Commit
•
3f73ee9
1
Parent(s):
d79a927
Training in progress, step 150000
Browse files- last-checkpoint/optimizer.pt +1 -1
- last-checkpoint/pytorch_model.bin +1 -1
- last-checkpoint/trainer_state.json +51 -51
- last-checkpoint/training_args.bin +1 -1
- pytorch_model.bin +1 -1
- runs/Feb20_15-56-09_t1v-n-9f780742-w-0/events.out.tfevents.1676908745.t1v-n-9f780742-w-0.3298479.0 +2 -2
- runs/Feb24_00-12-30_t1v-n-9f780742-w-0/1677197576.9197693/events.out.tfevents.1677197576.t1v-n-9f780742-w-0.3044920.1 +3 -0
- runs/Feb24_00-12-30_t1v-n-9f780742-w-0/events.out.tfevents.1677197576.t1v-n-9f780742-w-0.3044920.0 +3 -0
- runs/Feb24_00-19-36_t1v-n-9f780742-w-0/1677197998.1740472/events.out.tfevents.1677197998.t1v-n-9f780742-w-0.3062414.1 +3 -0
- runs/Feb24_00-19-36_t1v-n-9f780742-w-0/events.out.tfevents.1677197998.t1v-n-9f780742-w-0.3062414.0 +3 -0
- runs/Feb24_02-46-37_t1v-n-9f780742-w-0/1677206822.561554/events.out.tfevents.1677206822.t1v-n-9f780742-w-0.3088188.1 +3 -0
- runs/Feb24_02-46-37_t1v-n-9f780742-w-0/events.out.tfevents.1677206822.t1v-n-9f780742-w-0.3088188.0 +3 -0
- runs/Feb25_19-20-30_t1v-n-9f780742-w-0/1677352852.8462265/events.out.tfevents.1677352852.t1v-n-9f780742-w-0.3235622.1 +3 -0
- runs/Feb25_19-20-30_t1v-n-9f780742-w-0/events.out.tfevents.1677352852.t1v-n-9f780742-w-0.3235622.0 +3 -0
- runs/Feb27_19-46-56_t1v-n-9f780742-w-0/1677527238.5880394/events.out.tfevents.1677527238.t1v-n-9f780742-w-0.2687388.1 +3 -0
- runs/Feb27_19-46-56_t1v-n-9f780742-w-0/events.out.tfevents.1677527238.t1v-n-9f780742-w-0.2687388.0 +3 -0
- training_args.bin +1 -1
last-checkpoint/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 885325017
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c485e39025b832d2bd3bd464214f232ea6b3f0a5283d0f592f5b95f7afffe93d
|
| 3 |
size 885325017
|
last-checkpoint/pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 442675755
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6bc0d6bce669ef3b9a37ef6ee5196e838e490074197c122478da0bc35a3e8702
|
| 3 |
size 442675755
|
last-checkpoint/trainer_state.json
CHANGED
|
@@ -638,7 +638,7 @@
|
|
| 638 |
{
|
| 639 |
"epoch": 0.1,
|
| 640 |
"learning_rate": 9.923399386589933e-05,
|
| 641 |
-
"loss": 0.
|
| 642 |
"step": 103000
|
| 643 |
},
|
| 644 |
{
|
|
@@ -650,285 +650,285 @@
|
|
| 650 |
{
|
| 651 |
"epoch": 0.1,
|
| 652 |
"learning_rate": 9.917525374361912e-05,
|
| 653 |
-
"loss": 1.
|
| 654 |
"step": 105000
|
| 655 |
},
|
| 656 |
{
|
| 657 |
"epoch": 0.11,
|
| 658 |
"learning_rate": 9.914507686137019e-05,
|
| 659 |
-
"loss": 0.
|
| 660 |
"step": 106000
|
| 661 |
},
|
| 662 |
{
|
| 663 |
"epoch": 0.11,
|
| 664 |
"learning_rate": 9.911436253643445e-05,
|
| 665 |
-
"loss": 0.
|
| 666 |
"step": 107000
|
| 667 |
},
|
| 668 |
{
|
| 669 |
"epoch": 0.11,
|
| 670 |
"learning_rate": 9.90831111046988e-05,
|
| 671 |
-
"loss": 1.
|
| 672 |
"step": 108000
|
| 673 |
},
|
| 674 |
{
|
| 675 |
"epoch": 0.11,
|
| 676 |
"learning_rate": 9.905132290792394e-05,
|
| 677 |
-
"loss": 1.
|
| 678 |
"step": 109000
|
| 679 |
},
|
| 680 |
{
|
| 681 |
"epoch": 0.11,
|
| 682 |
"learning_rate": 9.901899829374047e-05,
|
| 683 |
-
"loss": 1.
|
| 684 |
"step": 110000
|
| 685 |
},
|
| 686 |
{
|
| 687 |
"epoch": 0.11,
|
| 688 |
"learning_rate": 9.89861376156452e-05,
|
| 689 |
-
"loss": 1.
|
| 690 |
"step": 111000
|
| 691 |
},
|
| 692 |
{
|
| 693 |
"epoch": 0.11,
|
| 694 |
"learning_rate": 9.895274123299723e-05,
|
| 695 |
-
"loss": 1.
|
| 696 |
"step": 112000
|
| 697 |
},
|
| 698 |
{
|
| 699 |
"epoch": 0.11,
|
| 700 |
"learning_rate": 9.891880951101407e-05,
|
| 701 |
-
"loss": 1.
|
| 702 |
"step": 113000
|
| 703 |
},
|
| 704 |
{
|
| 705 |
"epoch": 0.11,
|
| 706 |
"learning_rate": 9.888434282076758e-05,
|
| 707 |
-
"loss": 0.
|
| 708 |
"step": 114000
|
| 709 |
},
|
| 710 |
{
|
| 711 |
"epoch": 0.12,
|
| 712 |
"learning_rate": 9.884934153917997e-05,
|
| 713 |
-
"loss": 1.
|
| 714 |
"step": 115000
|
| 715 |
},
|
| 716 |
{
|
| 717 |
"epoch": 0.12,
|
| 718 |
"learning_rate": 9.881380604901964e-05,
|
| 719 |
-
"loss": 1.
|
| 720 |
"step": 116000
|
| 721 |
},
|
| 722 |
{
|
| 723 |
"epoch": 0.12,
|
| 724 |
"learning_rate": 9.877773673889701e-05,
|
| 725 |
-
"loss": 0.
|
| 726 |
"step": 117000
|
| 727 |
},
|
| 728 |
{
|
| 729 |
"epoch": 0.12,
|
| 730 |
"learning_rate": 9.87411340032603e-05,
|
| 731 |
-
"loss": 0.
|
| 732 |
"step": 118000
|
| 733 |
},
|
| 734 |
{
|
| 735 |
"epoch": 0.12,
|
| 736 |
"learning_rate": 9.870399824239117e-05,
|
| 737 |
-
"loss": 0.
|
| 738 |
"step": 119000
|
| 739 |
},
|
| 740 |
{
|
| 741 |
"epoch": 0.12,
|
| 742 |
"learning_rate": 9.86663298624003e-05,
|
| 743 |
-
"loss": 0.
|
| 744 |
"step": 120000
|
| 745 |
},
|
| 746 |
{
|
| 747 |
"epoch": 0.12,
|
| 748 |
"learning_rate": 9.862812927522309e-05,
|
| 749 |
-
"loss": 1.
|
| 750 |
"step": 121000
|
| 751 |
},
|
| 752 |
{
|
| 753 |
"epoch": 0.12,
|
| 754 |
"learning_rate": 9.858939689861506e-05,
|
| 755 |
-
"loss": 0.
|
| 756 |
"step": 122000
|
| 757 |
},
|
| 758 |
{
|
| 759 |
"epoch": 0.12,
|
| 760 |
"learning_rate": 9.855013315614725e-05,
|
| 761 |
-
"loss": 0.
|
| 762 |
"step": 123000
|
| 763 |
},
|
| 764 |
{
|
| 765 |
"epoch": 0.12,
|
| 766 |
"learning_rate": 9.851033847720166e-05,
|
| 767 |
-
"loss": 0.
|
| 768 |
"step": 124000
|
| 769 |
},
|
| 770 |
{
|
| 771 |
"epoch": 0.12,
|
| 772 |
"learning_rate": 9.847001329696653e-05,
|
| 773 |
-
"loss": 0.
|
| 774 |
"step": 125000
|
| 775 |
},
|
| 776 |
{
|
| 777 |
"epoch": 0.13,
|
| 778 |
"learning_rate": 9.842915805643155e-05,
|
| 779 |
-
"loss": 0.
|
| 780 |
"step": 126000
|
| 781 |
},
|
| 782 |
{
|
| 783 |
"epoch": 0.13,
|
| 784 |
"learning_rate": 9.838777320238312e-05,
|
| 785 |
-
"loss": 0.
|
| 786 |
"step": 127000
|
| 787 |
},
|
| 788 |
{
|
| 789 |
"epoch": 0.13,
|
| 790 |
"learning_rate": 9.834585918739936e-05,
|
| 791 |
-
"loss": 1.
|
| 792 |
"step": 128000
|
| 793 |
},
|
| 794 |
{
|
| 795 |
"epoch": 0.13,
|
| 796 |
"learning_rate": 9.830341646984521e-05,
|
| 797 |
-
"loss": 0.
|
| 798 |
"step": 129000
|
| 799 |
},
|
| 800 |
{
|
| 801 |
"epoch": 0.13,
|
| 802 |
"learning_rate": 9.826044551386744e-05,
|
| 803 |
-
"loss": 0.
|
| 804 |
"step": 130000
|
| 805 |
},
|
| 806 |
{
|
| 807 |
"epoch": 0.13,
|
| 808 |
"learning_rate": 9.821694678938953e-05,
|
| 809 |
-
"loss": 1.
|
| 810 |
"step": 131000
|
| 811 |
},
|
| 812 |
{
|
| 813 |
"epoch": 0.13,
|
| 814 |
"learning_rate": 9.817292077210659e-05,
|
| 815 |
-
"loss": 1.
|
| 816 |
"step": 132000
|
| 817 |
},
|
| 818 |
{
|
| 819 |
"epoch": 0.13,
|
| 820 |
"learning_rate": 9.812836794348004e-05,
|
| 821 |
-
"loss": 0.
|
| 822 |
"step": 133000
|
| 823 |
},
|
| 824 |
{
|
| 825 |
"epoch": 0.13,
|
| 826 |
"learning_rate": 9.808328879073251e-05,
|
| 827 |
-
"loss": 1.
|
| 828 |
"step": 134000
|
| 829 |
},
|
| 830 |
{
|
| 831 |
"epoch": 0.14,
|
| 832 |
"learning_rate": 9.803768380684242e-05,
|
| 833 |
-
"loss": 1.
|
| 834 |
"step": 135000
|
| 835 |
},
|
| 836 |
{
|
| 837 |
"epoch": 0.14,
|
| 838 |
"learning_rate": 9.799155349053851e-05,
|
| 839 |
-
"loss": 1.
|
| 840 |
"step": 136000
|
| 841 |
},
|
| 842 |
{
|
| 843 |
"epoch": 0.14,
|
| 844 |
"learning_rate": 9.794489834629455e-05,
|
| 845 |
-
"loss": 0.
|
| 846 |
"step": 137000
|
| 847 |
},
|
| 848 |
{
|
| 849 |
"epoch": 0.14,
|
| 850 |
"learning_rate": 9.789771888432375e-05,
|
| 851 |
-
"loss": 0.
|
| 852 |
"step": 138000
|
| 853 |
},
|
| 854 |
{
|
| 855 |
"epoch": 0.14,
|
| 856 |
"learning_rate": 9.785001562057309e-05,
|
| 857 |
-
"loss": 0.
|
| 858 |
"step": 139000
|
| 859 |
},
|
| 860 |
{
|
| 861 |
"epoch": 0.14,
|
| 862 |
"learning_rate": 9.780178907671789e-05,
|
| 863 |
-
"loss": 0.
|
| 864 |
"step": 140000
|
| 865 |
},
|
| 866 |
{
|
| 867 |
"epoch": 0.14,
|
| 868 |
"learning_rate": 9.775303978015585e-05,
|
| 869 |
-
"loss": 0.
|
| 870 |
"step": 141000
|
| 871 |
},
|
| 872 |
{
|
| 873 |
"epoch": 0.14,
|
| 874 |
"learning_rate": 9.77037682640015e-05,
|
| 875 |
-
"loss": 0.
|
| 876 |
"step": 142000
|
| 877 |
},
|
| 878 |
{
|
| 879 |
"epoch": 0.14,
|
| 880 |
"learning_rate": 9.765397506708023e-05,
|
| 881 |
-
"loss": 0.
|
| 882 |
"step": 143000
|
| 883 |
},
|
| 884 |
{
|
| 885 |
"epoch": 0.14,
|
| 886 |
"learning_rate": 9.760366073392246e-05,
|
| 887 |
-
"loss": 0.
|
| 888 |
"step": 144000
|
| 889 |
},
|
| 890 |
{
|
| 891 |
"epoch": 0.14,
|
| 892 |
"learning_rate": 9.755282581475769e-05,
|
| 893 |
-
"loss": 0.
|
| 894 |
"step": 145000
|
| 895 |
},
|
| 896 |
{
|
| 897 |
"epoch": 0.15,
|
| 898 |
"learning_rate": 9.750147086550844e-05,
|
| 899 |
-
"loss": 0.
|
| 900 |
"step": 146000
|
| 901 |
},
|
| 902 |
{
|
| 903 |
"epoch": 0.15,
|
| 904 |
"learning_rate": 9.744959644778422e-05,
|
| 905 |
-
"loss": 0.
|
| 906 |
"step": 147000
|
| 907 |
},
|
| 908 |
{
|
| 909 |
"epoch": 0.15,
|
| 910 |
"learning_rate": 9.739720312887535e-05,
|
| 911 |
-
"loss": 0.
|
| 912 |
"step": 148000
|
| 913 |
},
|
| 914 |
{
|
| 915 |
"epoch": 0.15,
|
| 916 |
"learning_rate": 9.734429148174675e-05,
|
| 917 |
-
"loss": 1.
|
| 918 |
"step": 149000
|
| 919 |
},
|
| 920 |
{
|
| 921 |
"epoch": 0.15,
|
| 922 |
"learning_rate": 9.729086208503174e-05,
|
| 923 |
-
"loss": 1.
|
| 924 |
"step": 150000
|
| 925 |
},
|
| 926 |
{
|
| 927 |
"epoch": 0.15,
|
| 928 |
-
"eval_loss": 0.
|
| 929 |
-
"eval_runtime":
|
| 930 |
-
"eval_samples_per_second":
|
| 931 |
-
"eval_steps_per_second": 1.
|
| 932 |
"step": 150000
|
| 933 |
}
|
| 934 |
],
|
|
|
|
| 638 |
{
|
| 639 |
"epoch": 0.1,
|
| 640 |
"learning_rate": 9.923399386589933e-05,
|
| 641 |
+
"loss": 0.6525,
|
| 642 |
"step": 103000
|
| 643 |
},
|
| 644 |
{
|
|
|
|
| 650 |
{
|
| 651 |
"epoch": 0.1,
|
| 652 |
"learning_rate": 9.917525374361912e-05,
|
| 653 |
+
"loss": 1.0445,
|
| 654 |
"step": 105000
|
| 655 |
},
|
| 656 |
{
|
| 657 |
"epoch": 0.11,
|
| 658 |
"learning_rate": 9.914507686137019e-05,
|
| 659 |
+
"loss": 0.9613,
|
| 660 |
"step": 106000
|
| 661 |
},
|
| 662 |
{
|
| 663 |
"epoch": 0.11,
|
| 664 |
"learning_rate": 9.911436253643445e-05,
|
| 665 |
+
"loss": 0.9396,
|
| 666 |
"step": 107000
|
| 667 |
},
|
| 668 |
{
|
| 669 |
"epoch": 0.11,
|
| 670 |
"learning_rate": 9.90831111046988e-05,
|
| 671 |
+
"loss": 1.0517,
|
| 672 |
"step": 108000
|
| 673 |
},
|
| 674 |
{
|
| 675 |
"epoch": 0.11,
|
| 676 |
"learning_rate": 9.905132290792394e-05,
|
| 677 |
+
"loss": 1.0215,
|
| 678 |
"step": 109000
|
| 679 |
},
|
| 680 |
{
|
| 681 |
"epoch": 0.11,
|
| 682 |
"learning_rate": 9.901899829374047e-05,
|
| 683 |
+
"loss": 1.0032,
|
| 684 |
"step": 110000
|
| 685 |
},
|
| 686 |
{
|
| 687 |
"epoch": 0.11,
|
| 688 |
"learning_rate": 9.89861376156452e-05,
|
| 689 |
+
"loss": 1.0618,
|
| 690 |
"step": 111000
|
| 691 |
},
|
| 692 |
{
|
| 693 |
"epoch": 0.11,
|
| 694 |
"learning_rate": 9.895274123299723e-05,
|
| 695 |
+
"loss": 1.0749,
|
| 696 |
"step": 112000
|
| 697 |
},
|
| 698 |
{
|
| 699 |
"epoch": 0.11,
|
| 700 |
"learning_rate": 9.891880951101407e-05,
|
| 701 |
+
"loss": 1.0837,
|
| 702 |
"step": 113000
|
| 703 |
},
|
| 704 |
{
|
| 705 |
"epoch": 0.11,
|
| 706 |
"learning_rate": 9.888434282076758e-05,
|
| 707 |
+
"loss": 0.9592,
|
| 708 |
"step": 114000
|
| 709 |
},
|
| 710 |
{
|
| 711 |
"epoch": 0.12,
|
| 712 |
"learning_rate": 9.884934153917997e-05,
|
| 713 |
+
"loss": 1.0298,
|
| 714 |
"step": 115000
|
| 715 |
},
|
| 716 |
{
|
| 717 |
"epoch": 0.12,
|
| 718 |
"learning_rate": 9.881380604901964e-05,
|
| 719 |
+
"loss": 1.0163,
|
| 720 |
"step": 116000
|
| 721 |
},
|
| 722 |
{
|
| 723 |
"epoch": 0.12,
|
| 724 |
"learning_rate": 9.877773673889701e-05,
|
| 725 |
+
"loss": 0.8,
|
| 726 |
"step": 117000
|
| 727 |
},
|
| 728 |
{
|
| 729 |
"epoch": 0.12,
|
| 730 |
"learning_rate": 9.87411340032603e-05,
|
| 731 |
+
"loss": 0.9133,
|
| 732 |
"step": 118000
|
| 733 |
},
|
| 734 |
{
|
| 735 |
"epoch": 0.12,
|
| 736 |
"learning_rate": 9.870399824239117e-05,
|
| 737 |
+
"loss": 0.9643,
|
| 738 |
"step": 119000
|
| 739 |
},
|
| 740 |
{
|
| 741 |
"epoch": 0.12,
|
| 742 |
"learning_rate": 9.86663298624003e-05,
|
| 743 |
+
"loss": 0.9112,
|
| 744 |
"step": 120000
|
| 745 |
},
|
| 746 |
{
|
| 747 |
"epoch": 0.12,
|
| 748 |
"learning_rate": 9.862812927522309e-05,
|
| 749 |
+
"loss": 1.023,
|
| 750 |
"step": 121000
|
| 751 |
},
|
| 752 |
{
|
| 753 |
"epoch": 0.12,
|
| 754 |
"learning_rate": 9.858939689861506e-05,
|
| 755 |
+
"loss": 0.9952,
|
| 756 |
"step": 122000
|
| 757 |
},
|
| 758 |
{
|
| 759 |
"epoch": 0.12,
|
| 760 |
"learning_rate": 9.855013315614725e-05,
|
| 761 |
+
"loss": 0.984,
|
| 762 |
"step": 123000
|
| 763 |
},
|
| 764 |
{
|
| 765 |
"epoch": 0.12,
|
| 766 |
"learning_rate": 9.851033847720166e-05,
|
| 767 |
+
"loss": 0.9928,
|
| 768 |
"step": 124000
|
| 769 |
},
|
| 770 |
{
|
| 771 |
"epoch": 0.12,
|
| 772 |
"learning_rate": 9.847001329696653e-05,
|
| 773 |
+
"loss": 0.9131,
|
| 774 |
"step": 125000
|
| 775 |
},
|
| 776 |
{
|
| 777 |
"epoch": 0.13,
|
| 778 |
"learning_rate": 9.842915805643155e-05,
|
| 779 |
+
"loss": 0.9251,
|
| 780 |
"step": 126000
|
| 781 |
},
|
| 782 |
{
|
| 783 |
"epoch": 0.13,
|
| 784 |
"learning_rate": 9.838777320238312e-05,
|
| 785 |
+
"loss": 0.9687,
|
| 786 |
"step": 127000
|
| 787 |
},
|
| 788 |
{
|
| 789 |
"epoch": 0.13,
|
| 790 |
"learning_rate": 9.834585918739936e-05,
|
| 791 |
+
"loss": 1.0081,
|
| 792 |
"step": 128000
|
| 793 |
},
|
| 794 |
{
|
| 795 |
"epoch": 0.13,
|
| 796 |
"learning_rate": 9.830341646984521e-05,
|
| 797 |
+
"loss": 0.9352,
|
| 798 |
"step": 129000
|
| 799 |
},
|
| 800 |
{
|
| 801 |
"epoch": 0.13,
|
| 802 |
"learning_rate": 9.826044551386744e-05,
|
| 803 |
+
"loss": 0.9909,
|
| 804 |
"step": 130000
|
| 805 |
},
|
| 806 |
{
|
| 807 |
"epoch": 0.13,
|
| 808 |
"learning_rate": 9.821694678938953e-05,
|
| 809 |
+
"loss": 1.0289,
|
| 810 |
"step": 131000
|
| 811 |
},
|
| 812 |
{
|
| 813 |
"epoch": 0.13,
|
| 814 |
"learning_rate": 9.817292077210659e-05,
|
| 815 |
+
"loss": 1.0766,
|
| 816 |
"step": 132000
|
| 817 |
},
|
| 818 |
{
|
| 819 |
"epoch": 0.13,
|
| 820 |
"learning_rate": 9.812836794348004e-05,
|
| 821 |
+
"loss": 0.9957,
|
| 822 |
"step": 133000
|
| 823 |
},
|
| 824 |
{
|
| 825 |
"epoch": 0.13,
|
| 826 |
"learning_rate": 9.808328879073251e-05,
|
| 827 |
+
"loss": 1.0035,
|
| 828 |
"step": 134000
|
| 829 |
},
|
| 830 |
{
|
| 831 |
"epoch": 0.14,
|
| 832 |
"learning_rate": 9.803768380684242e-05,
|
| 833 |
+
"loss": 1.2201,
|
| 834 |
"step": 135000
|
| 835 |
},
|
| 836 |
{
|
| 837 |
"epoch": 0.14,
|
| 838 |
"learning_rate": 9.799155349053851e-05,
|
| 839 |
+
"loss": 1.0479,
|
| 840 |
"step": 136000
|
| 841 |
},
|
| 842 |
{
|
| 843 |
"epoch": 0.14,
|
| 844 |
"learning_rate": 9.794489834629455e-05,
|
| 845 |
+
"loss": 0.916,
|
| 846 |
"step": 137000
|
| 847 |
},
|
| 848 |
{
|
| 849 |
"epoch": 0.14,
|
| 850 |
"learning_rate": 9.789771888432375e-05,
|
| 851 |
+
"loss": 0.9751,
|
| 852 |
"step": 138000
|
| 853 |
},
|
| 854 |
{
|
| 855 |
"epoch": 0.14,
|
| 856 |
"learning_rate": 9.785001562057309e-05,
|
| 857 |
+
"loss": 0.9485,
|
| 858 |
"step": 139000
|
| 859 |
},
|
| 860 |
{
|
| 861 |
"epoch": 0.14,
|
| 862 |
"learning_rate": 9.780178907671789e-05,
|
| 863 |
+
"loss": 0.9192,
|
| 864 |
"step": 140000
|
| 865 |
},
|
| 866 |
{
|
| 867 |
"epoch": 0.14,
|
| 868 |
"learning_rate": 9.775303978015585e-05,
|
| 869 |
+
"loss": 0.9877,
|
| 870 |
"step": 141000
|
| 871 |
},
|
| 872 |
{
|
| 873 |
"epoch": 0.14,
|
| 874 |
"learning_rate": 9.77037682640015e-05,
|
| 875 |
+
"loss": 0.989,
|
| 876 |
"step": 142000
|
| 877 |
},
|
| 878 |
{
|
| 879 |
"epoch": 0.14,
|
| 880 |
"learning_rate": 9.765397506708023e-05,
|
| 881 |
+
"loss": 0.9245,
|
| 882 |
"step": 143000
|
| 883 |
},
|
| 884 |
{
|
| 885 |
"epoch": 0.14,
|
| 886 |
"learning_rate": 9.760366073392246e-05,
|
| 887 |
+
"loss": 0.9223,
|
| 888 |
"step": 144000
|
| 889 |
},
|
| 890 |
{
|
| 891 |
"epoch": 0.14,
|
| 892 |
"learning_rate": 9.755282581475769e-05,
|
| 893 |
+
"loss": 0.9474,
|
| 894 |
"step": 145000
|
| 895 |
},
|
| 896 |
{
|
| 897 |
"epoch": 0.15,
|
| 898 |
"learning_rate": 9.750147086550844e-05,
|
| 899 |
+
"loss": 0.7895,
|
| 900 |
"step": 146000
|
| 901 |
},
|
| 902 |
{
|
| 903 |
"epoch": 0.15,
|
| 904 |
"learning_rate": 9.744959644778422e-05,
|
| 905 |
+
"loss": 0.6734,
|
| 906 |
"step": 147000
|
| 907 |
},
|
| 908 |
{
|
| 909 |
"epoch": 0.15,
|
| 910 |
"learning_rate": 9.739720312887535e-05,
|
| 911 |
+
"loss": 0.6778,
|
| 912 |
"step": 148000
|
| 913 |
},
|
| 914 |
{
|
| 915 |
"epoch": 0.15,
|
| 916 |
"learning_rate": 9.734429148174675e-05,
|
| 917 |
+
"loss": 1.0325,
|
| 918 |
"step": 149000
|
| 919 |
},
|
| 920 |
{
|
| 921 |
"epoch": 0.15,
|
| 922 |
"learning_rate": 9.729086208503174e-05,
|
| 923 |
+
"loss": 1.017,
|
| 924 |
"step": 150000
|
| 925 |
},
|
| 926 |
{
|
| 927 |
"epoch": 0.15,
|
| 928 |
+
"eval_loss": 0.791193425655365,
|
| 929 |
+
"eval_runtime": 27.8851,
|
| 930 |
+
"eval_samples_per_second": 179.307,
|
| 931 |
+
"eval_steps_per_second": 1.434,
|
| 932 |
"step": 150000
|
| 933 |
}
|
| 934 |
],
|
last-checkpoint/training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c6ec36a3db40f8f1ddaac5209f15ca1b14671366820c356e3429fb92e12f3c8
|
| 3 |
size 3503
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 442675755
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6bc0d6bce669ef3b9a37ef6ee5196e838e490074197c122478da0bc35a3e8702
|
| 3 |
size 442675755
|
runs/Feb20_15-56-09_t1v-n-9f780742-w-0/events.out.tfevents.1676908745.t1v-n-9f780742-w-0.3298479.0
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 sha256:54e06d4f36aa91c889360b5ae94a88b9ec9ece094ed9e243da1d8c15cdf4f251
|
| 3 |
+
size 13907
|
runs/Feb24_00-12-30_t1v-n-9f780742-w-0/1677197576.9197693/events.out.tfevents.1677197576.t1v-n-9f780742-w-0.3044920.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5a5fbd6e886b0c4dc4c944fc5ef5df06817655069ed402a843d6e82683f505c9
|
| 3 |
+
size 5494
|
runs/Feb24_00-12-30_t1v-n-9f780742-w-0/events.out.tfevents.1677197576.t1v-n-9f780742-w-0.3044920.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4259b30c6b0189c21e1f311b8a8317502cfb9427519242ed75c126f270891573
|
| 3 |
+
size 4191
|
runs/Feb24_00-19-36_t1v-n-9f780742-w-0/1677197998.1740472/events.out.tfevents.1677197998.t1v-n-9f780742-w-0.3062414.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:22336f6cbbbee2fac23968c04cc7ae7ee82056968dd22e2389cdd293772f96e8
|
| 3 |
+
size 5494
|
runs/Feb24_00-19-36_t1v-n-9f780742-w-0/events.out.tfevents.1677197998.t1v-n-9f780742-w-0.3062414.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed9e768531b20105cac38e46eaa3f23b96f7e74bd097e6264fe8b8fa85c6b808
|
| 3 |
+
size 4191
|
runs/Feb24_02-46-37_t1v-n-9f780742-w-0/1677206822.561554/events.out.tfevents.1677206822.t1v-n-9f780742-w-0.3088188.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7b08653912ee437c10feae38a9b02a6a7fefd406f0733910788a5789c5f275c6
|
| 3 |
+
size 5494
|
runs/Feb24_02-46-37_t1v-n-9f780742-w-0/events.out.tfevents.1677206822.t1v-n-9f780742-w-0.3088188.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:517b84445e0e1a867254ad82f284b566d3eb43350bdf62570c192a250de1decd
|
| 3 |
+
size 4191
|
runs/Feb25_19-20-30_t1v-n-9f780742-w-0/1677352852.8462265/events.out.tfevents.1677352852.t1v-n-9f780742-w-0.3235622.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:37c186d49eb69d4077969e00cc7a10fdd645eacab5bf553597a602349c63d45d
|
| 3 |
+
size 5494
|
runs/Feb25_19-20-30_t1v-n-9f780742-w-0/events.out.tfevents.1677352852.t1v-n-9f780742-w-0.3235622.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a8d91c3fb78aff0688ef117263ed04672aaf11c076152935b41b6c1c0db64896
|
| 3 |
+
size 3831
|
runs/Feb27_19-46-56_t1v-n-9f780742-w-0/1677527238.5880394/events.out.tfevents.1677527238.t1v-n-9f780742-w-0.2687388.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:61770893ad58894535db5b4cc776137aca01a3f871d3a5c3d812a443f7f3e655
|
| 3 |
+
size 5494
|
runs/Feb27_19-46-56_t1v-n-9f780742-w-0/events.out.tfevents.1677527238.t1v-n-9f780742-w-0.2687388.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:53cdd5d73ce64e8fdcfd9d8faae04f0a4434a994fe919e51e7a8311600e68ab6
|
| 3 |
+
size 12107
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c6ec36a3db40f8f1ddaac5209f15ca1b14671366820c356e3429fb92e12f3c8
|
| 3 |
size 3503
|