Qwen2VL_Dockerfile

ARG CUDA_VERSION=cu12

FROM openmmlab/lmdeploy:latest-cu12 AS cu12
ENV CUDA_VERSION_SHORT=cu123

FROM openmmlab/lmdeploy:latest-cu11 AS cu11
ENV CUDA_VERSION_SHORT=cu118

FROM ${CUDA_VERSION} AS final

# we use transformers to load vision part of qwen2_vl and it needs transformers > v4.44.2
RUN python3 -m pip install git+https://github.com/huggingface/transformers.git

RUN python3 -m pip install qwen_vl_utils