code update

code/slurm_start.sh

@@ -0,0 +1,58 @@
+#!/bin/bash
+
+# Parameter check
+if [[ $# -ne 5 ]]; then
+    echo "Usage: $0 <run_config_module> <partition> <gpu_type> <num_gpus> <num_cpus_per_gpu>"
+    exit 1
+fi
+
+# Assign parameters
+run_config=$1
+partition=$2
+gpu_type=$3
+num_gpus=$4
+num_cpus_per_gpu=$5
+
+# Derive run name
+run_name=$(basename "$run_config")
+run_name="${run_name%.*}"
+
+# Confirm inputs
+echo "Starting run with configuration: $run_config"
+echo "Partition: $partition"
+echo "GPU type: $gpu_type"
+echo "Number of GPUs: $num_gpus"
+echo "Number of CPUs per GPU: $num_cpus_per_gpu"
+
+# Set paths
+current_dir_path=$(dirname "$(realpath "$0")")
+venv_path=$(realpath "$current_dir_path/../../venv/bin/activate")
+
+# Create log directory
+log_dir="/work/rr41qemu-MA/logs/${run_name}_$(date +%Y-%m-%d_%H-%M-%S)"
+mkdir -p "$log_dir"
+echo "Logging to $log_dir"
+
+# Submit job
+sbatch <<EOF
+#!/bin/bash
+#SBATCH --job-name=$run_name
+#SBATCH --output=$log_dir/log.out
+#SBATCH --error=$log_dir/log.err
+#SBATCH --time=48:00:00
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=$(($num_gpus * $num_cpus_per_gpu))
+#SBATCH --mem=32G
+#SBATCH --partition=$partition
+#SBATCH --gpus=$gpu_type:$num_gpus
+
+echo "Loading python virtual environment..."
+source $venv_path
+
+echo "Loading python 3.10..."
+module load Python/3.10.4-GCCcore-11.3.0
+
+cd $current_dir_path
+
+torchrun --nproc_per_node=$num_gpus training_wrapper.py $run_config --item_limit -1
+EOF