[AMD] Add AMD Nightly Performance & VLMs Accuracy Tests (#15500)
This commit is contained in:
@@ -53,8 +53,22 @@ for key in "${!ENV_MAP[@]}"; do
|
||||
ENV_ARGS+=("-e" "$key=${ENV_MAP[$key]}")
|
||||
done
|
||||
|
||||
# Run docker exec
|
||||
# Run docker exec with retry logic for HF network issues
|
||||
# First attempt: normal mode
|
||||
if docker exec \
|
||||
-w "$WORKDIR" \
|
||||
"${ENV_ARGS[@]}" \
|
||||
ci_sglang "$@"; then
|
||||
exit 0
|
||||
fi
|
||||
|
||||
FIRST_EXIT_CODE=$?
|
||||
echo "First attempt failed with exit code $FIRST_EXIT_CODE"
|
||||
echo "Retrying with HF_HUB_OFFLINE=1 (offline mode)..."
|
||||
|
||||
# Second attempt: force HF offline mode to avoid network timeouts
|
||||
docker exec \
|
||||
-w "$WORKDIR" \
|
||||
"${ENV_ARGS[@]}" \
|
||||
-e HF_HUB_OFFLINE=1 \
|
||||
ci_sglang "$@"
|
||||
|
||||
@@ -168,6 +168,8 @@ docker run -dt --user root --device=/dev/kfd ${DEVICE_FLAG} \
|
||||
--cap-add=SYS_PTRACE \
|
||||
-e HF_TOKEN="${HF_TOKEN:-}" \
|
||||
-e HF_HOME=/sgl-data/hf-cache \
|
||||
-e HF_HUB_ETAG_TIMEOUT=300 \
|
||||
-e HF_HUB_DOWNLOAD_TIMEOUT=300 \
|
||||
-e MIOPEN_USER_DB_PATH=/sgl-data/miopen-cache \
|
||||
-e MIOPEN_CUSTOM_CACHE_DIR=/sgl-data/miopen-cache \
|
||||
--security-opt seccomp=unconfined \
|
||||
|
||||
Reference in New Issue
Block a user