Default branch

f1e2ace651 · Merge pull request #801 from ngc92/ngc92/fix-test · Updated 2025-05-10 19:24:10 -04:00

Branches

da88cb11b0 · Merge pull request #803 from ngc92/ngc92/llama3-tied-weights · Updated 2025-06-26 13:03:39 -04:00    aparler

6
75

18298f3a40 · i misspelled reduced · Updated 2024-08-16 14:16:46 -04:00    aparler

21
0
Included

755458d0e1 · fix tokenizer omg · Updated 2024-08-13 13:24:15 -04:00    aparler

29
0
Included

0aa9175569 · still wip just putting things up for comment · Updated 2024-08-04 16:26:41 -04:00    aparler

82
1

51dd102328 · remove confusing comment · Updated 2024-07-30 16:01:43 -04:00    aparler

97
0
Included

5afc57064c · document the imports from zero.cuh to be explicit · Updated 2024-07-15 15:09:52 -04:00    aparler

133
0
Included

220125f39c · ok let's allow T < max_seq_len because our tests use it. make warning very prominent · Updated 2024-07-15 14:21:34 -04:00    aparler

139
0
Included

7b92930021 · move things around a bit, a few comments, and also update python script to correspond to C more · Updated 2024-07-01 13:20:46 -04:00    aparler

180
0
Included

7fd7717816 · fix the overfit single batch behavior to actually overfit batch, not microbatch, we do this more cleanly by simply resetting the dataloader every step · Updated 2024-07-01 13:00:40 -04:00    aparler

187
0
Included

7c283df9c4 · use ?= for OPENMPI_DIR so it can be overriden · Updated 2024-06-26 12:06:54 -04:00    aparler

213
0
Included

e12d090483 · resolve merge conflicts · Updated 2024-06-26 00:05:46 -04:00    aparler

216
0
Included

11e2fd1cf2 · separate out write_checkpoint and delete_checkpoint · Updated 2024-06-25 16:46:03 -04:00    aparler

250
0
Included

dbe06082ca · delete spurious check its fine · Updated 2024-06-20 22:39:40 -04:00    aparler

292
1

f19202c858 · small fixes on top of @chinthysl PR to kick off the feature branch · Updated 2024-06-20 16:13:27 -04:00    aparler

307
3

1a66151a80 · small fixes and readme · Updated 2024-06-18 21:03:36 -04:00    aparler

403
0
Included

14d9cd671e · resolve merge conflict in staate file between use_master_weights and DataLoader should_shuffle · Updated 2024-06-17 15:16:42 -04:00    aparler

354
0
Included

cbe82e783a · consolidate memory by deleting the grad activation struct, move some of the last pieces of memory to the forward pass · Updated 2024-06-13 20:54:58 -04:00    aparler

423
0
Included

43a903e1b4 · revert old printf0, minor touchups and comments, add assert · Updated 2024-06-13 19:18:22 -04:00    aparler

425
0
Included

ee6b3c9325 · take out Async copies and memsets · Updated 2024-06-08 12:17:02 -04:00    aparler

479
0
Included

094988738b · adjust the comment on extern deviceProp to be more descriptive · Updated 2024-06-03 18:57:00 -04:00    aparler

506
0
Included