Making large AI models cheaper, faster and more accessible
COMMITS
August 12, 2024
P
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] committed
R
fix readme, tokenization fully tested
root committed
R
fix readme, tokenization fully tested
root committed
R
fix readme
root committed
August 7, 2024
E
[Docs] clarify launch port
Edenzzzz committed
August 2, 2024
Y
Merge pull request #5962 from hpcaitech/colossalchat
YeAnbang committed
Y
Support overall loss, update KTO logging
YeAnbang committed
W
[lora] lora support hybrid parallel plugin (#5956)
Wang Binluo committed
T
[feat] Dist Loader for Eval (#5950)
Tong Li committed
July 30, 2024
B
July 26, 2024
B
[moe] solve dp axis issue
botbw committed
B
[doc] add MoeHybridParallelPlugin docstring
botbw committed
July 25, 2024
H
[moe] remove force_overlap_comm flag and add warning instead
hxwang committed
H
[chore] docstring
hxwang committed
H
[misc] remove useless condition
haze188 committed
H
[misc] fix ci failure: change default value to false in moe plugin
haze188 committed
H
[misc] remove incompatible test config
haze188 committed
H
[chore] change moe_pg_mesh to private
hxwang committed
H
Revert "[moe] implement submesh initialization"
hxwang committed
H
[moe] refactor mesh assignment
hxwang committed
July 24, 2024
H
[misc] remove debug/print code
haze188 committed
H
[misc] skip redunant test
haze188 committed
July 23, 2024
H
[deepseek] replace attn (a workaround for bug in transformers)
hxwang committed
H
[test] add check
hxwang committed
H
[moe] deepseek moe sp support
haze188 committed
B
[bug] fix: somehow logger hangs the program
botbw committed
July 22, 2024
H
[test] fix test: test_zero1_2
hxwang committed
H
[moe] remove ops
hxwang committed
H
[chore] solve moe ckpt test failure and some other arg pass failure
hxwang committed
July 19, 2024
P
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] committed