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

Merge branch 'main' into pr-62

Browse Source
This commit is contained in:
Eugene Rakhmatulin
2026-02-26 14:57:32 -08:00
parent a276a76be2 4593931421
commit e9aa411e6c
4 changed files with 15 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
Dockerfile
View File

@@ -90,7 +90,7 @@ RUN --mount=type=cache,id=repo-cache,target=/repo-cache \
echo "Cache hit: Fetching flashinfer updates..." && \ echo "Cache hit: Fetching flashinfer updates..." && \
cd flashinfer && \ cd flashinfer && \
git fetch origin && \ git fetch origin && \
git fetch origin --tags && \ git fetch origin --tags --force && \
(git checkout --detach origin/${FLASHINFER_REF} 2>/dev/null || git checkout ${FLASHINFER_REF}) && \ (git checkout --detach origin/${FLASHINFER_REF} 2>/dev/null || git checkout ${FLASHINFER_REF}) && \
git submodule update --init --recursive && \ git submodule update --init --recursive && \
git clean -fdx && \ git clean -fdx && \
@@ -153,7 +153,7 @@ RUN --mount=type=cache,id=repo-cache,target=/repo-cache \
echo "Cache hit: Fetching updates..." && \ echo "Cache hit: Fetching updates..." && \
cd vllm && \ cd vllm && \
git fetch origin && \ git fetch origin && \
git fetch origin --tags && \ git fetch origin --tags --force && \
(git checkout --detach origin/${VLLM_REF} 2>/dev/null || git checkout ${VLLM_REF}) && \ (git checkout --detach origin/${VLLM_REF} 2>/dev/null || git checkout ${VLLM_REF}) && \
git submodule update --init --recursive && \ git submodule update --init --recursive && \
git clean -fdx && \ git clean -fdx && \

4
hf-download.sh
View File

@@ -1,7 +1,7 @@
#!/bin/bash #!/bin/bash
set -e set -e
HUB_PATH="$HOME/.cache/huggingface/hub" HUB_PATH="${HF_HOME:-$HOME/.cache/huggingface}/hub"
# Default values # Default values
COPY_HOSTS=() COPY_HOSTS=()
@@ -42,7 +42,7 @@ copy_model_to_host() {
local host_copy_start host_copy_end host_copy_time local host_copy_start host_copy_end host_copy_time
host_copy_start=$(date +%s) host_copy_start=$(date +%s)
if rsync -av --progress "$model_dir" "${SSH_USER}@${host}:$HUB_PATH/"; then if rsync -av --mkpath --progress "$model_dir" "${SSH_USER}@${host}:$HUB_PATH/"; then
host_copy_end=$(date +%s) host_copy_end=$(date +%s)
host_copy_time=$((host_copy_end - host_copy_start)) host_copy_time=$((host_copy_end - host_copy_start))
printf "Copy to %s completed in %02d:%02d:%02d\n" "$host" $((host_copy_time/3600)) $((host_copy_time%3600/60)) $((host_copy_time%60)) printf "Copy to %s completed in %02d:%02d:%02d\n" "$host" $((host_copy_time/3600)) $((host_copy_time%3600/60)) $((host_copy_time%60))

3
launch-cluster.sh
View File

@@ -3,8 +3,9 @@
# Default Configuration # Default Configuration
IMAGE_NAME="vllm-node" IMAGE_NAME="vllm-node"
DEFAULT_CONTAINER_NAME="vllm_node" DEFAULT_CONTAINER_NAME="vllm_node"
HF_CACHE_DIR="${HF_HOME:-$HOME/.cache/huggingface}"
# Modify these if you want to pass additional docker args or set VLLM_SPARK_EXTRA_DOCKER_ARGS variable # Modify these if you want to pass additional docker args or set VLLM_SPARK_EXTRA_DOCKER_ARGS variable
DOCKER_ARGS="-e NCCL_IGNORE_CPU_AFFINITY=1 -v $HOME/.cache/huggingface:/root/.cache/huggingface" DOCKER_ARGS="-e NCCL_IGNORE_CPU_AFFINITY=1 -v $HF_CACHE_DIR:/root/.cache/huggingface"
# Append additional arguments from environment variable # Append additional arguments from environment variable
if [[ -n "$VLLM_SPARK_EXTRA_DOCKER_ARGS" ]]; then if [[ -n "$VLLM_SPARK_EXTRA_DOCKER_ARGS" ]]; then

9
mods/fix-qwen3-next-autoround/run.sh Normal file
View File

@@ -0,0 +1,9 @@
#!/bin/bash
set -e
echo "Reverting PR #35156"
if curl -L https://patch-diff.githubusercontent.com/raw/vllm-project/vllm/pull/35156.diff | patch -p1 -R -d /usr/local/lib/python3.12/dist-packages; then
echo " OK"
else
echo " Patch can't be reversed, skipping"
fi