microsoft/DeepSpeed

Pull Requests Commits

Actually make it fail

loadams committed 2 years ago

5c8bb8af

Test failure mode

loadams committed 2 years ago

050dc595

DeepSpeed-Chat Llama2/stability release (#4240)

awan-10 committed 2 years ago

Verified 2420e23f

Add Mixed Precision ZeRO++ tutorial (#4241)

HeyangQin committed 2 years ago

Verified 69e89dd5

jeffra committed 2 years ago

dc7b65f6

Enable hpz when running with torch.no_grad (#4232)

HeyangQin committed 2 years ago

Verified 462def45

fix MegatronLayerPolicy to be compatible with the newest ParallelTransformerLayer (#4236)

Dino Chen committed 2 years ago

Verified 6cbf6661

Enable AMD MI200 and H100 to run on branches for testing (#4238)

loadams committed 2 years ago

Verified 5dbc5313

fix: linker issues in conda environments #3929 (#4235)

maximegmd committed 2 years ago

Verified 194b0043

Empty tensor size check (#4186)

jomayeri committed 2 years ago

Verified 57d629a1

added port argument for ssh (#4117)

Hiromasa-H committed 2 years ago

Verified 8145b5e4

Fix fused qkv sizing for bloom (#4161)

molly-smith committed 2 years ago

Verified 042115c8

add npu support dtypes (#4223)

CurryRice233 committed 2 years ago

Verified 60d7b0a3

feat(activation_checkpointing): add `non_reentrant_checkpoint` to support inputs require no grad (#4118)

hughpu committed 2 years ago

Verified 42c1e916

Fix pipline dataloader when batch elements contain tuple (#565)

ghosthamlet committed 2 years ago

Verified c69bd1f7

Fixes timer error referenced in #4212 (#4213)

bjoernpl committed 2 years ago

Verified 0b7a760c

add meta onDevice support for LLAMA2 (#4147)

Dino Chen committed 2 years ago

Verified 0712e299

Simplify Gradient Attribute Names (#4214)

jomayeri committed 2 years ago

Verified f6903190

Add MuP optimizers (#2043)

mrwyattii committed 2 years ago

Verified 9647ea79

add ulysses blog index (#4215)

conglongli committed 2 years ago

Verified d6c2e6b0

DeepSpeed Ulysses Chinese blog translation (#4210)

HeyangQin committed 2 years ago

Verified 63e17769

Add Japanese blog of DS-Ulysses (#4209)

tohtana committed 2 years ago

Verified 3808273c

Update README.md (#4211)

samadejacobs committed 2 years ago

Verified c274e512

Update Ulyssess (#4205)

samadejacobs committed 2 years ago

Verified 10bef7ac

DS-Ulysses formating (#4204)

samadejacobs committed 2 years ago

Verified 961827be

[docs] fix pypi badge

jeffra committed 2 years ago

Verified 3e82cb64

Ds ulysses news (#4202)

samadejacobs committed 2 years ago

Verified 5de0662c

Deepspeed-Ulysses blog (#4201)

samadejacobs committed 2 years ago

Verified 4e5d39fe

DeepSpeed Ulysses release (#4198)

samadejacobs committed 2 years ago

Verified a855405e

Load z3 checkpoints for inference (#4171)

tjruwase committed 2 years ago

Verified 6df15873

Older