microsoft/DeepSpeed

Pull Requests Commits

engine.py: remove unused _curr_save_path (#4844)

nelyahu committed 2 years ago

Verified 18643914

Checkpointing: Avoid assigning tensor storage with different device (#4836)

deepcharm committed 2 years ago

Verified 9e455d76

Capture short kernel sequences to graph (#4318)

inkcherry committed 2 years ago

Verified d5a7c1e0

Add cuda_accelerator.py to triggers for A6000 test (#4848)

mrwyattii committed 2 years ago

Verified 060a8e18

[NPU] load EXPORT_ENV based on different accelerators to support multi-node training on other devices (#4830)

minchao-sun committed 2 years ago

Verified 6d7b44a8

Update zeropp.md (#4835)

goodship1 committed 2 years ago

Verified a00bdde8

inference engine: check if accelerator supports FP16 (#4832)

nelyahu committed 2 years ago

Verified 3f1c3c2a

support baichuan model: (#4721)

baodii committed 2 years ago

Verified c20f6fa4

Update CODEOWNERS (#4838)

mrwyattii committed 2 years ago

Verified 83fa673a

Inference changes for incorporating meta loading checkpoint (#4692)

oelayan7 committed 2 years ago

Verified 4c2cac03

Cache metadata for TP activations and grads (#4360)

BacharL committed 2 years ago

Verified 4559dadd

Update version.txt after 0.12.5 release (#4826)

mrwyattii committed 2 years ago

Verified 4d866bd5

Fix 4649 (#4650)

Alienfeel committed 2 years ago

Verified 65b77277

Refactor launcher user arg parsing (#4824)

mrwyattii committed 2 years ago

Verified d37fc25d

mv DeepSpeedEngine param_names dict init post _configure_distributed_model (#4803)

nelyahu committed 2 years ago

Verified 449e454f

fix falcon model load from_config meta_data error (#4783)

baodii committed 2 years ago

Verified faa00b13

Minor updates to CI workflows (#4823)

mrwyattii committed 2 years ago

Verified bc1b5a6c

Update imports from Transformers (#4817)

loadams committed 2 years ago

Verified b83b1c2e

Update broken link in docs (#4822)

mrwyattii committed 2 years ago

Verified d1f1d45f

fix [BUG] 'DeepSpeedGPTInference' object has no attribute 'dtype' for… (#4814)

jxysoft committed 2 years ago

Verified 4a6e0c06

Accelerate CI fix (#4819)

mrwyattii committed 2 years ago

Verified 84eaf5ac

Universal Checkpoint for Sequence Parallelism (#4752)

samadejacobs committed 2 years ago

Verified 8998707a

fix for tests using torch<2.1 (#4818)

mrwyattii committed 2 years ago

Verified 7914b195

Switch paths-ignore to single quotes, update paths-ignore on nv-pre-compile-ops (#4805)

loadams committed 2 years ago

Verified 3324efd9

Add support of Falcon models (7b, 40b, 180b) to DeepSpeed-FastGen (#4790)

arashb committed 2 years ago

Verified a7900bcc

Fix for stage3 when setting different communication data type (#4540)

BacharL committed 2 years ago

Verified b1868169

[NPU] Add HcclBackend for 1-bit adam, 1-bit lamb, 0/1 adam (#4733)

RUAN-ZX committed 2 years ago

Verified 2ce6bf8c

improve the way to determine whether a variable is None (#4782)

RUAN-ZX committed 2 years ago

Verified 7b818ee9

[BUG] partition_balanced return wrong result. (#4312)

zjjMaiMai committed 2 years ago

Verified 2bdf061f

BF16_Optimizer: add support for bf16 grad acc (#4713)

nelyahu committed 2 years ago

Verified ce607080

Newer Older