microsoft/DeepSpeed

Pull Requests Commits

Add _load_checkpoint function to InferenceEngine

lekurile committed 2 years ago

4af3deb6

Remove _load_checkpoint function from InferenceEngine

lekurile committed 2 years ago

0e8f7464

Add additional assertions

lekurile committed 2 years ago

ebc228ad

Merge branch 'master' into lekurile/update_inf_ckpt_load

lekurile committed 2 years ago

f80103fb

fix spelling error with docs/index.md (#3443)

digger-yu committed 2 years ago

Verified 58c4d230

fix spelling error with deepspeed/ (#3494)

digger-yu committed 2 years ago

Verified 19816642

fix a small type error on bf16+pp (#3441)

ys950902 committed 2 years ago

Verified 195563a2

fix spelling error (#3482)

luliyucoordinate committed 2 years ago

Verified f1fab902

Fix spelling errors in comments and documents (#3486)

digger-yu committed 2 years ago

Verified 87edbc8d

Doc corrections (#3435)

goodship1 committed 2 years ago

Verified 41321180

Build: Update license in setup (#3484)

PabloEmidio committed 2 years ago

Verified f3f4c449

[Docs] Update automatic-tensor-parallelism.md (#3198)

sywangyi committed 2 years ago

Verified 9715d802

launcher/multinode_runner.py: mapping env variables in running cmd for mpich runner (#3372)

YizhouZ committed 2 years ago

Verified 4e886f05

add sharded checkpoint loading for AutoTP path to reduce the peak mem… (#3102)

sywangyi committed 2 years ago

Verified d10b8ca0

Hybrid Engine Refactor and Llama Inference Support (#3425)

cmikeh2 committed 2 years ago

Verified 0a61d5d6

add skip for mismatched cuda versions (#3436)

mrwyattii committed 2 years ago

Verified d2e6110e

Enable auto TP policy for llama model (#3170)

jianan-gu committed 2 years ago

Verified 2c63e349

jeffra committed 2 years ago

5314c9a1

Update 2020-09-09-sparse-attention.md (#3432)

goodship1 committed 2 years ago

Verified e0e80850

Adagrad support in ZeRO (#3401)

jomayeri committed 2 years ago

Verified d3550dc8

[zero_to_fp32] fix shared param recovery (#3407)

stas00 committed 2 years ago

Verified 77ebf760

Add ZeRO 1 support to PP for BF16. (#3399)

jomayeri committed 2 years ago

Verified dcb4a7d6

OPT Activation Function Hotfix (#3400)

cmikeh2 committed 2 years ago

Verified 52d7e80a

Fix `PipelineEngine.eval_batch` result (#3316)

nrailg committed 2 years ago

Verified b0d9c4d0

Update links (#3424)

loadams committed 2 years ago

Verified b2ac7f63

Add HE support for the rest of model containers (#3191)

RezaYazdaniAminabadi committed 2 years ago

Verified 3e856464

save tensors in context of memory_efficient_linear (#3413)

tohtana committed 2 years ago

Verified 42858a98

doc updates (#3415)

goodship1 committed 2 years ago

Verified b4b63f52

Fix pipeline module evaluation when contiguous activation checkpointing is enabled (#3005)

BacharL committed 2 years ago

Verified 7ddc3b01

remove megatron-lm, no longer pip installable (#3389)

jeffra committed 2 years ago

Verified a094c976

Older