Code

Support qwen2 vl#2689

Code

Merged

drbh merged 18 commits intomainhuggingface/text-generation-inference:mainfrom support-qwen2-vlhuggingface/text-generation-inference:support-qwen2-vlCopy head branch name to clipboardOct 30, 2024

+705-10

Commits

Commits on Oct 28, 2024

• feat: add support for qwen2 vl model

  drbh

  committed
  d96eef2

• feat: fix token padding, enable warmup and process basic request

  drbh

  committed
  09ac4fb

• fix: improve get_position_ids, add lift embed_tokens

  drbh

  committed
  22fdf93

• fix: remove get_cos_sin_hack dev function

  drbh

  committed
  ec93328

• feat: add simple test chat with meesage and text

  drbh

  committed
  80ea4f0

• fix: lint test

  drbh

  committed
  e1114c2

• fix: adjust positional embeddings for multi dimensional position ids

  drbh

  committed
  279b114

• fix: update docs and lint unused vars

  drbh

  committed
  670d75b

• fix: include linted file

  drbh

  committed
  aa2aa9f

• fix: add norm after text output

  drbh

  committed
  65558b3

• fix: format model file

  drbh

  committed
  6208d10

• fix: adjust for ruff lints

  drbh

  committed
  f2a1b1b

• fix: remove unused rotate_half

  drbh

  committed
  831a07f

• feat: refactors and calc num features

  drbh

  committed
  fb1ae6d

Commits on Oct 29, 2024

• fix: prefer position_ids passed from vlm causal lm and reset ids on batch

  drbh

  committed
  77c81a2

• fix: adjust get_position_ids if not available and add required args to signatures

  drbh

  committed
  4f90db4

• fix: adjust resize case for qwen2_vl warmup

  drbh

  committed
  77eb07f

• fix: avoid qwen2 vl specific paths with qwen2

  drbh

  committed
  620769e