chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[AMD] Add 8-GPU MX35X test running DSR1-MXFP4 model for AMD CI (#13602)

Browse Source
This commit is contained in:
Hubert Lu
2026-01-07 01:43:11 -08:00
committed by GitHub
parent 48381c3b6d
commit b86bbf841e
5 changed files with 243 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
scripts/ci/amd_ci_exec.sh
View File

@@ -60,9 +60,10 @@ if docker exec \
"${ENV_ARGS[@]}" \
ci_sglang "$@"; then
exit 0
else
FIRST_EXIT_CODE=$?
fi
FIRST_EXIT_CODE=$?
echo "First attempt failed with exit code $FIRST_EXIT_CODE"
echo "Retrying with HF_HUB_OFFLINE=1 (offline mode)..."

5
scripts/ci/amd_ci_install_dependency.sh
View File

@@ -99,10 +99,11 @@ case "${GPU_ARCH}" in
mi35x)
echo "Runner uses ${GPU_ARCH}; will fetch mi35x image."
docker exec ci_sglang rm -rf python/pyproject.toml && mv python/pyproject_other.toml python/pyproject.toml
install_with_retry docker exec ci_sglang pip install --cache-dir=/sgl-data/pip-cache -e "python[${EXTRAS}]" --no-deps # TODO: only for mi35x
# Follow the same dependency installation flow as mi30x/mi300/mi325.
install_with_retry docker exec ci_sglang pip install --cache-dir=/sgl-data/pip-cache -e "python[${EXTRAS}]"
# For lmms_evals evaluating MMMU
docker exec -w / ci_sglang git clone --branch v0.4.1 --depth 1 https://github.com/EvolvingLMMs-Lab/lmms-eval.git
install_with_retry docker exec -w /lmms-eval ci_sglang pip install --cache-dir=/sgl-data/pip-cache -e . --no-deps
install_with_retry docker exec -w /lmms-eval ci_sglang pip install --cache-dir=/sgl-data/pip-cache -e .
;;
mi30x|mi300|mi325)
echo "Runner uses ${GPU_ARCH}; will fetch mi30x image."