microsoft/DeepSpeed

Pull Requests Commits

Update test_compression as it does not need the other version

loadams committed 2 years ago

1553fc49

loadams committed 2 years ago

98ff2edf

Remove more whitespace and make files match more

loadams committed 2 years ago

1ac135fa

Add back BertSelfAttention

loadams committed 2 years ago

2611477e

Revert "Remove BertLayerNorm code too"

loadams committed 2 years ago

ed72937e

Remove BertLayerNorm code too

loadams committed 2 years ago

da9770d3

loadams committed 2 years ago

76a311ab

Include BertConfig though

loadams committed 2 years ago

d530bbc7

Try replacing everything except BertEncoder

loadams committed 2 years ago

9aaf6406

Clean up modeling code (#4320)

loadams committed 2 years ago

Verified 180dd397

Handle empty parameter groups (#4277)

tjruwase committed 2 years ago

Verified 9adc73ff

Update README.md (#4316)

NinoRisteski committed 2 years ago

Verified 48f2192a

Formatting whitespace

loadams committed 2 years ago

3e3e34a0

Switch away from local modeling.py files to importing from transformers

loadams committed 2 years ago

55dee96d

Merge branch 'master' into loadams/tar-vuln

loadams committed 2 years ago

dcc8dc42

README update (#4303)

tjruwase committed 2 years ago

Verified 581e44dd

Update release and bump patch versioning flow (#4286)

loadams committed 2 years ago

Verified 8c1eed2e

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

stephen-youn committed 2 years ago

Verified ffd82bb0

ZeRO-Inference v2 release

tjruwase committed 2 years ago

94174f88

jeffra committed 2 years ago

c700195d

Update index.md (#4297)

NinoRisteski committed 2 years ago

Verified 04a6fedf

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

YudiZh committed 2 years ago

Verified e75c285a

ZeRO-Inference refresh (#4197)

tjruwase committed 2 years ago

Verified aa4a7401

AMD Kernel Compatibility Fixes (#3180)

cmikeh2 committed 2 years ago

Verified 542dc0d5

Merge branch 'master' into loadams/tar-vuln

loadams committed 2 years ago

Verified 0c25c0cb

Support iterators with incompletely defined __len__ functions (#2445)

codedecde committed 2 years ago

Verified e1949565

Keep hpz secondary tensor in forward pass (#4288)

HeyangQin committed 2 years ago

Verified 1f0a44d9

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

stephen-youn committed 2 years ago

Verified e8ed7419

Update README.md (#4284)

NinoRisteski committed 2 years ago

Verified c55fad9f

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

cli99 committed 2 years ago

Verified bce6ed1c

Older