microsoft/DeepSpeed

Pull Requests Commits

add note regarding prepend_path modifcation

mrwyattii committed 2 years ago

4bca8f19

fix absolute path for JIT

mrwyattii committed 2 years ago

9c656de5

fix for importing submodules from deepspeed/

mrwyattii committed 2 years ago

f50137ee

Merge branch 'master' into mrwyattii/remove-symlinks

mrwyattii committed 2 years ago

Verified 4e32cd9e

flake8 and cuda-check changes

mrwyattii committed 2 years ago

b15d662a

formatting - yapf

mrwyattii committed 2 years ago

2da39822

added prepend path to fix source and include paths

mrwyattii committed 2 years ago

03a20ed0

update setup.py imports

mrwyattii committed 2 years ago

a1ec2eea

remove symlinks and replace with dirs

mrwyattii committed 2 years ago

672eee96

Update README.md (#4316)

NinoRisteski committed 2 years ago

Verified 48f2192a

Handle empty parameter groups (#4277)

tjruwase committed 2 years ago

Verified 9adc73ff

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

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

Use clearer naming (#2548)

Crispig committed 2 years ago

Verified 18a04d04

use ```non_reentrant_checkpoint``` fix requires_grad of input must be true for activation checkpoint layer in pipeline train. (#4224)

inkcherry committed 2 years ago

Verified 60a3e89e

check for zero (empty) param groups in llama + hf/accelerate. (#4270)

awan-10 committed 2 years ago

Verified 16d8953a

Fix nv-inference/un-pin transformers (#4269)

loadams committed 2 years ago

Verified 55d9964c

skipping redundant MoE optimizer state loading (#4120)

Alexander Jipa committed 2 years ago

Verified e801e6d7

Older