microsoft/mscclpp
1
0
Fork 0
mirror of https://github.com/microsoft/mscclpp.git synced 2026-05-12 01:10:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix remote run and clean up files

Browse Source
This commit is contained in:
empyreus
2026-03-27 21:13:07 +00:00
parent 324254d57c
commit 38552a6f9c
2 changed files with 8 additions and 56 deletions
Download Patch File Download Diff File Expand all files Collapse all files

60
.azure-pipelines/templates/sglang-test.yml
View File

@@ -66,59 +66,7 @@ steps:
remoteScript: |
python -m sglang.bench_one_batch --model-path Qwen/Qwen3-8B --batch <bs> --input-len 256 --output-len 256 --tp-size 8 --disable-custom-all-reduce --enable-mscclpp
# - task: Bash@3
# name: SGLangSetup
# displayName: SGLang Setup
# inputs:
# targetType: inline
# script: |
# hostname
# set -e
# HOSTFILE=$(System.DefaultWorkingDirectory)/test/deploy/hostfile_ci
# SSH_OPTION="StrictHostKeyChecking=no"
# KeyFilePath=${SSHKEYFILE_SECUREFILEPATH}
# : > azureuser@10.0.0.4
# tail -f azureuser@10.0.0.4 &
# CHILD_PID=$!
# parallel-ssh -o . -t 0 -h ${HOSTFILE} -x "-i ${KeyFilePath}" \
# -O $SSH_OPTION 'sudo docker run -itd --name=mscclpp-sglang-test --privileged --net=host --ipc=host --gpus=all -w /root -v /mnt:/mnt lmsysorg/sglang:latest bash && \
# sudo docker exec -t mscclpp-sglang-test bash -c " \
# python3 -m venv /root/venv && \
# git clone https://github.com/microsoft/mscclpp.git && \
# cd mscclpp && \
# mkdir build && \
# cd build && \
# cmake -DCMAKE_BUILD_TYPE=Release .. && \
# make -j && \
# cd .. && \
# /root/venv/bin/pip install . && \
# /root/venv/bin/pip install -r ./python/requirements_cuda12.txt \
# "'
# kill $CHILD_PID
# workingDirectory: '$(System.DefaultWorkingDirectory)'
# - task: Bash@3
# name: AllGatherTest
# displayName: Run mscclpp AllGather test
# inputs:
# targetType: inline
# script: |
# set -e
# HOSTFILE=$(System.DefaultWorkingDirectory)/test/deploy/hostfile_ci
# SSH_OPTION="StrictHostKeyChecking=no"
# KeyFilePath=${SSHKEYFILE_SECUREFILEPATH}
# : > azureuser@10.0.0.4
# tail -f azureuser@10.0.0.4 &
# CHILD_PID=$!
# parallel-ssh -o . -t 0 -h ${HOSTFILE} -x "-i ${KeyFilePath}" \
# -O $SSH_OPTION 'sudo docker exec -t mscclpp-sglang-test bash -c " \
# export PATH=/usr/local/mpi/bin:\$PATH; \
# export LD_LIBRARY_PATH=/root/mscclpp/build/lib:\$LD_LIBRARY_PATH; \
# echo Running on \$(hostname -i); \
# kill $CHILD_PID
# workingDirectory: '$(System.DefaultWorkingDirectory)'
- template: stop.yml
parameters:
subscription: ${{ parameters.subscription }}
vmssName: ${{ parameters.vmssName }}
# - template: stop.yml
# parameters:
# subscription: ${{ parameters.subscription }}
# vmssName: ${{ parameters.vmssName }}

4
test/deploy/run-remote.sh
View File

@@ -105,6 +105,10 @@ if $USE_DOCKER; then
parallel-ssh -i "${PSSH_COMMON[@]}" \
"sudo docker rm -f ${CONTAINER_NAME} 2>/dev/null; \
sudo docker run -itd --name=${CONTAINER_NAME} --privileged --net=host --ipc=host --gpus=all -w /root -v /mnt:/mnt lmsysorg/sglang:latest bash"
INNER="set -euxo pipefail;"
INNER+=" CMD_B64='${CMD_B64}';"
INNER+=" printf '%s' \\\"\\\$CMD_B64\\\" | base64 -d | bash -euxo pipefail"
else
INNER="set -euxo pipefail;"
INNER+=" cd /root/mscclpp;"