Dockerfile.rocm 2.51 KB
Newer Older
1
# Usage (to build SGLang ROCm docker image):
2
#   docker build --build-arg SGL_BRANCH=v0.4.3.post2 -t v0.4.3.post2-rocm630 -f Dockerfile.rocm .
3
4

# default base image
HAI's avatar
HAI committed
5
ARG BASE_IMAGE="rocm/vllm-dev:20250114"
6
7
8
9
10

FROM $BASE_IMAGE AS base
USER root

WORKDIR /sgl-workspace
11
ARG BUILD_TYPE=all
12
13
14
15
ARG SGL_REPO="https://github.com/sgl-project/sglang"
ENV SGL_DEFAULT="main"
ARG SGL_BRANCH=${SGL_DEFAULT}

HAI's avatar
HAI committed
16
17
ARG TRITON_REPO="https://github.com/ROCm/triton.git"
ARG TRITON_COMMIT="improve_fa_decode_3.0.0"
18

19

20
21
ARG AITER_REPO="https://github.com/ROCm/aiter.git"
ARG AITER_COMMIT="dev/testx"
22

23
24
25
26
27
28
29
30
RUN git clone ${SGL_REPO} \
    && cd sglang \
    && if [ "${SGL_BRANCH}" = ${SGL_DEFAULT} ]; then \
         echo "Using ${SGL_DEFAULT}, default branch."; \
       else \
         echo "Using ${SGL_BRANCH} branch."; \
         git checkout ${SGL_BRANCH}; \
       fi \
31
32
33
    && cd sgl-kernel \
    && python setup_rocm.py install \
    && cd .. \
34
35
36
37
38
39
40
41
42
    && if [ "$BUILD_TYPE" = "srt" ]; then \
         python -m pip --no-cache-dir install -e "python[srt_hip]"; \
       else \
         python -m pip --no-cache-dir install -e "python[all_hip]"; \
       fi

RUN cp -r /sgl-workspace/sglang /sglang
RUN python -m pip cache purge

kk's avatar
kk committed
43
44
45
46
47
48
RUN pip install IPython \
    && pip install orjson \
    && pip install python-multipart \
    && pip install torchao \
    && pip install pybind11

49
50
51
52
53
54
55
RUN pip uninstall -y triton
RUN git clone ${TRITON_REPO} \
    && cd triton \
    && git checkout ${TRITON_COMMIT} \
    && cd python \
    && python3 setup.py install

56
57
58
RUN git clone ${AITER_REPO} \
    && cd aiter \
    && git checkout ${AITER_COMMIT} \
59
60
61
    && git submodule update --init --recursive \
    && PREBUILD_KERNELS=1 GPU_ARCHS=gfx942 python3 setup.py develop

62
63
64
65
66
# Copy config files to support MI300X in virtualized environments (MI300X_VF).  Symlinks will not be created in image build.
RUN find /sgl-workspace/sglang/python/sglang/srt/layers/quantization/configs/ \
         /sgl-workspace/sglang/python/sglang/srt/layers/moe/fused_moe_triton/configs/ \
         -type f -name '*MI300X*' | xargs -I {} sh -c 'vf_config=$(echo "$1" | sed "s/MI300X/MI300X_VF/"); cp "$1" "$vf_config"' -- {}

67
68
69
# Performance environment variable.

ENV HIP_FORCE_DEV_KERNARG=1
HAI's avatar
HAI committed
70
ENV HSA_NO_SCRATCH_RECLAIM=1
71
ENV SGLANG_SET_CPU_AFFINITY=1
72
73
74
75
76
77
78
79
ENV SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1
ENV NCCL_MIN_NCHANNELS=112

ENV MOE_PADDING=1
ENV VLLM_FP8_PADDING=1
ENV VLLM_FP8_ACT_PADDING=1
ENV VLLM_FP8_WEIGHT_PADDING=1
ENV VLLM_FP8_REDUCE_CONV=1
80
81
ENV TORCHINDUCTOR_MAX_AUTOTUNE=1
ENV TORCHINDUCTOR_MAX_AUTOTUNE_POINTWISE=1
82
83

CMD ["/bin/bash"]