_roc_auc": 0.966796875,
|
552 |
-
"eval_runtime":
|
553 |
-
"eval_samples_per_second":
|
554 |
"eval_specificity": 0.921875,
|
555 |
-
"eval_steps_per_second": 0.
|
556 |
"eval_tn": 59,
|
557 |
"eval_tp": 61,
|
558 |
"step": 1536
|
@@ -571,10 +571,10 @@
|
|
571 |
"eval_precision": 0.9242424242424242,
|
572 |
"eval_recall": 0.953125,
|
573 |
"eval_roc_auc": 0.964599609375,
|
574 |
-
"eval_runtime":
|
575 |
-
"eval_samples_per_second":
|
576 |
"eval_specificity": 0.921875,
|
577 |
-
"eval_steps_per_second": 0.
|
578 |
"eval_tn": 59,
|
579 |
"eval_tp": 61,
|
580 |
"step": 1600
|
@@ -593,10 +593,10 @@
|
|
593 |
"eval_precision": 0.9508196721311475,
|
594 |
"eval_recall": 0.90625,
|
595 |
"eval_roc_auc": 0.958740234375,
|
596 |
-
"eval_runtime":
|
597 |
-
"eval_samples_per_second":
|
598 |
"eval_specificity": 0.953125,
|
599 |
-
"eval_steps_per_second": 0.
|
600 |
"eval_tn": 61,
|
601 |
"eval_tp": 58,
|
602 |
"step": 1664
|
@@ -615,10 +615,10 @@
|
|
615 |
"eval_precision": 0.921875,
|
616 |
"eval_recall": 0.921875,
|
617 |
"eval_roc_auc": 0.963134765625,
|
618 |
-
"eval_runtime":
|
619 |
-
"eval_samples_per_second":
|
620 |
"eval_specificity": 0.921875,
|
621 |
-
"eval_steps_per_second": 0.
|
622 |
"eval_tn": 59,
|
623 |
"eval_tp": 59,
|
624 |
"step": 1728
|
@@ -637,10 +637,10 @@
|
|
637 |
"eval_precision": 0.9375,
|
638 |
"eval_recall": 0.9375,
|
639 |
"eval_roc_auc": 0.967041015625,
|
640 |
-
"eval_runtime":
|
641 |
-
"eval_samples_per_second":
|
642 |
"eval_specificity": 0.9375,
|
643 |
-
"eval_steps_per_second": 0.
|
644 |
"eval_tn": 60,
|
645 |
"eval_tp": 60,
|
646 |
"step": 1792
|
@@ -659,10 +659,10 @@
|
|
659 |
"eval_precision": 0.9354838709677419,
|
660 |
"eval_recall": 0.90625,
|
661 |
"eval_roc_auc": 0.95654296875,
|
662 |
-
"eval_runtime":
|
663 |
-
"eval_samples_per_second":
|
664 |
"eval_specificity": 0.9375,
|
665 |
-
"eval_steps_per_second": 0.
|
666 |
"eval_tn": 60,
|
667 |
"eval_tp": 58,
|
668 |
"step": 1856
|
@@ -681,10 +681,10 @@
|
|
681 |
"eval_precision": 0.9516129032258065,
|
682 |
"eval_recall": 0.921875,
|
683 |
"eval_roc_auc": 0.96875,
|
684 |
-
"eval_runtime":
|
685 |
-
"eval_samples_per_second":
|
686 |
"eval_specificity": 0.953125,
|
687 |
-
"eval_steps_per_second": 0.
|
688 |
"eval_tn": 61,
|
689 |
"eval_tp": 59,
|
690 |
"step": 1920
|
@@ -703,10 +703,10 @@
|
|
703 |
"eval_precision": 0.9104477611940298,
|
704 |
"eval_recall": 0.953125,
|
705 |
"eval_roc_auc": 0.97265625,
|
706 |
-
"eval_runtime":
|
707 |
-
"eval_samples_per_second":
|
708 |
"eval_specificity": 0.90625,
|
709 |
-
"eval_steps_per_second": 0.
|
710 |
"eval_tn": 58,
|
711 |
"eval_tp": 61,
|
712 |
"step": 1984
|
@@ -732,10 +732,10 @@
|
|
732 |
"eval_precision": 0.9375,
|
733 |
"eval_recall": 0.9375,
|
734 |
"eval_roc_auc": 0.96875,
|
735 |
-
"eval_runtime":
|
736 |
-
"eval_samples_per_second":
|
737 |
"eval_specificity": 0.9375,
|
738 |
-
"eval_steps_per_second": 0.
|
739 |
"eval_tn": 60,
|
740 |
"eval_tp": 60,
|
741 |
"step": 2048
|
@@ -754,10 +754,10 @@
|
|
754 |
"eval_precision": 0.9242424242424242,
|
755 |
"eval_recall": 0.953125,
|
756 |
"eval_roc_auc": 0.966064453125,
|
757 |
-
"eval_runtime":
|
758 |
-
"eval_samples_per_second":
|
759 |
"eval_specificity": 0.921875,
|
760 |
-
"eval_steps_per_second": 0.
|
761 |
"eval_tn": 59,
|
762 |
"eval_tp": 61,
|
763 |
"step": 2112
|
@@ -776,10 +776,10 @@
|
|
776 |
"eval_precision": 0.9104477611940298,
|
777 |
"eval_recall": 0.953125,
|
778 |
"eval_roc_auc": 0.970703125,
|
779 |
-
"eval_runtime":
|
780 |
-
"eval_samples_per_second":
|
781 |
"eval_specificity": 0.90625,
|
782 |
-
"eval_steps_per_second": 0.
|
783 |
"eval_tn": 58,
|
784 |
"eval_tp": 61,
|
785 |
"step": 2176
|
@@ -798,10 +798,10 @@
|
|
798 |
"eval_precision": 0.9242424242424242,
|
799 |
"eval_recall": 0.953125,
|
800 |
"eval_roc_auc": 0.963134765625,
|
801 |
-
"eval_runtime":
|
802 |
-
"eval_samples_per_second":
|
803 |
"eval_specificity": 0.921875,
|
804 |
-
"eval_steps_per_second": 0.
|
805 |
"eval_tn": 59,
|
806 |
"eval_tp": 61,
|
807 |
"step": 2240
|
@@ -820,10 +820,10 @@
|
|
820 |
"eval_precision": 0.9516129032258065,
|
821 |
"eval_recall": 0.921875,
|
822 |
"eval_roc_auc": 0.96875,
|
823 |
-
"eval_runtime":
|
824 |
-
"eval_samples_per_second":
|
825 |
"eval_specificity": 0.953125,
|
826 |
-
"eval_steps_per_second": 0.
|
827 |
"eval_tn": 61,
|
828 |
"eval_tp": 59,
|
829 |
"step": 2304
|
@@ -842,10 +842,10 @@
|
|
842 |
"eval_precision": 0.9384615384615385,
|
843 |
"eval_recall": 0.953125,
|
844 |
"eval_roc_auc": 0.9677734375,
|
845 |
-
"eval_runtime":
|
846 |
-
"eval_samples_per_second":
|
847 |
"eval_specificity": 0.9375,
|
848 |
-
"eval_steps_per_second": 0.
|
849 |
"eval_tn": 60,
|
850 |
"eval_tp": 61,
|
851 |
"step": 2368
|
@@ -864,10 +864,10 @@
|
|
864 |
"eval_precision": 0.9384615384615385,
|
865 |
"eval_recall": 0.953125,
|
866 |
"eval_roc_auc": 0.961181640625,
|
867 |
-
"eval_runtime":
|
868 |
-
"eval_samples_per_second":
|
869 |
"eval_specificity": 0.9375,
|
870 |
-
"eval_steps_per_second": 0.
|
871 |
"eval_tn": 60,
|
872 |
"eval_tp": 61,
|
873 |
"step": 2432
|
@@ -886,10 +886,10 @@
|
|
886 |
"eval_precision": 0.9384615384615385,
|
887 |
"eval_recall": 0.953125,
|
888 |
"eval_roc_auc": 0.96435546875,
|
889 |
-
"eval_runtime":
|
890 |
-
"eval_samples_per_second":
|
891 |
"eval_specificity": 0.9375,
|
892 |
-
"eval_steps_per_second": 0.
|
893 |
"eval_tn": 60,
|
894 |
"eval_tp": 61,
|
895 |
"step": 2496
|
@@ -915,13 +915,240 @@
|
|
915 |
"eval_precision": 0.953125,
|
916 |
"eval_recall": 0.953125,
|
917 |
"eval_roc_auc": 0.96240234375,
|
918 |
-
"eval_runtime":
|
919 |
-
"eval_samples_per_second":
|
920 |
"eval_specificity": 0.953125,
|
921 |
-
"eval_steps_per_second": 0.
|
922 |
"eval_tn": 61,
|
923 |
"eval_tp": 61,
|
924 |
"step": 2560
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
925 |
}
|
926 |
],
|
927 |
"logging_steps": 500,
|
@@ -936,12 +1163,12 @@
|
|
936 |
"should_evaluate": false,
|
937 |
"should_log": false,
|
938 |
"should_save": true,
|
939 |
-
"should_training_stop":
|
940 |
},
|
941 |
"attributes": {}
|
942 |
}
|
943 |
},
|
944 |
-
"total_flos":
|
945 |
"train_batch_size": 8,
|
946 |
"trial_name": null,
|
947 |
"trial_params": null
|
|
|
1 |
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 50.0,
|
5 |
"eval_steps": 500,
|
6 |
+
"global_step": 3200,
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
|
|
22 |
"eval_precision": 0.0,
|
23 |
"eval_recall": 0.0,
|
24 |
"eval_roc_auc": 0.630615234375,
|
25 |
+
"eval_runtime": 5.1525,
|
26 |
+
"eval_samples_per_second": 24.843,
|
27 |
"eval_specificity": 1.0,
|
28 |
+
"eval_steps_per_second": 0.194,
|
29 |
"eval_tn": 64,
|
30 |
"eval_tp": 0,
|
31 |
"step": 64
|
|
|
44 |
"eval_precision": 0.6276595744680851,
|
45 |
"eval_recall": 0.921875,
|
46 |
"eval_roc_auc": 0.797119140625,
|
47 |
+
"eval_runtime": 4.5248,
|
48 |
+
"eval_samples_per_second": 28.288,
|
49 |
"eval_specificity": 0.453125,
|
50 |
+
"eval_steps_per_second": 0.221,
|
51 |
"eval_tn": 29,
|
52 |
"eval_tp": 59,
|
53 |
"step": 128
|
|
|
66 |
"eval_precision": 0.8518518518518519,
|
67 |
"eval_recall": 0.71875,
|
68 |
"eval_roc_auc": 0.891845703125,
|
69 |
+
"eval_runtime": 4.5497,
|
70 |
+
"eval_samples_per_second": 28.134,
|
71 |
"eval_specificity": 0.875,
|
72 |
+
"eval_steps_per_second": 0.22,
|
73 |
"eval_tn": 56,
|
74 |
"eval_tp": 46,
|
75 |
"step": 192
|
|
|
88 |
"eval_precision": 0.9090909090909091,
|
89 |
"eval_recall": 0.78125,
|
90 |
"eval_roc_auc": 0.934814453125,
|
91 |
+
"eval_runtime": 3.0168,
|
92 |
+
"eval_samples_per_second": 42.43,
|
93 |
"eval_specificity": 0.921875,
|
94 |
+
"eval_steps_per_second": 0.331,
|
95 |
"eval_tn": 59,
|
96 |
"eval_tp": 50,
|
97 |
"step": 256
|
|
|
110 |
"eval_precision": 0.8507462686567164,
|
111 |
"eval_recall": 0.890625,
|
112 |
"eval_roc_auc": 0.938232421875,
|
113 |
+
"eval_runtime": 4.3385,
|
114 |
+
"eval_samples_per_second": 29.504,
|
115 |
"eval_specificity": 0.84375,
|
116 |
+
"eval_steps_per_second": 0.23,
|
117 |
"eval_tn": 54,
|
118 |
"eval_tp": 57,
|
119 |
"step": 320
|
|
|
132 |
"eval_precision": 0.8378378378378378,
|
133 |
"eval_recall": 0.96875,
|
134 |
"eval_roc_auc": 0.951171875,
|
135 |
+
"eval_runtime": 4.3284,
|
136 |
+
"eval_samples_per_second": 29.572,
|
137 |
"eval_specificity": 0.8125,
|
138 |
+
"eval_steps_per_second": 0.231,
|
139 |
"eval_tn": 52,
|
140 |
"eval_tp": 62,
|
141 |
"step": 384
|
|
|
154 |
"eval_precision": 0.921875,
|
155 |
"eval_recall": 0.921875,
|
156 |
"eval_roc_auc": 0.96142578125,
|
157 |
+
"eval_runtime": 4.3421,
|
158 |
+
"eval_samples_per_second": 29.479,
|
159 |
"eval_specificity": 0.921875,
|
160 |
+
"eval_steps_per_second": 0.23,
|
161 |
"eval_tn": 59,
|
162 |
"eval_tp": 59,
|
163 |
"step": 448
|
|
|
183 |
"eval_precision": 0.8076923076923077,
|
184 |
"eval_recall": 0.984375,
|
185 |
"eval_roc_auc": 0.953369140625,
|
186 |
+
"eval_runtime": 4.3535,
|
187 |
+
"eval_samples_per_second": 29.401,
|
188 |
"eval_specificity": 0.765625,
|
189 |
+
"eval_steps_per_second": 0.23,
|
190 |
"eval_tn": 49,
|
191 |
"eval_tp": 63,
|
192 |
"step": 512
|
|
|
205 |
"eval_precision": 0.8181818181818182,
|
206 |
"eval_recall": 0.984375,
|
207 |
"eval_roc_auc": 0.95751953125,
|
208 |
+
"eval_runtime": 4.5389,
|
209 |
+
"eval_samples_per_second": 28.201,
|
210 |
"eval_specificity": 0.78125,
|
211 |
+
"eval_steps_per_second": 0.22,
|
212 |
"eval_tn": 50,
|
213 |
"eval_tp": 63,
|
214 |
"step": 576
|
|
|
227 |
"eval_precision": 0.9322033898305084,
|
228 |
"eval_recall": 0.859375,
|
229 |
"eval_roc_auc": 0.9658203125,
|
230 |
+
"eval_runtime": 4.4902,
|
231 |
+
"eval_samples_per_second": 28.507,
|
232 |
"eval_specificity": 0.9375,
|
233 |
+
"eval_steps_per_second": 0.223,
|
234 |
"eval_tn": 60,
|
235 |
"eval_tp": 55,
|
236 |
"step": 640
|
|
|
249 |
"eval_precision": 0.8923076923076924,
|
250 |
"eval_recall": 0.90625,
|
251 |
"eval_roc_auc": 0.959716796875,
|
252 |
+
"eval_runtime": 4.487,
|
253 |
+
"eval_samples_per_second": 28.527,
|
254 |
"eval_specificity": 0.890625,
|
255 |
+
"eval_steps_per_second": 0.223,
|
256 |
"eval_tn": 57,
|
257 |
"eval_tp": 58,
|
258 |
"step": 704
|
|
|
271 |
"eval_precision": 0.9333333333333333,
|
272 |
"eval_recall": 0.875,
|
273 |
"eval_roc_auc": 0.96240234375,
|
274 |
+
"eval_runtime": 4.5247,
|
275 |
+
"eval_samples_per_second": 28.289,
|
276 |
"eval_specificity": 0.9375,
|
277 |
+
"eval_steps_per_second": 0.221,
|
278 |
"eval_tn": 60,
|
279 |
"eval_tp": 56,
|
280 |
"step": 768
|
|
|
293 |
"eval_precision": 0.9655172413793104,
|
294 |
"eval_recall": 0.875,
|
295 |
"eval_roc_auc": 0.963134765625,
|
296 |
+
"eval_runtime": 4.5632,
|
297 |
+
"eval_samples_per_second": 28.051,
|
298 |
"eval_specificity": 0.96875,
|
299 |
+
"eval_steps_per_second": 0.219,
|
300 |
"eval_tn": 62,
|
301 |
"eval_tp": 56,
|
302 |
"step": 832
|
|
|
315 |
"eval_precision": 0.9180327868852459,
|
316 |
"eval_recall": 0.875,
|
317 |
"eval_roc_auc": 0.962158203125,
|
318 |
+
"eval_runtime": 4.5111,
|
319 |
+
"eval_samples_per_second": 28.375,
|
320 |
"eval_specificity": 0.921875,
|
321 |
+
"eval_steps_per_second": 0.222,
|
322 |
"eval_tn": 59,
|
323 |
"eval_tp": 56,
|
324 |
"step": 896
|
|
|
337 |
"eval_precision": 0.9180327868852459,
|
338 |
"eval_recall": 0.875,
|
339 |
"eval_roc_auc": 0.947509765625,
|
340 |
+
"eval_runtime": 4.5543,
|
341 |
+
"eval_samples_per_second": 28.105,
|
342 |
"eval_specificity": 0.921875,
|
343 |
+
"eval_steps_per_second": 0.22,
|
344 |
"eval_tn": 59,
|
345 |
"eval_tp": 56,
|
346 |
"step": 960
|
|
|
366 |
"eval_precision": 0.9206349206349206,
|
367 |
"eval_recall": 0.90625,
|
368 |
"eval_roc_auc": 0.94677734375,
|
369 |
+
"eval_runtime": 4.5536,
|
370 |
+
"eval_samples_per_second": 28.11,
|
371 |
"eval_specificity": 0.921875,
|
372 |
+
"eval_steps_per_second": 0.22,
|
373 |
"eval_tn": 59,
|
374 |
"eval_tp": 58,
|
375 |
"step": 1024
|
|
|
388 |
"eval_precision": 0.9322033898305084,
|
389 |
"eval_recall": 0.859375,
|
390 |
"eval_roc_auc": 0.96875,
|
391 |
+
"eval_runtime": 4.548,
|
392 |
+
"eval_samples_per_second": 28.144,
|
393 |
"eval_specificity": 0.9375,
|
394 |
+
"eval_steps_per_second": 0.22,
|
395 |
"eval_tn": 60,
|
396 |
"eval_tp": 55,
|
397 |
"step": 1088
|
|
|
410 |
"eval_precision": 0.9180327868852459,
|
411 |
"eval_recall": 0.875,
|
412 |
"eval_roc_auc": 0.95947265625,
|
413 |
+
"eval_runtime": 4.5489,
|
414 |
+
"eval_samples_per_second": 28.139,
|
415 |
"eval_specificity": 0.921875,
|
416 |
+
"eval_steps_per_second": 0.22,
|
417 |
"eval_tn": 59,
|
418 |
"eval_tp": 56,
|
419 |
"step": 1152
|
|
|
432 |
"eval_precision": 0.9230769230769231,
|
433 |
"eval_recall": 0.9375,
|
434 |
"eval_roc_auc": 0.958740234375,
|
435 |
+
"eval_runtime": 4.5524,
|
436 |
+
"eval_samples_per_second": 28.117,
|
437 |
"eval_specificity": 0.921875,
|
438 |
+
"eval_steps_per_second": 0.22,
|
439 |
"eval_tn": 59,
|
440 |
"eval_tp": 60,
|
441 |
"step": 1216
|
|
|
454 |
"eval_precision": 0.9104477611940298,
|
455 |
"eval_recall": 0.953125,
|
456 |
"eval_roc_auc": 0.951416015625,
|
457 |
+
"eval_runtime": 4.5492,
|
458 |
+
"eval_samples_per_second": 28.137,
|
459 |
"eval_specificity": 0.90625,
|
460 |
+
"eval_steps_per_second": 0.22,
|
461 |
"eval_tn": 58,
|
462 |
"eval_tp": 61,
|
463 |
"step": 1280
|
|
|
476 |
"eval_precision": 0.9491525423728814,
|
477 |
"eval_recall": 0.875,
|
478 |
"eval_roc_auc": 0.95654296875,
|
479 |
+
"eval_runtime": 4.5431,
|
480 |
+
"eval_samples_per_second": 28.175,
|
481 |
"eval_specificity": 0.953125,
|
482 |
+
"eval_steps_per_second": 0.22,
|
483 |
"eval_tn": 61,
|
484 |
"eval_tp": 56,
|
485 |
"step": 1344
|
|
|
498 |
"eval_precision": 0.9104477611940298,
|
499 |
"eval_recall": 0.953125,
|
500 |
"eval_roc_auc": 0.95703125,
|
501 |
+
"eval_runtime": 4.54,
|
502 |
+
"eval_samples_per_second": 28.194,
|
503 |
"eval_specificity": 0.90625,
|
504 |
+
"eval_steps_per_second": 0.22,
|
505 |
"eval_tn": 58,
|
506 |
"eval_tp": 61,
|
507 |
"step": 1408
|
|
|
520 |
"eval_precision": 0.9253731343283582,
|
521 |
"eval_recall": 0.96875,
|
522 |
"eval_roc_auc": 0.966796875,
|
523 |
+
"eval_runtime": 4.5302,
|
524 |
+
"eval_samples_per_second": 28.255,
|
525 |
"eval_specificity": 0.921875,
|
526 |
+
"eval_steps_per_second": 0.221,
|
527 |
"eval_tn": 59,
|
528 |
"eval_tp": 62,
|
529 |
"step": 1472
|
|
|
549 |
"eval_precision": 0.9242424242424242,
|
550 |
"eval_recall": 0.953125,
|
551 |
"eval_roc_auc": 0.966796875,
|
552 |
+
"eval_runtime": 4.5437,
|
553 |
+
"eval_samples_per_second": 28.171,
|
554 |
"eval_specificity": 0.921875,
|
555 |
+
"eval_steps_per_second": 0.22,
|
556 |
"eval_tn": 59,
|
557 |
"eval_tp": 61,
|
558 |
"step": 1536
|
|
|
571 |
"eval_precision": 0.9242424242424242,
|
572 |
"eval_recall": 0.953125,
|
573 |
"eval_roc_auc": 0.964599609375,
|
574 |
+
"eval_runtime": 4.5503,
|
575 |
+
"eval_samples_per_second": 28.13,
|
576 |
"eval_specificity": 0.921875,
|
577 |
+
"eval_steps_per_second": 0.22,
|
578 |
"eval_tn": 59,
|
579 |
"eval_tp": 61,
|
580 |
"step": 1600
|
|
|
593 |
"eval_precision": 0.9508196721311475,
|
594 |
"eval_recall": 0.90625,
|
595 |
"eval_roc_auc": 0.958740234375,
|
596 |
+
"eval_runtime": 4.5413,
|
597 |
+
"eval_samples_per_second": 28.186,
|
598 |
"eval_specificity": 0.953125,
|
599 |
+
"eval_steps_per_second": 0.22,
|
600 |
"eval_tn": 61,
|
601 |
"eval_tp": 58,
|
602 |
"step": 1664
|
|
|
615 |
"eval_precision": 0.921875,
|
616 |
"eval_recall": 0.921875,
|
617 |
"eval_roc_auc": 0.963134765625,
|
618 |
+
"eval_runtime": 4.5504,
|
619 |
+
"eval_samples_per_second": 28.129,
|
620 |
"eval_specificity": 0.921875,
|
621 |
+
"eval_steps_per_second": 0.22,
|
622 |
"eval_tn": 59,
|
623 |
"eval_tp": 59,
|
624 |
"step": 1728
|
|
|
637 |
"eval_precision": 0.9375,
|
638 |
"eval_recall": 0.9375,
|
639 |
"eval_roc_auc": 0.967041015625,
|
640 |
+
"eval_runtime": 4.5474,
|
641 |
+
"eval_samples_per_second": 28.148,
|
642 |
"eval_specificity": 0.9375,
|
643 |
+
"eval_steps_per_second": 0.22,
|
644 |
"eval_tn": 60,
|
645 |
"eval_tp": 60,
|
646 |
"step": 1792
|
|
|
659 |
"eval_precision": 0.9354838709677419,
|
660 |
"eval_recall": 0.90625,
|
661 |
"eval_roc_auc": 0.95654296875,
|
662 |
+
"eval_runtime": 4.5528,
|
663 |
+
"eval_samples_per_second": 28.115,
|
664 |
"eval_specificity": 0.9375,
|
665 |
+
"eval_steps_per_second": 0.22,
|
666 |
"eval_tn": 60,
|
667 |
"eval_tp": 58,
|
668 |
"step": 1856
|
|
|
681 |
"eval_precision": 0.9516129032258065,
|
682 |
"eval_recall": 0.921875,
|
683 |
"eval_roc_auc": 0.96875,
|
684 |
+
"eval_runtime": 4.5887,
|
685 |
+
"eval_samples_per_second": 27.895,
|
686 |
"eval_specificity": 0.953125,
|
687 |
+
"eval_steps_per_second": 0.218,
|
688 |
"eval_tn": 61,
|
689 |
"eval_tp": 59,
|
690 |
"step": 1920
|
|
|
703 |
"eval_precision": 0.9104477611940298,
|
704 |
"eval_recall": 0.953125,
|
705 |
"eval_roc_auc": 0.97265625,
|
706 |
+
"eval_runtime": 4.5517,
|
707 |
+
"eval_samples_per_second": 28.121,
|
708 |
"eval_specificity": 0.90625,
|
709 |
+
"eval_steps_per_second": 0.22,
|
710 |
"eval_tn": 58,
|
711 |
"eval_tp": 61,
|
712 |
"step": 1984
|
|
|
732 |
"eval_precision": 0.9375,
|
733 |
"eval_recall": 0.9375,
|
734 |
"eval_roc_auc": 0.96875,
|
735 |
+
"eval_runtime": 4.6396,
|
736 |
+
"eval_samples_per_second": 27.588,
|
737 |
"eval_specificity": 0.9375,
|
738 |
+
"eval_steps_per_second": 0.216,
|
739 |
"eval_tn": 60,
|
740 |
"eval_tp": 60,
|
741 |
"step": 2048
|
|
|
754 |
"eval_precision": 0.9242424242424242,
|
755 |
"eval_recall": 0.953125,
|
756 |
"eval_roc_auc": 0.966064453125,
|
757 |
+
"eval_runtime": 4.676,
|
758 |
+
"eval_samples_per_second": 27.374,
|
759 |
"eval_specificity": 0.921875,
|
760 |
+
"eval_steps_per_second": 0.214,
|
761 |
"eval_tn": 59,
|
762 |
"eval_tp": 61,
|
763 |
"step": 2112
|
|
|
776 |
"eval_precision": 0.9104477611940298,
|
777 |
"eval_recall": 0.953125,
|
778 |
"eval_roc_auc": 0.970703125,
|
779 |
+
"eval_runtime": 4.6816,
|
780 |
+
"eval_samples_per_second": 27.341,
|
781 |
"eval_specificity": 0.90625,
|
782 |
+
"eval_steps_per_second": 0.214,
|
783 |
"eval_tn": 58,
|
784 |
"eval_tp": 61,
|
785 |
"step": 2176
|
|
|
798 |
"eval_precision": 0.9242424242424242,
|
799 |
"eval_recall": 0.953125,
|
800 |
"eval_roc_auc": 0.963134765625,
|
801 |
+
"eval_runtime": 4.6414,
|
802 |
+
"eval_samples_per_second": 27.578,
|
803 |
"eval_specificity": 0.921875,
|
804 |
+
"eval_steps_per_second": 0.215,
|
805 |
"eval_tn": 59,
|
806 |
"eval_tp": 61,
|
807 |
"step": 2240
|
|
|
820 |
"eval_precision": 0.9516129032258065,
|
821 |
"eval_recall": 0.921875,
|
822 |
"eval_roc_auc": 0.96875,
|
823 |
+
"eval_runtime": 4.5146,
|
824 |
+
"eval_samples_per_second": 28.352,
|
825 |
"eval_specificity": 0.953125,
|
826 |
+
"eval_steps_per_second": 0.222,
|
827 |
"eval_tn": 61,
|
828 |
"eval_tp": 59,
|
829 |
"step": 2304
|
|
|
842 |
"eval_precision": 0.9384615384615385,
|
843 |
"eval_recall": 0.953125,
|
844 |
"eval_roc_auc": 0.9677734375,
|
845 |
+
"eval_runtime": 4.5189,
|
846 |
+
"eval_samples_per_second": 28.325,
|
847 |
"eval_specificity": 0.9375,
|
848 |
+
"eval_steps_per_second": 0.221,
|
849 |
"eval_tn": 60,
|
850 |
"eval_tp": 61,
|
851 |
"step": 2368
|
|
|
864 |
"eval_precision": 0.9384615384615385,
|
865 |
"eval_recall": 0.953125,
|
866 |
"eval_roc_auc": 0.961181640625,
|
867 |
+
"eval_runtime": 4.6226,
|
868 |
+
"eval_samples_per_second": 27.69,
|
869 |
"eval_specificity": 0.9375,
|
870 |
+
"eval_steps_per_second": 0.216,
|
871 |
"eval_tn": 60,
|
872 |
"eval_tp": 61,
|
873 |
"step": 2432
|
|
|
886 |
"eval_precision": 0.9384615384615385,
|
887 |
"eval_recall": 0.953125,
|
888 |
"eval_roc_auc": 0.96435546875,
|
889 |
+
"eval_runtime": 4.5344,
|
890 |
+
"eval_samples_per_second": 28.229,
|
891 |
"eval_specificity": 0.9375,
|
892 |
+
"eval_steps_per_second": 0.221,
|
893 |
"eval_tn": 60,
|
894 |
"eval_tp": 61,
|
895 |
"step": 2496
|
|
|
915 |
"eval_precision": 0.953125,
|
916 |
"eval_recall": 0.953125,
|
917 |
"eval_roc_auc": 0.96240234375,
|
918 |
+
"eval_runtime": 4.5169,
|
919 |
+
"eval_samples_per_second": 28.338,
|
920 |
"eval_specificity": 0.953125,
|
921 |
+
"eval_steps_per_second": 0.221,
|
922 |
"eval_tn": 61,
|
923 |
"eval_tp": 61,
|
924 |
"step": 2560
|
925 |
+
},
|
926 |
+
{
|
927 |
+
"epoch": 41.0,
|
928 |
+
"eval_accuracy": 0.9296875,
|
929 |
+
"eval_auc": 0.96435546875,
|
930 |
+
"eval_balanced_acc": 0.9296875,
|
931 |
+
"eval_f1": 0.928,
|
932 |
+
"eval_fn": 6,
|
933 |
+
"eval_fp": 3,
|
934 |
+
"eval_gmean": 0.9293920223726907,
|
935 |
+
"eval_loss": 0.23136085271835327,
|
936 |
+
"eval_matthews_correlation": 0.8603206963184278,
|
937 |
+
"eval_precision": 0.9508196721311475,
|
938 |
+
"eval_recall": 0.90625,
|
939 |
+
"eval_roc_auc": 0.96435546875,
|
940 |
+
"eval_runtime": 4.5098,
|
941 |
+
"eval_samples_per_second": 28.383,
|
942 |
+
"eval_specificity": 0.953125,
|
943 |
+
"eval_steps_per_second": 0.222,
|
944 |
+
"eval_tn": 61,
|
945 |
+
"eval_tp": 58,
|
946 |
+
"step": 2624
|
947 |
+
},
|
948 |
+
{
|
949 |
+
"epoch": 42.0,
|
950 |
+
"eval_accuracy": 0.9453125,
|
951 |
+
"eval_auc": 0.9619140625,
|
952 |
+
"eval_balanced_acc": 0.9453125,
|
953 |
+
"eval_f1": 0.9457364341085271,
|
954 |
+
"eval_fn": 3,
|
955 |
+
"eval_fp": 4,
|
956 |
+
"eval_gmean": 0.9452802163908859,
|
957 |
+
"eval_loss": 0.22938130795955658,
|
958 |
+
"eval_matthews_correlation": 0.8907337387831413,
|
959 |
+
"eval_precision": 0.9384615384615385,
|
960 |
+
"eval_recall": 0.953125,
|
961 |
+
"eval_roc_auc": 0.9619140625,
|
962 |
+
"eval_runtime": 4.5043,
|
963 |
+
"eval_samples_per_second": 28.418,
|
964 |
+
"eval_specificity": 0.9375,
|
965 |
+
"eval_steps_per_second": 0.222,
|
966 |
+
"eval_tn": 60,
|
967 |
+
"eval_tp": 61,
|
968 |
+
"step": 2688
|
969 |
+
},
|
970 |
+
{
|
971 |
+
"epoch": 43.0,
|
972 |
+
"eval_accuracy": 0.9375,
|
973 |
+
"eval_auc": 0.955810546875,
|
974 |
+
"eval_balanced_acc": 0.9375,
|
975 |
+
"eval_f1": 0.9365079365079365,
|
976 |
+
"eval_fn": 5,
|
977 |
+
"eval_fp": 3,
|
978 |
+
"eval_gmean": 0.9373697826231652,
|
979 |
+
"eval_loss": 0.24056339263916016,
|
980 |
+
"eval_matthews_correlation": 0.8754275592730114,
|
981 |
+
"eval_precision": 0.9516129032258065,
|
982 |
+
"eval_recall": 0.921875,
|
983 |
+
"eval_roc_auc": 0.955810546875,
|
984 |
+
"eval_runtime": 4.5092,
|
985 |
+
"eval_samples_per_second": 28.387,
|
986 |
+
"eval_specificity": 0.953125,
|
987 |
+
"eval_steps_per_second": 0.222,
|
988 |
+
"eval_tn": 61,
|
989 |
+
"eval_tp": 59,
|
990 |
+
"step": 2752
|
991 |
+
},
|
992 |
+
{
|
993 |
+
"epoch": 44.0,
|
994 |
+
"eval_accuracy": 0.953125,
|
995 |
+
"eval_auc": 0.957763671875,
|
996 |
+
"eval_balanced_acc": 0.953125,
|
997 |
+
"eval_f1": 0.953125,
|
998 |
+
"eval_fn": 3,
|
999 |
+
"eval_fp": 3,
|
1000 |
+
"eval_gmean": 0.953125,
|
1001 |
+
"eval_loss": 0.22141942381858826,
|
1002 |
+
"eval_matthews_correlation": 0.90625,
|
1003 |
+
"eval_precision": 0.953125,
|
1004 |
+
"eval_recall": 0.953125,
|
1005 |
+
"eval_roc_auc": 0.957763671875,
|
1006 |
+
"eval_runtime": 4.5337,
|
1007 |
+
"eval_samples_per_second": 28.233,
|
1008 |
+
"eval_specificity": 0.953125,
|
1009 |
+
"eval_steps_per_second": 0.221,
|
1010 |
+
"eval_tn": 61,
|
1011 |
+
"eval_tp": 61,
|
1012 |
+
"step": 2816
|
1013 |
+
},
|
1014 |
+
{
|
1015 |
+
"epoch": 45.0,
|
1016 |
+
"eval_accuracy": 0.953125,
|
1017 |
+
"eval_auc": 0.961669921875,
|
1018 |
+
"eval_balanced_acc": 0.953125,
|
1019 |
+
"eval_f1": 0.953125,
|
1020 |
+
"eval_fn": 3,
|
1021 |
+
"eval_fp": 3,
|
1022 |
+
"eval_gmean": 0.953125,
|
1023 |
+
"eval_loss": 0.22277112305164337,
|
1024 |
+
"eval_matthews_correlation": 0.90625,
|
1025 |
+
"eval_precision": 0.953125,
|
1026 |
+
"eval_recall": 0.953125,
|
1027 |
+
"eval_roc_auc": 0.961669921875,
|
1028 |
+
"eval_runtime": 2.3022,
|
1029 |
+
"eval_samples_per_second": 55.599,
|
1030 |
+
"eval_specificity": 0.953125,
|
1031 |
+
"eval_steps_per_second": 0.434,
|
1032 |
+
"eval_tn": 61,
|
1033 |
+
"eval_tp": 61,
|
1034 |
+
"step": 2880
|
1035 |
+
},
|
1036 |
+
{
|
1037 |
+
"epoch": 46.0,
|
1038 |
+
"eval_accuracy": 0.953125,
|
1039 |
+
"eval_auc": 0.95458984375,
|
1040 |
+
"eval_balanced_acc": 0.953125,
|
1041 |
+
"eval_f1": 0.953125,
|
1042 |
+
"eval_fn": 3,
|
1043 |
+
"eval_fp": 3,
|
1044 |
+
"eval_gmean": 0.953125,
|
1045 |
+
"eval_loss": 0.2253548949956894,
|
1046 |
+
"eval_matthews_correlation": 0.90625,
|
1047 |
+
"eval_precision": 0.953125,
|
1048 |
+
"eval_recall": 0.953125,
|
1049 |
+
"eval_roc_auc": 0.95458984375,
|
1050 |
+
"eval_runtime": 4.5066,
|
1051 |
+
"eval_samples_per_second": 28.403,
|
1052 |
+
"eval_specificity": 0.953125,
|
1053 |
+
"eval_steps_per_second": 0.222,
|
1054 |
+
"eval_tn": 61,
|
1055 |
+
"eval_tp": 61,
|
1056 |
+
"step": 2944
|
1057 |
+
},
|
1058 |
+
{
|
1059 |
+
"epoch": 46.875,
|
1060 |
+
"grad_norm": 0.10018020868301392,
|
1061 |
+
"learning_rate": 3.289473684210526e-05,
|
1062 |
+
"loss": 0.0659,
|
1063 |
+
"step": 3000
|
1064 |
+
},
|
1065 |
+
{
|
1066 |
+
"epoch": 47.0,
|
1067 |
+
"eval_accuracy": 0.953125,
|
1068 |
+
"eval_auc": 0.96337890625,
|
1069 |
+
"eval_balanced_acc": 0.953125,
|
1070 |
+
"eval_f1": 0.953125,
|
1071 |
+
"eval_fn": 3,
|
1072 |
+
"eval_fp": 3,
|
1073 |
+
"eval_gmean": 0.953125,
|
1074 |
+
"eval_loss": 0.22502213716506958,
|
1075 |
+
"eval_matthews_correlation": 0.90625,
|
1076 |
+
"eval_precision": 0.953125,
|
1077 |
+
"eval_recall": 0.953125,
|
1078 |
+
"eval_roc_auc": 0.96337890625,
|
1079 |
+
"eval_runtime": 4.5194,
|
1080 |
+
"eval_samples_per_second": 28.322,
|
1081 |
+
"eval_specificity": 0.953125,
|
1082 |
+
"eval_steps_per_second": 0.221,
|
1083 |
+
"eval_tn": 61,
|
1084 |
+
"eval_tp": 61,
|
1085 |
+
"step": 3008
|
1086 |
+
},
|
1087 |
+
{
|
1088 |
+
"epoch": 48.0,
|
1089 |
+
"eval_accuracy": 0.953125,
|
1090 |
+
"eval_auc": 0.963623046875,
|
1091 |
+
"eval_balanced_acc": 0.953125,
|
1092 |
+
"eval_f1": 0.953125,
|
1093 |
+
"eval_fn": 3,
|
1094 |
+
"eval_fp": 3,
|
1095 |
+
"eval_gmean": 0.953125,
|
1096 |
+
"eval_loss": 0.2254004329442978,
|
1097 |
+
"eval_matthews_correlation": 0.90625,
|
1098 |
+
"eval_precision": 0.953125,
|
1099 |
+
"eval_recall": 0.953125,
|
1100 |
+
"eval_roc_auc": 0.963623046875,
|
1101 |
+
"eval_runtime": 4.5174,
|
1102 |
+
"eval_samples_per_second": 28.335,
|
1103 |
+
"eval_specificity": 0.953125,
|
1104 |
+
"eval_steps_per_second": 0.221,
|
1105 |
+
"eval_tn": 61,
|
1106 |
+
"eval_tp": 61,
|
1107 |
+
"step": 3072
|
1108 |
+
},
|
1109 |
+
{
|
1110 |
+
"epoch": 49.0,
|
1111 |
+
"eval_accuracy": 0.953125,
|
1112 |
+
"eval_auc": 0.96826171875,
|
1113 |
+
"eval_balanced_acc": 0.953125,
|
1114 |
+
"eval_f1": 0.953125,
|
1115 |
+
"eval_fn": 3,
|
1116 |
+
"eval_fp": 3,
|
1117 |
+
"eval_gmean": 0.953125,
|
1118 |
+
"eval_loss": 0.2249780297279358,
|
1119 |
+
"eval_matthews_correlation": 0.90625,
|
1120 |
+
"eval_precision": 0.953125,
|
1121 |
+
"eval_recall": 0.953125,
|
1122 |
+
"eval_roc_auc": 0.96826171875,
|
1123 |
+
"eval_runtime": 4.5274,
|
1124 |
+
"eval_samples_per_second": 28.272,
|
1125 |
+
"eval_specificity": 0.953125,
|
1126 |
+
"eval_steps_per_second": 0.221,
|
1127 |
+
"eval_tn": 61,
|
1128 |
+
"eval_tp": 61,
|
1129 |
+
"step": 3136
|
1130 |
+
},
|
1131 |
+
{
|
1132 |
+
"epoch": 50.0,
|
1133 |
+
"eval_accuracy": 0.953125,
|
1134 |
+
"eval_auc": 0.9677734375,
|
1135 |
+
"eval_balanced_acc": 0.953125,
|
1136 |
+
"eval_f1": 0.953125,
|
1137 |
+
"eval_fn": 3,
|
1138 |
+
"eval_fp": 3,
|
1139 |
+
"eval_gmean": 0.953125,
|
1140 |
+
"eval_loss": 0.2252984642982483,
|
1141 |
+
"eval_matthews_correlation": 0.90625,
|
1142 |
+
"eval_precision": 0.953125,
|
1143 |
+
"eval_recall": 0.953125,
|
1144 |
+
"eval_roc_auc": 0.9677734375,
|
1145 |
+
"eval_runtime": 4.5509,
|
1146 |
+
"eval_samples_per_second": 28.127,
|
1147 |
+
"eval_specificity": 0.953125,
|
1148 |
+
"eval_steps_per_second": 0.22,
|
1149 |
+
"eval_tn": 61,
|
1150 |
+
"eval_tp": 61,
|
1151 |
+
"step": 3200
|
1152 |
}
|
1153 |
],
|
1154 |
"logging_steps": 500,
|
|
|
1163 |
"should_evaluate": false,
|
1164 |
"should_log": false,
|
1165 |
"should_save": true,
|
1166 |
+
"should_training_stop": true
|
1167 |
},
|
1168 |
"attributes": {}
|
1169 |
}
|
1170 |
},
|
1171 |
+
"total_flos": 6180004916191008.0,
|
1172 |
"train_batch_size": 8,
|
1173 |
"trial_name": null,
|
1174 |
"trial_params": null
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0836c1d1011341b50b6969f934d0a13595b9d18eb2158049c63cbba331fef33a
|
3 |
+
size 5496
|