qwen-dflash2-spark / scripts /Dockerfile.df2
cfontes's picture
Qwen3.8-27B DFlash2 on 2x DGX Spark: one-stop stack (135 tok/s C1, lm_head BF16 fix)
5be2256 verified
Raw
History Blame Contribute Delete
213 Bytes
FROM vllm/vllm-openai:v0.27.1-aarch64
COPY o.tgz /tmp/o.tgz
RUN mkdir -p /tmp/px && tar xzf /tmp/o.tgz -C /tmp/px && cp -r /tmp/px/vllm/* /usr/local/lib/python3.12/dist-packages/vllm/ && rm -rf /tmp/px /tmp/o.tgz