Commit Graph
28 Commits
Author SHA1 Message Date
lb203 429391840f Merge pull request #121 from sysuyy/add_multi_node_script
[feat]: use accelerate on multi-node

Former-commit-id: 67b01dc1d934d3d86d7ecca00c5dd1a11fd4d22d
2024-03-12 14:05:08 +08:00
lb203 e0dac4fb40 Create train_256.sh
Former-commit-id: bf62151617feb49663f8dfe0c78aa01f41e62092
2024-03-12 12:55:41 +08:00
sysuyy 330e46e9e8 [feat]: use accelerate on multi-node
Former-commit-id: 0b12fc31f1d2a63f9da38821174ab63d9b3a2d72
2024-03-11 14:45:23 +00:00
touale 76e172e08c [fix]: correct attention_mode to attention-mode
Former-commit-id: 9957e2eebb2166b02600c6052fb54908f5d12bad
2024-03-11 22:14:34 +08:00
LinB203 a7165b506e clean script
Former-commit-id: 02c61f63a8539a31516b92f5bb7c2d2e428ca7d7
2024-03-10 18:34:37 +08:00
LinB203 b57709108a train 1080p video
Former-commit-id: 74625142ef892f9e7fa459b2f98f2f7ae5b36d6a
2024-03-10 18:08:23 +08:00
LinB203 6dae0b1c0c deepspeed
Former-commit-id: 6179a4b0e7d06caed4201a27679abc64615febe6
2024-03-10 15:08:12 +08:00
sennnnn f3274235d8 Support deepspeed zero2 and zero2_offload for latte training.
Former-commit-id: c87e95dd357e6fa6d6fb2dcfee781bc49c039e13
2024-03-10 11:27:33 +08:00
sennnnn ee3810a585 Decouple mixed_precision and gradient_accumulation to command args.
Former-commit-id: 903d473d1432eb492a87efe184d03439f643b92d
2024-03-10 11:24:39 +08:00
LinB203 6df13b07bb flash-attn train and sample
Former-commit-id: e336565bd5ff2aaf4d0b4ad5f42732e89cd216b3
2024-03-09 16:30:09 +08:00
LinB203 b0a9e72008 extract script
Former-commit-id: 4139f2f68ae09695302ce49afdb7318df83cabfd
2024-03-09 01:03:05 +08:00
LinB203 21fad0d34a update script
Former-commit-id: 43f539b68af150c45040d79918a84de1dce47c33
2024-03-08 17:21:06 +08:00
lb203 e37dae81ac Delete scripts/train_vqvae.sh
Former-commit-id: 952a4ce3a3b0da159d599ff3a0a052265f48c2dd
2024-03-08 11:21:00 +08:00
Liu Hanchen 2448aa7c8c commit
Former-commit-id: 81359cf67248b6c594ff79f121a8a38558c1cc70
2024-03-08 01:04:16 +08:00
LinB203 9c8b50a29a train script
Former-commit-id: 3fd1ccf099c2deed20af50bf9556e2fc3b7a4382
2024-03-07 22:17:39 +08:00
qqingzheng cb86e18190 [refactor] rename train_videogpt.sh
Former-commit-id: e2ea2caacaf5be046f48cf50a53ccafa7a3cc351
2024-03-07 06:15:14 +00:00
Chestnut 589da07fec Merge branch 'main' into videogpt
Former-commit-id: 714cc434cea5823fc09ef6fe502b1fe83170cae2
2024-03-07 01:25:58 +08:00
qqingzheng b409ef399a [refactor] added more methods in abstract classes
Former-commit-id: a535e3a06eb766bebe7baa3d9df64b04d04d40e6
2024-03-06 17:23:40 +00:00
lb203 921d927a27 Update train.sh
Former-commit-id: 66aa3709f80d89a30d0a57c5e033ce1ae46f998f
2024-03-07 01:06:20 +08:00
lb203 06423d785d Update train.sh
Former-commit-id: 6c4957d715eaa489691be21dd0abdb9309660e0d
2024-03-07 00:57:14 +08:00
LinB203 38bf4afdfa support accelerate training
Former-commit-id: b32671c9991b9a9db66cb148822ccb3e8be7dc11
2024-03-07 00:55:35 +08:00
Chestnut 140eb267d9 Merge branch 'main' into videogpt
Former-commit-id: 0a45d76fdcc3faf167239dfe5dc055b30f3f4e05
2024-03-06 23:36:14 +08:00
LinB203 1b99cf2789 add sample script
Former-commit-id: 9ce847651bb6dba28c51ce5807e222ae2902a78a
2024-03-06 23:23:17 +08:00
qqingzheng 107f140b56 [fix] fix a bug when using ucf101_stride4x4x4
Former-commit-id: 9911403c688321db731450d9383bc02923a4be2d
2024-03-06 14:51:35 +00:00
qqingzheng 1bed11aee3 [refactor] reformat videogpt, support training videogpt on accelerate
Former-commit-id: 75385ed47fe60008b61b1bcf89f61ea696d095bc
2024-03-06 13:45:16 +00:00
qqingzheng 3810227fd6 first commit
Former-commit-id: ec3481bc35a8120d519082dc76585428f0896470
2024-03-06 08:16:58 +00:00
LinB203 cf40ce1e7a add latte
Former-commit-id: 01ba10e5d65c9d2d957472fbe7958e1b07a971e9
2024-03-06 15:28:20 +08:00
Junwu Zhang a02832797a reformat code
Former-commit-id: 7a3dccf2c738eaf535e62996211c8bc5f3de5683
2024-03-04 14:06:39 +00:00