microsoft/DeepSpeed

Pull Requests Commits

Update to add workflow dispatch

loadams committed 1 year ago

714840a4

Nvme offload checkpoint (#4707)

eisene committed 1 year ago

Verified d058d4b3

fix num_kv_heads sharding in uneven autoTP for Falcon-40b (#4712)

Yejing-Lai committed 1 year ago

Verified 1787673e

[NPU]Add ZeRO-Infinity feature for NPU (#4809)

misstek committed 1 year ago

Verified b596963b

Release overlap_comm & contiguous_gradients restrictions for ZeRO 1 (#4887)

li-plus committed 1 year ago

Verified af033831

enable starcode((kv_head=1)) autotp (#4896)

Yejing-Lai committed 1 year ago

Verified 85132adc

[XPU] XPU accelerator support for Intel GPU device (#4547)

delock committed 1 year ago

Verified f4f31317

add sharded loading for safetensors in AutoTP (#4854)

sywangyi committed 1 year ago

Verified c8c57b8c

Support cpu tensors without direct device invocation (#3842)

abhilash1910 committed 1 year ago

Verified c84c28d2

Add support of Microsoft Phi-2 model to DeepSpeed-FastGen (#4812)

arashb committed 1 year ago

Verified 83427253

zero_to_fp32.py: Handle a case where shape doesn't have numel attr (#4842)

nelyahu committed 1 year ago

Verified 691458f8

Pipeline: Add support to eval micro bs configuration (#4859)

nelyahu committed 1 year ago

Verified ac84cf3f

Partition parameters: Minor refactoring of use_secondary_tensor condition (#4868)

deepcharm committed 1 year ago

Verified 81cc3207

[NPU] Fix npu offload bug (#4883)

CurryRice233 committed 1 year ago

Verified d873ce61

Fix f-string messages (#4865)

li-plus committed 1 year ago

Verified 9ec55bd9

optimize grad_norm calculation in stage3.py (#4436)

mmhab committed 1 year ago

Verified ea0d8114

Remove hooks on gradient accumulation on engine/optimizer destroy (#4858)

chiragjn committed 1 year ago

Verified 40342055

deepspeed engine: fp16 support validation on init (#4843)

nelyahu committed 1 year ago

Verified 3e94f8c7

Fix exception handling in get_all_ranks_from_group() function (#4862)

HeyangQin committed 1 year ago

Verified c37fe9cb

doc corrections (#4861)

goodship1 committed 1 year ago

Verified 75c77202

Update version.txt after 0.12.6 release (#4850)

mrwyattii committed 1 year ago

Verified 827e1ca8

Mixtral FastGen Support (#4828)

cmikeh2 committed 1 year ago

Verified c00388a2

engine.py: remove unused _curr_save_path (#4844)

nelyahu committed 1 year ago

Verified 18643914

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

deepcharm committed 1 year ago

Verified 9e455d76

Capture short kernel sequences to graph (#4318)

inkcherry committed 1 year ago

Verified d5a7c1e0

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

mrwyattii committed 1 year ago

Verified 060a8e18

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

minchao-sun committed 1 year ago

Verified 6d7b44a8

Update zeropp.md (#4835)

goodship1 committed 1 year ago

Verified a00bdde8

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

nelyahu committed 1 year ago

Verified 3f1c3c2a

support baichuan model: (#4721)

baodii committed 1 year ago

Verified c20f6fa4

Older