Skip to content

Commit fb3640c

Browse files
committed
$CUDA_HOME_PATH
1 parent da61165 commit fb3640c

File tree

2 files changed

+31
-1
lines changed

2 files changed

+31
-1
lines changed

.azure/gpu-tests-fabric.yml

Lines changed: 15 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -86,8 +86,23 @@ jobs:
8686
condition: endsWith(variables['Agent.JobName'], 'future')
8787
displayName: "extend env. vars 4 future"
8888
89+
- bash: |
90+
set -e
91+
# Try to find nvcc and set CUDA_HOME so that deepspeed can be installed
92+
NVCC_PATH=$(find /usr/local -name "nvcc" | head -n 1)
93+
if [[ -z "$NVCC_PATH" ]]; then
94+
echo "nvcc not found, skipping setting CUDA_HOME"
95+
else
96+
CUDA_HOME_PATH=$(dirname $(dirname "$NVCC_PATH"))
97+
echo "Found nvcc at: ${NVCC_PATH}"
98+
echo "Setting CUDA_HOME to: ${CUDA_HOME_PATH}"
99+
echo "##vso[task.setvariable variable=CUDA_HOME]${CUDA_HOME_PATH}"
100+
fi
101+
displayName: "Set CUDA_HOME for DeepSpeed"
102+
89103
- bash: |
90104
echo $(DEVICES)
105+
echo $CUDA_HOME_PATH
91106
echo $CUDA_VISIBLE_DEVICES
92107
echo $CUDA_VERSION_MM
93108
echo $TORCH_URL

.azure/gpu-tests-pytorch.yml

Lines changed: 16 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -90,8 +90,23 @@ jobs:
9090
condition: endsWith(variables['Agent.JobName'], 'future')
9191
displayName: "extend env. vars 4 future"
9292
93+
- bash: |
94+
set -e
95+
# Try to find nvcc and set CUDA_HOME so that deepspeed can be installed
96+
NVCC_PATH=$(find /usr/local -name "nvcc" | head -n 1)
97+
if [[ -z "$NVCC_PATH" ]]; then
98+
echo "nvcc not found, skipping setting CUDA_HOME"
99+
else
100+
CUDA_HOME_PATH=$(dirname $(dirname "$NVCC_PATH"))
101+
echo "Found nvcc at: ${NVCC_PATH}"
102+
echo "Setting CUDA_HOME to: ${CUDA_HOME_PATH}"
103+
echo "##vso[task.setvariable variable=CUDA_HOME]${CUDA_HOME_PATH}"
104+
fi
105+
displayName: "Set CUDA_HOME for DeepSpeed"
106+
93107
- bash: |
94108
echo $(DEVICES)
109+
echo $CUDA_HOME_PATH
95110
echo $CUDA_VISIBLE_DEVICES
96111
echo $CUDA_VERSION_MM
97112
echo $TORCH_URL
@@ -212,4 +227,4 @@ jobs:
212227
workingDirectory: examples
213228
env:
214229
PL_USE_MOCKED_MNIST: "1"
215-
displayName: "Testing: PyTorch examples"
230+
displayName: "Testing: PyTorch examples"

0 commit comments

Comments
 (0)