From ae247fdafe72c21662c8035abd635130f8754b93 Mon Sep 17 00:00:00 2001
From: milan jaros <milan.jaros@vsb.cz>
Date: Wed, 31 Jan 2024 12:58:18 +0200
Subject: [PATCH] fix srun in scripts, fix path to singularity
---
README.md | 11 ++---------
.../05-install-container-torch2.0.1-rocm5.5.1.sh | 3 ++-
.../01-simple-test-direct-torch1.13.1-rocm5.2.3.sh | 2 +-
.../02-simple-test-venv-torch1.13.1-rocm5.2.3.sh | 2 +-
.../03-simple-test-conda-torch1.13.1-rocm5.2.3.sh | 2 +-
.../04-simple-test-source-torch1.13.1-rocm5.2.3.sh | 2 +-
.../05-simple-test-container-torch2.0.1-rocm5.5.1.sh | 5 +++--
7 files changed, 11 insertions(+), 16 deletions(-)
diff --git a/README.md b/README.md
index 0b91e9d..14c55b0 100644
--- a/README.md
+++ b/README.md
@@ -33,17 +33,10 @@ module load rocm/5.2.3.lua
## Pytorch simple test (calling of torch.cuda.device_count())
-### Base environment
+### Run an interactive job
```
-module purge
-module load CrayEnv
-module load PrgEnv-cray/8.3.3
-module load craype-accel-amd-gfx90a
-module load cray-python
-
-# Default ROCm – more recent versions are preferable (e.g. ROCm 5.6.0)
-module load rocm/5.2.3.lua
+salloc -A project_XXX --partition=standard-g -N 1 -n 1 --gres=gpu:8 -t 01:00:00
```
### Scripts
diff --git a/scripts/install/05-install-container-torch2.0.1-rocm5.5.1.sh b/scripts/install/05-install-container-torch2.0.1-rocm5.5.1.sh
index 51ae02e..22dede6 100755
--- a/scripts/install/05-install-container-torch2.0.1-rocm5.5.1.sh
+++ b/scripts/install/05-install-container-torch2.0.1-rocm5.5.1.sh
@@ -14,7 +14,8 @@ wd=$(pwd)
set -x
# download from https://hub.docker.com/r/rocm/pytorch or use from Samuel Antao:
-SIF=/pfs/lustrep2/projappl/project_462000125/samantao-public/containers/lumi-pytorch-rocm-5.5.1-python-3.10-pytorch-v2.0.1-dockerhash-4305da4654f4.sif
+#SIF=/pfs/lustrep2/projappl/project_462000125/samantao-public/containers/lumi-pytorch-rocm-5.5.1-python-3.10-pytorch-v2.0.1-dockerhash-4305da4654f4.sif
+SIF=/pfs/lustrep2/projappl/project_462000125/samantao-public/containers/lumi-pytorch-rocm-5.5.1-python-3.10-pytorch-v2.0.1-dockerhash-d55f9163ed80.sif
rm -rf $wd/run-me.sh
cat > $wd/run-me.sh << EOF
diff --git a/scripts/tests/01-simple-test-direct-torch1.13.1-rocm5.2.3.sh b/scripts/tests/01-simple-test-direct-torch1.13.1-rocm5.2.3.sh
index 00b096e..9159f91 100755
--- a/scripts/tests/01-simple-test-direct-torch1.13.1-rocm5.2.3.sh
+++ b/scripts/tests/01-simple-test-direct-torch1.13.1-rocm5.2.3.sh
@@ -28,5 +28,5 @@ if [ ! -d $wd/pip-installs ] ; then
fi
PYTHONPATH=$wd/pip-installs \
- srun -n1 --gpus 8 \
+ srun -n 1 \
python -c 'import torch; print("I have this many devices:", torch.cuda.device_count())'
diff --git a/scripts/tests/02-simple-test-venv-torch1.13.1-rocm5.2.3.sh b/scripts/tests/02-simple-test-venv-torch1.13.1-rocm5.2.3.sh
index 1696487..be9bf2e 100755
--- a/scripts/tests/02-simple-test-venv-torch1.13.1-rocm5.2.3.sh
+++ b/scripts/tests/02-simple-test-venv-torch1.13.1-rocm5.2.3.sh
@@ -29,5 +29,5 @@ else
source cray-python-virtualenv/bin/activate
fi
-srun -n1 --gpus 8 \
+srun -n 1 \
python -c 'import torch; print("I have this many devices:", torch.cuda.device_count())'
diff --git a/scripts/tests/03-simple-test-conda-torch1.13.1-rocm5.2.3.sh b/scripts/tests/03-simple-test-conda-torch1.13.1-rocm5.2.3.sh
index b11da4c..56efb63 100755
--- a/scripts/tests/03-simple-test-conda-torch1.13.1-rocm5.2.3.sh
+++ b/scripts/tests/03-simple-test-conda-torch1.13.1-rocm5.2.3.sh
@@ -28,5 +28,5 @@ else
source $wd/miniconda3/bin/activate pytorch
fi
-srun -n1 --gpus 8 \
+srun -n 1 \
python -c 'import torch; print("I have this many devices:", torch.cuda.device_count())'
diff --git a/scripts/tests/04-simple-test-source-torch1.13.1-rocm5.2.3.sh b/scripts/tests/04-simple-test-source-torch1.13.1-rocm5.2.3.sh
index eeb0d30..03cf58d 100755
--- a/scripts/tests/04-simple-test-source-torch1.13.1-rocm5.2.3.sh
+++ b/scripts/tests/04-simple-test-source-torch1.13.1-rocm5.2.3.sh
@@ -31,5 +31,5 @@ fi
# Make sure conda libs can be loaded.
export LD_LIBRARY_PATH=$CONDA_PREFIX/lib:$LD_LIBRARY_PATH
-srun -n1 --gpus 8 \
+srun -n 1 \
python -c 'import torch; print("I have this many devices:", torch.cuda.device_count())'
diff --git a/scripts/tests/05-simple-test-container-torch2.0.1-rocm5.5.1.sh b/scripts/tests/05-simple-test-container-torch2.0.1-rocm5.5.1.sh
index d16c020..c03e814 100755
--- a/scripts/tests/05-simple-test-container-torch2.0.1-rocm5.5.1.sh
+++ b/scripts/tests/05-simple-test-container-torch2.0.1-rocm5.5.1.sh
@@ -14,7 +14,8 @@ wd=$(pwd)
set -x
# download from https://hub.docker.com/r/rocm/pytorch or use from Samuel Antao:
-SIF=/pfs/lustrep2/projappl/project_462000125/samantao-public/containers/lumi-pytorch-rocm-5.5.1-python-3.10-pytorch-v2.0.1-dockerhash-4305da4654f4.sif
+#SIF=/pfs/lustrep2/projappl/project_462000125/samantao-public/containers/lumi-pytorch-rocm-5.5.1-python-3.10-pytorch-v2.0.1-dockerhash-4305da4654f4.sif
+SIF=/pfs/lustrep2/projappl/project_462000125/samantao-public/containers/lumi-pytorch-rocm-5.5.1-python-3.10-pytorch-v2.0.1-dockerhash-d55f9163ed80.sif
# check script
if [ ! -f $wd/run-me.sh ] ; then
@@ -22,7 +23,7 @@ if [ ! -f $wd/run-me.sh ] ; then
exit 1
fi
-srun -n1 --gpus 8 \
+srun -n 1 \
singularity exec \
-B /var/spool/slurmd:/var/spool/slurmd \
-B /opt/cray:/opt/cray \
--
GitLab