Commit
•
abd3c44
1
Parent(s):
962e35f
Training in progress, step 3000
Browse files- last-checkpoint/optimizer.pt +1 -1
- last-checkpoint/pytorch_model.bin +1 -1
- last-checkpoint/rng_state.pth +1 -1
- last-checkpoint/scaler.pt +1 -1
- last-checkpoint/scheduler.pt +1 -1
- last-checkpoint/trainer_state.json +254 -5
- pytorch_model.bin +1 -1
- runs/Jan28_23-19-38_gpu07.cyverse.org/events.out.tfevents.1674973221.gpu07.cyverse.org.79624.0 +2 -2
last-checkpoint/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 12347192855
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:28231ed2fa6db3a8ba9dec24805c24d474d5a98e2fc81e01f9dd3ec22fef6617
|
| 3 |
size 12347192855
|
last-checkpoint/pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 6173655480
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:188ae5e3d4f5fd4f3ea79612ae9050f69c98af92f2396885f41ca19d41d84d28
|
| 3 |
size 6173655480
|
last-checkpoint/rng_state.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14575
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fedfd6bd95ed17c2f4580cf3014f35b56bcb8755a40fc6136021f6ca87c22803
|
| 3 |
size 14575
|
last-checkpoint/scaler.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 557
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2ada1c4bf7218841e7f1bffbf7fa805df783718f35f0ea75965dbcfe82b6b0a0
|
| 3 |
size 557
|
last-checkpoint/scheduler.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 627
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:90be8dc230b164dd2d5b3c32eb6c2a7edeaf871fd8d5b65aa972b35423500fc7
|
| 3 |
size 627
|
last-checkpoint/trainer_state.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "/storage/elsayedissa/whisper-large-v2-arabic-24h/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -504,11 +504,260 @@
|
|
| 504 |
"eval_steps_per_second": 0.147,
|
| 505 |
"eval_wer": 0.42719787333206743,
|
| 506 |
"step": 2000
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 507 |
}
|
| 508 |
],
|
| 509 |
"max_steps": 5000,
|
| 510 |
"num_train_epochs": 9,
|
| 511 |
-
"total_flos":
|
| 512 |
"trial_name": null,
|
| 513 |
"trial_params": null
|
| 514 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.4207764409383577,
|
| 3 |
+
"best_model_checkpoint": "/storage/elsayedissa/whisper-large-v2-arabic-24h/checkpoint-3000",
|
| 4 |
+
"epoch": 5.347593582887701,
|
| 5 |
+
"global_step": 3000,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 504 |
"eval_steps_per_second": 0.147,
|
| 505 |
"eval_wer": 0.42719787333206743,
|
| 506 |
"step": 2000
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 3.61,
|
| 510 |
+
"learning_rate": 6.6222222222222236e-06,
|
| 511 |
+
"loss": 0.051,
|
| 512 |
+
"step": 2025
|
| 513 |
+
},
|
| 514 |
+
{
|
| 515 |
+
"epoch": 3.65,
|
| 516 |
+
"learning_rate": 6.566666666666667e-06,
|
| 517 |
+
"loss": 0.0552,
|
| 518 |
+
"step": 2050
|
| 519 |
+
},
|
| 520 |
+
{
|
| 521 |
+
"epoch": 3.7,
|
| 522 |
+
"learning_rate": 6.511111111111112e-06,
|
| 523 |
+
"loss": 0.0559,
|
| 524 |
+
"step": 2075
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 3.74,
|
| 528 |
+
"learning_rate": 6.455555555555556e-06,
|
| 529 |
+
"loss": 0.0562,
|
| 530 |
+
"step": 2100
|
| 531 |
+
},
|
| 532 |
+
{
|
| 533 |
+
"epoch": 3.79,
|
| 534 |
+
"learning_rate": 6.4000000000000006e-06,
|
| 535 |
+
"loss": 0.0597,
|
| 536 |
+
"step": 2125
|
| 537 |
+
},
|
| 538 |
+
{
|
| 539 |
+
"epoch": 3.83,
|
| 540 |
+
"learning_rate": 6.3444444444444454e-06,
|
| 541 |
+
"loss": 0.0512,
|
| 542 |
+
"step": 2150
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"epoch": 3.88,
|
| 546 |
+
"learning_rate": 6.28888888888889e-06,
|
| 547 |
+
"loss": 0.0533,
|
| 548 |
+
"step": 2175
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"epoch": 3.92,
|
| 552 |
+
"learning_rate": 6.2333333333333335e-06,
|
| 553 |
+
"loss": 0.0545,
|
| 554 |
+
"step": 2200
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"epoch": 3.97,
|
| 558 |
+
"learning_rate": 6.177777777777778e-06,
|
| 559 |
+
"loss": 0.0538,
|
| 560 |
+
"step": 2225
|
| 561 |
+
},
|
| 562 |
+
{
|
| 563 |
+
"epoch": 4.01,
|
| 564 |
+
"learning_rate": 6.124444444444445e-06,
|
| 565 |
+
"loss": 0.05,
|
| 566 |
+
"step": 2250
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"epoch": 4.06,
|
| 570 |
+
"learning_rate": 6.06888888888889e-06,
|
| 571 |
+
"loss": 0.0303,
|
| 572 |
+
"step": 2275
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 4.1,
|
| 576 |
+
"learning_rate": 6.013333333333335e-06,
|
| 577 |
+
"loss": 0.0259,
|
| 578 |
+
"step": 2300
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"epoch": 4.14,
|
| 582 |
+
"learning_rate": 5.957777777777778e-06,
|
| 583 |
+
"loss": 0.0257,
|
| 584 |
+
"step": 2325
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 4.19,
|
| 588 |
+
"learning_rate": 5.902222222222223e-06,
|
| 589 |
+
"loss": 0.0303,
|
| 590 |
+
"step": 2350
|
| 591 |
+
},
|
| 592 |
+
{
|
| 593 |
+
"epoch": 4.23,
|
| 594 |
+
"learning_rate": 5.846666666666667e-06,
|
| 595 |
+
"loss": 0.0292,
|
| 596 |
+
"step": 2375
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 4.28,
|
| 600 |
+
"learning_rate": 5.791111111111112e-06,
|
| 601 |
+
"loss": 0.0307,
|
| 602 |
+
"step": 2400
|
| 603 |
+
},
|
| 604 |
+
{
|
| 605 |
+
"epoch": 4.32,
|
| 606 |
+
"learning_rate": 5.735555555555557e-06,
|
| 607 |
+
"loss": 0.0274,
|
| 608 |
+
"step": 2425
|
| 609 |
+
},
|
| 610 |
+
{
|
| 611 |
+
"epoch": 4.37,
|
| 612 |
+
"learning_rate": 5.68e-06,
|
| 613 |
+
"loss": 0.0277,
|
| 614 |
+
"step": 2450
|
| 615 |
+
},
|
| 616 |
+
{
|
| 617 |
+
"epoch": 4.41,
|
| 618 |
+
"learning_rate": 5.624444444444445e-06,
|
| 619 |
+
"loss": 0.0277,
|
| 620 |
+
"step": 2475
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"epoch": 4.46,
|
| 624 |
+
"learning_rate": 5.56888888888889e-06,
|
| 625 |
+
"loss": 0.0327,
|
| 626 |
+
"step": 2500
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 4.5,
|
| 630 |
+
"learning_rate": 5.513333333333334e-06,
|
| 631 |
+
"loss": 0.0268,
|
| 632 |
+
"step": 2525
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 4.55,
|
| 636 |
+
"learning_rate": 5.4577777777777785e-06,
|
| 637 |
+
"loss": 0.0317,
|
| 638 |
+
"step": 2550
|
| 639 |
+
},
|
| 640 |
+
{
|
| 641 |
+
"epoch": 4.59,
|
| 642 |
+
"learning_rate": 5.402222222222223e-06,
|
| 643 |
+
"loss": 0.0289,
|
| 644 |
+
"step": 2575
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 4.63,
|
| 648 |
+
"learning_rate": 5.346666666666667e-06,
|
| 649 |
+
"loss": 0.0304,
|
| 650 |
+
"step": 2600
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 4.68,
|
| 654 |
+
"learning_rate": 5.2911111111111115e-06,
|
| 655 |
+
"loss": 0.0304,
|
| 656 |
+
"step": 2625
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 4.72,
|
| 660 |
+
"learning_rate": 5.235555555555556e-06,
|
| 661 |
+
"loss": 0.0286,
|
| 662 |
+
"step": 2650
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 4.77,
|
| 666 |
+
"learning_rate": 5.18e-06,
|
| 667 |
+
"loss": 0.0285,
|
| 668 |
+
"step": 2675
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 4.81,
|
| 672 |
+
"learning_rate": 5.124444444444445e-06,
|
| 673 |
+
"loss": 0.0303,
|
| 674 |
+
"step": 2700
|
| 675 |
+
},
|
| 676 |
+
{
|
| 677 |
+
"epoch": 4.86,
|
| 678 |
+
"learning_rate": 5.06888888888889e-06,
|
| 679 |
+
"loss": 0.0281,
|
| 680 |
+
"step": 2725
|
| 681 |
+
},
|
| 682 |
+
{
|
| 683 |
+
"epoch": 4.9,
|
| 684 |
+
"learning_rate": 5.013333333333333e-06,
|
| 685 |
+
"loss": 0.0291,
|
| 686 |
+
"step": 2750
|
| 687 |
+
},
|
| 688 |
+
{
|
| 689 |
+
"epoch": 4.95,
|
| 690 |
+
"learning_rate": 4.957777777777778e-06,
|
| 691 |
+
"loss": 0.0286,
|
| 692 |
+
"step": 2775
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"epoch": 4.99,
|
| 696 |
+
"learning_rate": 4.902222222222222e-06,
|
| 697 |
+
"loss": 0.0298,
|
| 698 |
+
"step": 2800
|
| 699 |
+
},
|
| 700 |
+
{
|
| 701 |
+
"epoch": 5.04,
|
| 702 |
+
"learning_rate": 4.846666666666667e-06,
|
| 703 |
+
"loss": 0.0238,
|
| 704 |
+
"step": 2825
|
| 705 |
+
},
|
| 706 |
+
{
|
| 707 |
+
"epoch": 5.08,
|
| 708 |
+
"learning_rate": 4.791111111111111e-06,
|
| 709 |
+
"loss": 0.0122,
|
| 710 |
+
"step": 2850
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"epoch": 5.12,
|
| 714 |
+
"learning_rate": 4.735555555555556e-06,
|
| 715 |
+
"loss": 0.0148,
|
| 716 |
+
"step": 2875
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 5.17,
|
| 720 |
+
"learning_rate": 4.680000000000001e-06,
|
| 721 |
+
"loss": 0.0123,
|
| 722 |
+
"step": 2900
|
| 723 |
+
},
|
| 724 |
+
{
|
| 725 |
+
"epoch": 5.21,
|
| 726 |
+
"learning_rate": 4.624444444444445e-06,
|
| 727 |
+
"loss": 0.0162,
|
| 728 |
+
"step": 2925
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 5.26,
|
| 732 |
+
"learning_rate": 4.568888888888889e-06,
|
| 733 |
+
"loss": 0.015,
|
| 734 |
+
"step": 2950
|
| 735 |
+
},
|
| 736 |
+
{
|
| 737 |
+
"epoch": 5.3,
|
| 738 |
+
"learning_rate": 4.513333333333333e-06,
|
| 739 |
+
"loss": 0.0161,
|
| 740 |
+
"step": 2975
|
| 741 |
+
},
|
| 742 |
+
{
|
| 743 |
+
"epoch": 5.35,
|
| 744 |
+
"learning_rate": 4.457777777777778e-06,
|
| 745 |
+
"loss": 0.0125,
|
| 746 |
+
"step": 3000
|
| 747 |
+
},
|
| 748 |
+
{
|
| 749 |
+
"epoch": 5.35,
|
| 750 |
+
"eval_loss": 0.2745016813278198,
|
| 751 |
+
"eval_runtime": 4396.7332,
|
| 752 |
+
"eval_samples_per_second": 2.374,
|
| 753 |
+
"eval_steps_per_second": 0.149,
|
| 754 |
+
"eval_wer": 0.4207764409383577,
|
| 755 |
+
"step": 3000
|
| 756 |
}
|
| 757 |
],
|
| 758 |
"max_steps": 5000,
|
| 759 |
"num_train_epochs": 9,
|
| 760 |
+
"total_flos": 3.18401342134272e+20,
|
| 761 |
"trial_name": null,
|
| 762 |
"trial_params": null
|
| 763 |
}
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 6173655480
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:188ae5e3d4f5fd4f3ea79612ae9050f69c98af92f2396885f41ca19d41d84d28
|
| 3 |
size 6173655480
|
runs/Jan28_23-19-38_gpu07.cyverse.org/events.out.tfevents.1674973221.gpu07.cyverse.org.79624.0
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 sha256:4c5e71c8a208596e260a1b1a53ca941ba4612a5a88ad5428422911ad98f8b3d5
|
| 3 |
+
size 24192
|