Code

Refactor model conversion#296

Code

Merged

lzhangzz merged 27 commits intoInternLM:mainInternLM/lmdeploy:mainfrom irexyc:deployirexyc/lmdeploy:deployCopy head branch name to clipboardNov 3, 2023

+1,743-1,050

Commits

Commits on Oct 31, 2023

• split deploy.py

  irexyc

  committed
  19ed1f9

• fix get_cuda_tensor

  irexyc

  committed
  10e3884

• deploy qwen_awq

  irexyc

  committed
  b5d4ece

• fix lint

  irexyc

  committed
  8c457a0

• add docstring

  irexyc

  committed
  edc60dd

• fix

  irexyc

  committed
  36c6f35

• support baichuan/baichuan-awq

  irexyc

  committed
  43b350a

• parameterizing size_per_head

  irexyc

  committed
  0985fbe

• remove try/except

  irexyc

  committed
  e6f1b2c

• limit input model_format

  irexyc

  committed
  cd6f01a

• add quant_path param

  irexyc

  committed
  b0146c0

• remove old deploy.py

  irexyc

  committed
  44be235

• fix path

  irexyc

  committed
  a4101db

Commits on Nov 1, 2023

• fix transformer layer range when load bins

  irexyc

  committed
  36a0f1a

• fix qwen init

  irexyc

  committed
  f1e4456

• split & save log

  irexyc

  committed
  db39b68

• relative import

  irexyc

  committed
  4617668

• update get_config

  irexyc

  committed
  0449c69

• WeightFileMgr -> Reader

  irexyc

  committed
  e65127c

• rename

  irexyc

  committed
  2a0405c

• update

  irexyc

  committed
  c8fe088

• fix init_layer_id

  irexyc

  committed
  f80a4c8

• rename llama.py -> meta_llama.py, hf.py -> llama.py

  irexyc

  committed
  b6255cb

• reduce code

  irexyc

  committed
  e092715

Commits on Nov 2, 2023

• update arg description

  irexyc

  committed
  dfc54d6

• fix meta llama

  irexyc

  committed
  794b620

• manually cleanup meta model params

  irexyc

  committed
  8224989