microsoft/DeepSpeed

Pull Requests Commits

Merge branch 'master' into cholmes/migrate-to-dequant-lib

cmikeh2 committed 3 years ago

Verified 68984426

Add checkpoint sharding unit tests (#2561)

mrwyattii committed 3 years ago

Verified ccb8eb81

Merge branch 'master' into cholmes/migrate-to-dequant-lib

cmikeh2 committed 3 years ago

Verified 1e47c174

Support N-dimension input in quantization kernel (#2575)

lokoppakmsft committed 3 years ago

Verified 591744eb

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

Quentin-Anthony committed 3 years ago

Verified 18d55e54

Fix MegatronLayerPolicy to have megatron_v2=True (#2579)

lekurile committed 3 years ago

Verified 731965db

Migrate W8A16 inference to dequant utils

cmikeh2 committed 3 years ago

b72d895c

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

RezaYazdaniAminabadi committed 3 years ago

Verified 35b350b2

Drop Maxwell Support (#2574)

cmikeh2 committed 3 years ago

Verified b8416282

Support fp32 gradaccum for bf16 model (#2566)

delock committed 3 years ago

Verified 06938835

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

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

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

aphedges committed 3 years ago

Verified 4abf637f

Add MII tests (#2533)

mrwyattii committed 3 years ago

Verified c5ee27f7

Make DS-Inference config readable from JSON (#2537)

mrwyattii committed 3 years ago

Verified 8b4318b9

Ensure is initialized for SD (#2534)

cmikeh2 committed 3 years ago

Verified 57e0a550

Fixes for torch 1.14 due to new torch.numel return type (#2522)

jeffra committed 3 years ago

Verified 21105521

Initial dequant library implementation (#2521)

cmikeh2 committed 3 years ago

Verified 30c8d8a8

Update codeowners (#2525)

jeffra committed 3 years ago

Verified 0b265326

Add note about nvcc/hipcc requirement (#2519)

jeffra committed 3 years ago

Verified 7ce371b1

Older