software-engineering/spark-vllm-docker
2
0
Fork 0
Code Issues Actions Packages Projects Releases Wiki Activity

Merge branch 'main' into pytorch-base

Browse Source
This commit is contained in:
Eugene Rakhmatulin
2026-01-30 17:06:29 -08:00
parent a4b524625a 4a4b4e7610
commit a6d6bafa69
2 changed files with 11 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
Dockerfile.mxfp4
View File

@@ -256,12 +256,9 @@ RUN mkdir -p tiktoken_encodings && \
wget -O tiktoken_encodings/cl100k_base.tiktoken "https://openaipublic.blob.core.windows.net/encodings/cl100k_base.tiktoken" wget -O tiktoken_encodings/cl100k_base.tiktoken "https://openaipublic.blob.core.windows.net/encodings/cl100k_base.tiktoken"
# Copy artifacts from Builder Stage # Copy artifacts from Builder Stage
# We copy the python packages and executables RUN --mount=type=bind,from=builder,source=/workspace/wheels,target=/mount/wheels \
# No need to copy source code, as it's already in the site-packages --mount=type=cache,id=uv-cache,target=/root/.cache/uv \
COPY --from=builder /workspace/wheels /workspace/wheels uv pip install /mount/wheels/*.whl
RUN --mount=type=cache,id=uv-cache,target=/root/.cache/uv \
uv pip install /workspace/wheels/*.whl && \
rm -rf /workspace/wheels
# Setup Env for Runtime # Setup Env for Runtime
ENV TORCH_CUDA_ARCH_LIST="12.0;12.1" ENV TORCH_CUDA_ARCH_LIST="12.0;12.1"

11
launch-cluster.sh
View File

@@ -144,24 +144,22 @@ for i in "${!MOD_PATHS[@]}"; do
done done
# --- Auto-Detection Logic --- # --- Auto-Detection Logic ---
# Source autodiscover module # Source autodiscover module
source "$(dirname "$0")/autodiscover.sh" source "$(dirname "$0")/autodiscover.sh"
# Perform auto-detection
detect_interfaces || exit 1
if [[ "$SOLO_MODE" == "true" ]]; then if [[ "$SOLO_MODE" == "true" ]]; then
if [[ -n "$NODES_ARG" ]]; then if [[ -n "$NODES_ARG" ]]; then
echo "Error: --solo is incompatible with -n/--nodes." echo "Error: --solo is incompatible with -n/--nodes."
exit 1 exit 1
fi fi
# Solo mode: skip node detection, just get local IP # Solo mode: skip node detection, just get local IP
detect_local_ip || exit 1 LOCAL_IP="127.0.0.1"
NODES_ARG="$LOCAL_IP" NODES_ARG="$LOCAL_IP"
PEER_NODES=() PEER_NODES=()
echo "Solo mode enabled. Skipping node detection." echo "Solo mode enabled. Skipping node detection."
else else
# Perform auto-detection
detect_interfaces || exit 1
detect_nodes || exit 1 detect_nodes || exit 1
fi fi
@@ -173,8 +171,11 @@ fi
# Split nodes into array # Split nodes into array
IFS=',' read -r -a ALL_NODES <<< "$NODES_ARG" IFS=',' read -r -a ALL_NODES <<< "$NODES_ARG"
if [[ "$SOLO_MODE" != "true" ]]; then
# Detect Head IP (Local IP) # Detect Head IP (Local IP)
detect_local_ip || exit 1 detect_local_ip || exit 1
fi
HEAD_IP="$LOCAL_IP" HEAD_IP="$LOCAL_IP"
# Verify HEAD_IP is in ALL_NODES # Verify HEAD_IP is in ALL_NODES