分支列表

默认分支

f1e2ace651 · Merge pull request #801 from ngc92/ngc92/fix-test · 最后更新于 2025-05-10 23:24:10 +00:00

分支列表

90a7bfecf1 · move fused classifier · 最后更新于 2024-06-03 22:35:20 +00:00    frostbyte_neo

509
0
已包含

2186cd6fe7 · move attention · 最后更新于 2024-06-03 22:19:47 +00:00    frostbyte_neo

511
0
已包含

d8d55e0a08 · move gelu to llmc · 最后更新于 2024-06-03 21:57:02 +00:00    frostbyte_neo

513
0
已包含

49b9776e50 · also move the launchers · 最后更新于 2024-06-03 21:37:18 +00:00    frostbyte_neo

514
0
已包含

001ad78fe1 · resolve deviceProp and cudaCheck dupes · 最后更新于 2024-06-03 19:54:32 +00:00    frostbyte_neo

529
0
已包含

6434009298 · attempt to remove PATH_SEP from Makefile · 最后更新于 2024-06-03 15:30:45 +00:00    frostbyte_neo

534
0
已包含

22d1607f6d · delete these files, they are now in scripts/ folder and better · 最后更新于 2024-05-30 20:11:59 +00:00    frostbyte_neo

581
0
已包含

5959851d2d · small comments for pyrun124M.sh · 最后更新于 2024-05-30 14:47:43 +00:00    frostbyte_neo

596
0
已包含

24a8c03cb6 · Merge branch 'master' into feature/resume · 最后更新于 2024-05-27 20:30:30 +00:00    frostbyte_neo

621
0
已包含

05fb642ae1 · attempt fix zero 1 · 最后更新于 2024-05-25 23:49:16 +00:00    frostbyte_neo

648
0
已包含

a13b2bd10b · fix bug check null · 最后更新于 2024-05-25 13:50:56 +00:00    frostbyte_neo

656
0
已包含

9f08882051 · add weight decay, but only for 2D tensors, as done in GPT series and in general too. this forces us to break up our adamw kernel again into one call per tensor, so there is a small throughput hit, of about 0.5% for me. but we have to break up this kernel in near future anyway · 最后更新于 2024-05-25 00:14:10 +00:00    frostbyte_neo

674
0
已包含

1f91bfc442 · fix small bug on eval logging · 最后更新于 2024-05-23 18:19:25 +00:00    frostbyte_neo

707
0
已包含

da59861d38 · make hellaswag optional eval yay · 最后更新于 2024-05-22 22:42:37 +00:00    frostbyte_neo

716
0
已包含

099d30f814 · add a super small crappy glob for windows that only matches a single unique file. this will make CI happy but we can't train on sharded data on windows until this is improved · 最后更新于 2024-05-22 13:08:18 +00:00    frostbyte_neo

720
0
已包含

54ccbd300c · docs on master-breaking changes around dataset file representation · 最后更新于 2024-05-21 18:31:19 +00:00    frostbyte_neo

727
0
已包含

8d55c4a6ac · fix a bug, we have to be careful to make sure we evaluate loss at the token just before the first completion token, because that is the prediction for the first token, and its accuracy · 最后更新于 2024-05-19 15:45:28 +00:00    frostbyte_neo

762
0
已包含

d7581fc542 · make recompute be an int instead of bool, so we can strengthen it over time just like ZeRO stages, as we recompute more and more of the model in the future possibly. and make it default on because it is awesome · 最后更新于 2024-05-16 19:07:39 +00:00    frostbyte_neo

793
0
已包含

8eb3a432d9 · revert all streams and synchronization events, we'll bring them back but only one at a time and very very carefully, because they cause really subtle bugs and issues · 最后更新于 2024-05-15 21:09:46 +00:00    frostbyte_neo

800
0
已包含

3048a3a780 · oops have to use labels= · 最后更新于 2024-05-06 18:04:35 +00:00    frostbyte_neo

891
2