microsoft/DeepSpeed

Pull Requests Commits

Update conda env to have max pydantic version

loadams committed 1 year ago

f5991a8d

fix deepspeed4science links (#4358)

conglongli committed 1 year ago

Verified 78c3b148

DeepSpeed4Science (#4357)

conglongli committed 1 year ago

Verified f876d81d

Support InternLM (#4137)

wangruohui committed 1 year ago

Verified 367d6f9c

Pass base_dir to model files can be loaded for auto-tp/meta-tensor. (#4348)

awan-10 committed 1 year ago

Verified b9d719a6

Fix autotune to support Triton 2.1 (#4340)

stephen-youn committed 1 year ago

Verified 3f3e9fb1

Fix skipped inference tests (#4336)

mrwyattii committed 1 year ago

Verified 8c7f7fd2

Suppress noise (#4310)

tjruwase committed 1 year ago

Verified 3ea42c6e

Fix a bug in the implementation of dequantization for inference (#3433)

sakogan committed 1 year ago

Verified 9bf77782

DS-Chat BLOOM: Fix Attention mask (#4338)

lekurile committed 1 year ago

Verified 0c83436a

clear redundant timers (#4308)

starkhu committed 1 year ago

Verified e20e4a9d

Add release version checking (#4328)

loadams committed 1 year ago

Verified b5977d41

Fix Zero3 contiguous grads, reduce scatter false accuracy issue (#4321)

nelyahu committed 1 year ago

Verified 11aa880e

Clean up modeling code (#4320)

loadams committed 1 year ago

Verified 180dd397

Handle empty parameter groups (#4277)

tjruwase committed 1 year ago

Verified 9adc73ff

Update README.md (#4316)

NinoRisteski committed 1 year ago

Verified 48f2192a

README update (#4303)

tjruwase committed 1 year ago

Verified 581e44dd

Update release and bump patch versioning flow (#4286)

loadams committed 1 year ago

Verified 8c1eed2e

added a bert-model check for triton (#4266)

stephen-youn committed 1 year ago

Verified ffd82bb0

ZeRO-Inference v2 release

tjruwase committed 1 year ago

94174f88

jeffra committed 1 year ago

c700195d

Update index.md (#4297)

NinoRisteski committed 1 year ago

Verified 04a6fedf

fix user args parsing of string with spaces on runner (#4265)

YudiZh committed 1 year ago

Verified e75c285a

ZeRO-Inference refresh (#4197)

tjruwase committed 1 year ago

Verified aa4a7401

AMD Kernel Compatibility Fixes (#3180)

cmikeh2 committed 1 year ago

Verified 542dc0d5

Support iterators with incompletely defined __len__ functions (#2445)

codedecde committed 1 year ago

Verified e1949565

Keep hpz secondary tensor in forward pass (#4288)

HeyangQin committed 1 year ago

Verified 1f0a44d9

update deepspeed to run with the most recent triton 2.1.0 (#4278)

stephen-youn committed 1 year ago

Verified e8ed7419

Update README.md (#4284)

NinoRisteski committed 1 year ago

Verified c55fad9f

fix iteration timing used in autotuning when gradient_accumulation_steps > 1 (#2888)

cli99 committed 1 year ago

Verified bce6ed1c

Older