8000 Commits · parasail-ai/torchtitan · GitHub

More Web Proxy on the site http://driver.im/

Commits

Commits on Jun 12, 2024

update all toml files to use experimental section (#392)
wanchaol
authored

Commits on Jun 9, 2024

Commits on May 22, 2024

torch.compile each TransformerBlock instead of the whole model (#268)
wanchaol
authored

Commits on May 8, 2024

simplify embedding + first transformer block TP (#314)
wanchaol
authored

Commits on May 3, 2024

turn off dynamic shape for torch.compile (#297)
wanchaol
authored

Commits on Apr 30, 2024

fix 3d mesh order (#288)
wanchaol
authored

Commits on Apr 26, 2024

some misc changes (#278)
wanchaol
authored

Commits on Apr 24, 2024

warn supported dataset checks instead of throw (#260)
wanchaol
authored

Commits on Apr 23, 2024

reenable integration tests with a test tokenizer (#259)
wanchaol
authored

Commits on Apr 20, 2024

Support Llama3 8b/70b (#256)
wanchaol
authored

Commits on Apr 17, 2024

Commits on Apr 11, 2024

Rename to torchtitan (#221)
wanchaol
authored

Commits on Apr 10, 2024

fix SAC to use the correct reduce_scatter op (#215)
wanchaol
authored

Commits on Apr 4, 2024

bump comm.train_timeout_seconds (#189)
wanchaol
authored

Commits on Mar 27, 2024

add basic AC configs for 13B and 70B (#169)
wanchaol
authored

Commits on Mar 25, 2024

rename sequence_parallel to tensor_parallel (#162)
wanchaol
authored

Commits on Mar 22, 2024

refactor config manager and support cmd overrides (#157)
wanchaol
authored

Commits on Mar 21, 2024

disable buffer reuse for compile for now (#156)
wanchaol
authored

Commits on Mar 13, 2024

use SequenceParallel style in tp/sp (#133)
wanchaol
authored

Commits on Mar 9, 2024

Commits on Feb 29, 2024

Add llama 7B config (#100)
wanchaol
authored

0