Code

feat: improve qwen2-vl startup #2802

Code

Open

drbh wants to merge 3 commits intomainhuggingface/text-generation-inference:mainfrom improve-qwen2-vl-warmuphuggingface/text-generation-inference:improve-qwen2-vl-warmupCopy head branch name to clipboard

+101-24

Commits

Commits on Jan 7, 2025

• feat: tokenize each request individually and increase warmup image size

  drbh

  committed
  901156c

  Copy full SHA for 901156c

• feat: adjust rotary embed and avoid cuda graphs of size 2 and smaller

  drbh

  committed
  90b83e2

  Copy full SHA for 90b83e2

• fix: address image resize and rebase changes

  drbh

  committed
  d671f6e

  Copy full SHA for d671f6e