Commits

Branch selector

 main

User selector

yangw1234

Datepicker

All time

Commit History

Commits on Sep 6, 2024

• Optimize broadcast for npu llama (#12028)

  yangw1234

  committedSep 6, 2024
  58555bd

Commits on Sep 2, 2024

• Support Qwen2-7b MLP in int4 and transpose_value_cache=True (#11968)

  yangw1234

  committedSep 2, 2024
  c48817b

Commits on Aug 29, 2024

• remove obselete npu code (#11967)

  yangw1234

  committedAug 29, 2024
  fbf088f

Commits on Aug 21, 2024

• Refactor npu mp to make it easier to integrate new models (#11873)

  yangw1234

  committedAug 21, 2024
  209d42a

Commits on Aug 20, 2024

• Fix run_decoders bug (#11871)

  yangw1234

  committedAug 20, 2024
  bdaeee1

Commits on Aug 19, 2024

• separate prefill into a process (#11787)

  

  yangw1234

  and

  sgwhat

  committedAug 19, 2024
  99b05ba

Commits on Aug 14, 2024

• follow up on experimental support of fused decoder layer for llama2 (#11785)

  yangw1234

  committedAug 14, 2024
  51bcac1

Commits on Apr 24, 2024

• Support the desc_act feature in GPTQ model (#10851)

  yangw1234

  committedApr 24, 2024
  1ce8d7b

Commits on Apr 18, 2024

• Initial llama3 example (#10799)

  

  yangw1234

  and

  Oscilloscope98

  committedApr 18, 2024
  8153c30

• use config rope_theta (#10787)

  yangw1234

  committedApr 18, 2024
  952e517

Commits on Apr 11, 2024

• Fuse MOE indexes computation (#10716)

  yangw1234

  committedApr 11, 2024
  019293e

Commits on Apr 8, 2024

• support fp8 in xetla (#10555)

  yangw1234

  committedApr 8, 2024
  5a1f446

Commits on Apr 4, 2024

• Update llama_cpp_quickstart.md (#10663)

  yangw1234

  committedApr 4, 2024
  ac65ab6

Commits on Mar 18, 2024

• Support running pipeline parallel inference by vertically partitioning model to different devices (#10392)

  yangw1234

  committedMar 18, 2024
  9e763b0

Commits on Mar 1, 2024

• use fused qkv forward in qwen2 (#10185)

  yangw1234

  committedMar 1, 2024
  f4d7dbc

Commits on Feb 27, 2024

• draft mmint4 (#10031)

  yangw1234

  committedFeb 27, 2024
  c581c6d

Commits on Dec 19, 2023

• fusing qkv project and rope (#9612)

  yangw1234

  committedDec 19, 2023
  f4fb58d

Commits on Nov 14, 2023

• Support directly loading gptq models from huggingface (#9391)

  yangw1234

  committedNov 14, 2023
  51d07a9

Commits on Nov 1, 2023

• fix import ipex problem (#9323)

  yangw1234

  committedNov 1, 2023
  e1bc18f

Commits on Oct 27, 2023

• Support qlora in CPU (#9233)

  yangw1234

  committedOct 27, 2023
  163d033

• Add deepspeed autotp example readme (#9289)

  yangw1234

  committedOct 27, 2023
  8838707

Commits on Oct 25, 2023

• Support deepspeed AutoTP (#9230)

  yangw1234

  committedOct 25, 2023
  067c7e8

Commits on Oct 20, 2023

• support transformers 4.34+ for llama (#9229)

  yangw1234

  committedOct 20, 2023
  118249b

Commits on Oct 18, 2023

• Fix removing convert dtype bug (#9216)

  yangw1234

  committedOct 18, 2023
  b0ddde0

Commits on Oct 17, 2023

• Support XPU DDP training and autocast for LowBitMatmul (#9167)

  yangw1234

  committedOct 17, 2023
  7160afd

Commits on Oct 15, 2023

• Fixes for xpu Bf16 training (#9156)

  yangw1234

  committedOct 15, 2023
  7a2de00

Commits on Oct 6, 2023

• Fix llama when rope scaling is not None (#9086)

  yangw1234

  committedOct 6, 2023
  36dd4af

• using bigdl-llm fused rope for llama (#9066)

  yangw1234

  committedOct 6, 2023
  fcb1c61

Commits on Oct 5, 2023

• add export merged model example (#9018)

  yangw1234

  committedOct 5, 2023
  88565c7

• Use ipex fused rms norm for llama (#9081)

  yangw1234

  committedOct 5, 2023
  0cd8f1c

Commits on Sep 19, 2023

• Experiment XPU QLora Finetuning (#8937)

  yangw1234

  committedSep 19, 2023
  c88f6ec

Commits on Sep 12, 2023

• Make llama attention stateless (#8928)

  yangw1234

  committedSep 12, 2023
  16761c5

Commits on Sep 8, 2023

• Support latest transformer version (#8923)

  yangw1234

  committedSep 8, 2023
  ee98cdd

• Fix chatglm2 attention and kv cache (#8924)

  yangw1234

  committedSep 8, 2023
  25428b2

Commits on Sep 7, 2023

• Use new layout for xpu qlinear (#8896)

  yangw1234

  committedSep 7, 2023
  c34400e

Pagination

PreviousNext