microsoft/DeepSpeed

Pull Requests Commits

AISC launcher fixes (#3637)

jeffra committed 3 years ago

Verified 49a73549

Align InferenceEngine to store ms in _model_times (#3501)

HolyFalafel committed 3 years ago

Verified d755b9d6

attn layer_past print

molly-smith committed 3 years ago

085e469a

add attention prints

molly-smith committed 3 years ago

e82e81cf

add flags to run base mlp and ds mlp.

awan-10 committed 3 years ago

81d9c1aa

add enable/disable debug flag.

awan-10 committed 3 years ago

535342bb

Expose Consecutive Hysteresis to Users (#3553)

Quentin-Anthony committed 3 years ago

Verified 0411a9f8

DS init should not broadcast or move zero.Init models (#3611)

tjruwase committed 3 years ago

Verified d39c311f

bug fix (#3609)

conglongli committed 3 years ago

Verified 736bf185

baseline mlp done. matches HF e2e.

awan-10 committed 3 years ago

10540aa0

Fix op_builder against PyTorch nightly (#3596)

malfet committed 3 years ago

Verified 6622776c

add several debug prints/instrumentation to do side by side comparison.

awan-10 committed 3 years ago

d0f765b0

Fix Hybrid Engine for BLOOM (#3580)

lekurile committed 3 years ago

Verified 76679884

Fixing bf16 test (#3551)

jomayeri committed 3 years ago

Verified 49d399cd

awan-10 committed 3 years ago

f66b6a80

fix_typo (#3559)

inkcherry committed 3 years ago

Verified d1c3c0df

rraminen committed 3 years ago

Verified 2fc56841

Clone tensors to avoid torch.save bloat (#3348)

tjruwase committed 3 years ago

Verified 5c3ebd7e

[CPU] Support Intel CPU inference (#3041)

delock committed 3 years ago

Verified 1f72082f

[ROCm] Hip headers fix (#3532)

rraminen committed 3 years ago

Verified 5147b90a

fix typo in comments with deepspeed/ (#3537)

digger-yu committed 3 years ago

Verified c8d3f5eb

Syncing FusedAdam with new Apex features (#3434)

jomayeri committed 3 years ago

Verified 9685eb92

share inflight registry between PartitionedParameterCoordinators (#3462)

HeyangQin committed 3 years ago

Verified 4716b0f7

Fix broadcast error on multi-node training with ZeroStage3 and TensorParallel=2 (#2999)

YizhouZ committed 3 years ago

Verified 9f4a8763

Fix passing offload_optimizer_config instead of offload_param_config to (#3420)

mmhab committed 3 years ago

Verified 5c6da1f0

fix if statement for self.mpu (#3497)

mrwyattii committed 3 years ago

Verified 1824cb55

Doc updates (#3520)

goodship1 committed 3 years ago

Verified 82991df0

fix typo with docs/ (#3523)

digger-yu committed 3 years ago

Verified d73f99d8

change actions/checkout@v2 to v3 (#3526)

digger-yu committed 3 years ago

Verified b3956dc9

Changing monitor loss to aggregate loss over gradient accumulation steps (#3428)

jomayeri committed 3 years ago

Verified 4d269c6e

Newer Older