v0.7.0-fix-xcomposer2d5

fix default vl template

fix-llava-hf

fix llava model when input images have size of (x, 1)

vision-slow

fix slow inference for vlm models with pytorch backend

peer_fix

v0.0.4

Merge remote-tracking branch 'zl/peer-allocator' into mt-batch

plora_20240605

fix penalty

test cuda-12

v0.0.3

add cuda version to release page

test cuda12

Merge pull request #7 from irexyc/use-cuda12

default use cuda12

permanent link

Merge pull request #5 from irexyc/cuda12-whl-release

add cuda12-whl-release ci