microsoft/DeepSpeed

Pull Requests Commits

Merge branch 'master' into lokoppak/quantization_3d

cmikeh2 committed 3 years ago

Verified 165efda8

Update barrier and reduce_scatter_base to conform to PyTorch signatures (#2570)

Quentin-Anthony committed 3 years ago

Verified 18d55e54

Merge branch 'master' into lokoppak/quantization_3d

cmikeh2 committed 3 years ago

Verified c4179eba

Merge branch 'lokoppak/quantization_3d' of https://github.com/microsoft/DeepSpeed into lokoppak/quantization_3d

lokoppakmsft committed 3 years ago

a254430b

lokoppakmsft committed 3 years ago

0ea6597a

Fix MegatronLayerPolicy to have megatron_v2=True (#2579)

lekurile committed 3 years ago

Verified 731965db

Fix quantized-inference & Add generic support of checkpoint loading (#2547)

RezaYazdaniAminabadi committed 3 years ago

Verified 35b350b2

Merge branch 'master' into lokoppak/quantization_3d

cmikeh2 committed 3 years ago

Verified 89118e88

Drop Maxwell Support (#2574)

cmikeh2 committed 3 years ago

Verified b8416282

lokoppakmsft committed 3 years ago

3bbafe69

Support fp32 gradaccum for bf16 model (#2566)

delock committed 3 years ago

Verified 06938835

lokoppakmsft committed 3 years ago

527dc219

merge master and fix format

lokoppakmsft committed 3 years ago

eb4e94c7

Merge remote-tracking branch 'origin/master' into lokoppak/quantization_3d

lokoppakmsft committed 3 years ago

fd5584b9

Add N-Dim support to Dequant kernel

lokoppakmsft committed 3 years ago

b3bf7f63

Add Determined to open-source DL frameworks (#2573)

sirredbeard committed 3 years ago

Verified 2d8f3f56

use get_global_rank if available (#2567)

jeffra committed 3 years ago

Verified d4cab2ce

docs: Update the recent url for Megatron-LM (#2564)

rapsealk committed 3 years ago

Verified bbe030c5

jeffra committed 3 years ago

c77d42dc

Fix invalid check of recorded parameter orders in zero stage3. (#2550)

inkcherry committed 3 years ago

Verified aeda7f9f

Abstract accelerator (step 1) (#2504)

delock committed 3 years ago

Verified ffcf3846

add missing moe deprecated fields to inference config (#2556)

mrwyattii committed 3 years ago

Verified c5f85858

encoded ds config into command line argument when launching child processes in autotuning (#2524)

cli99 committed 3 years ago

Verified abe4fc6b

Report progress at gradient accumulation boundary (#2553)

ShijieZZZZ committed 3 years ago

Verified 340fc0cf

Adding Gradient Accumulation Data Type Config (#2512)

jomayeri committed 3 years ago

Verified 21c28029

lokoppakmsft committed 3 years ago

9944f172

Pass down the new DS inference config to replace_transformer_layer. (#2539)

awan-10 committed 3 years ago

Verified 90ae6884

Change Where DS/Triton is Used in Stable Diffusion (#2536)

cmikeh2 committed 3 years ago

Verified 5df1eea7

Add support for inputs > 2D

lokoppakmsft committed 3 years ago

cdb3c23a

Remove mutable default parameter in init_inference() (#2540)

aphedges committed 3 years ago

Verified 4abf637f

Older