deepspeedai/DeepSpeed

Pull Requests Commits

add a base moe container skeleton for brainstorming.

awan-10 committed 3 years ago

56f8c51f

lekurile committed 3 years ago

9b24af91

Merge branch 'master' into inference-refactor-v1

lekurile committed 3 years ago

74233437

lekurile committed 3 years ago

09f37abb

Add ds_megatron_gpt model implementation + make explicit megatron is GPT specific

lekurile committed 3 years ago

9a36258d

Megatron container working w/ moe=1

lekurile committed 3 years ago

75003bed

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

Container based OPT is working

lekurile committed 3 years ago

8ca77499

awan-10 committed 3 years ago

3f511079

Public inference v2 (#387)

awan-10 committed 3 years ago

Verified 102a1bdb

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

Add missing Inference sub-configs (#2518)

mrwyattii committed 3 years ago

Verified fe678544

Fix backward compatibility for InferenceConfig (#2516)

mrwyattii committed 3 years ago

Verified e59f8054

Deepspeed quantization library v0.1 (#2450)

lokoppakmsft committed 3 years ago

Verified 78d4ca1f

Add max_tokens alias to max_out_tokens arg to maintain backwards compatibility (#2508)

lekurile committed 3 years ago

Verified d40a15fc

Update docs to autogenerate pydantic config model docs (#2509)

mrwyattii committed 3 years ago

Verified 43bf035c

DeepSpeed inference config. (#2459) (#2472)

awan-10 committed 3 years ago

Verified b5d18a6a

jeffra committed 3 years ago

a4ceabb6

Older