Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[CI] Assign larger /dev/shm to NCCL (#5966)

Browse Source 
* [CI] Assign larger /dev/shm to NCCL

* Use 10.2 artifact to run multi-GPU Python tests

* Add CUDA 10.0 -> 11.0 cross-version test; remove CUDA 10.0 target
This commit is contained in:
Philip Hyunsu Cho 2020-07-31 10:05:04 -07:00 committed by GitHub
parent 3fcfaad577
commit 5f3c811e84
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
Jenkinsfile vendored
View File

@ -89,10 +89,10 @@ pipeline {
script { script {
parallel ([ parallel ([
'test-python-cpu': { TestPythonCPU() }, 'test-python-cpu': { TestPythonCPU() },
'test-python-gpu-cuda10.0': { TestPythonGPU(host_cuda_version: '10.0') },
'test-python-gpu-cuda10.2': { TestPythonGPU(host_cuda_version: '10.2') }, 'test-python-gpu-cuda10.2': { TestPythonGPU(host_cuda_version: '10.2') },
'test-python-gpu-cuda11.0-cross': { TestPythonGPU(artifact_cuda_version: '10.0', host_cuda_version: '11.0') },
'test-python-gpu-cuda11.0': { TestPythonGPU(artifact_cuda_version: '11.0', host_cuda_version: '11.0') }, 'test-python-gpu-cuda11.0': { TestPythonGPU(artifact_cuda_version: '11.0', host_cuda_version: '11.0') },
'test-python-mgpu-cuda10.2': { TestPythonGPU(host_cuda_version: '10.2', multi_gpu: true) }, 'test-python-mgpu-cuda10.2': { TestPythonGPU(artifact_cuda_version: '10.2', host_cuda_version: '10.2', multi_gpu: true) },
'test-cpp-gpu-cuda10.2': { TestCppGPU(artifact_cuda_version: '10.2', host_cuda_version: '10.2') }, 'test-cpp-gpu-cuda10.2': { TestCppGPU(artifact_cuda_version: '10.2', host_cuda_version: '10.2') },
'test-cpp-gpu-cuda11.0': { TestCppGPU(artifact_cuda_version: '11.0', host_cuda_version: '11.0') }, 'test-cpp-gpu-cuda11.0': { TestCppGPU(artifact_cuda_version: '11.0', host_cuda_version: '11.0') },
'test-jvm-jdk8-cuda10.0': { CrossTestJVMwithJDKGPU(artifact_cuda_version: '10.0', host_cuda_version: '10.0') }, 'test-jvm-jdk8-cuda10.0': { CrossTestJVMwithJDKGPU(artifact_cuda_version: '10.0', host_cuda_version: '10.0') },
@ -368,8 +368,10 @@ def TestPythonGPU(args) {
def docker_args = "--build-arg CUDA_VERSION=${args.host_cuda_version}" def docker_args = "--build-arg CUDA_VERSION=${args.host_cuda_version}"
if (args.multi_gpu) { if (args.multi_gpu) {
echo "Using multiple GPUs" echo "Using multiple GPUs"
// Allocate extra space in /dev/shm to enable NCCL
def docker_extra_params = "CI_DOCKER_EXTRA_PARAMS_INIT='--shm-size=4g'"
sh """ sh """
${dockerRun} ${container_type} ${docker_binary} ${docker_args} tests/ci_build/test_python.sh mgpu ${docker_extra_params} ${dockerRun} ${container_type} ${docker_binary} ${docker_args} tests/ci_build/test_python.sh mgpu
""" """
} else { } else {
echo "Using a single GPU" echo "Using a single GPU"