diff options
Diffstat (limited to 'scripts/submit_single_job.sh')
| -rwxr-xr-x | scripts/submit_single_job.sh | 32 |
1 files changed, 32 insertions, 0 deletions
diff --git a/scripts/submit_single_job.sh b/scripts/submit_single_job.sh new file mode 100755 index 0000000..7fe7492 --- /dev/null +++ b/scripts/submit_single_job.sh @@ -0,0 +1,32 @@ +#!/bin/bash +# submit_single_job.sh +# Submit a single training job +# Usage: ./submit_single_job.sh <precision_mode> <seed> +# Example: ./submit_single_job.sh bf16 1 + +PRECISION_MODE=${1:-"bf16"} +SEED=${2:-1} + +PROJECT_DIR="/projects/bfqt/users/yurenh2/ml-projects/rl-floating-noise" +cd "$PROJECT_DIR" + +mkdir -p results/slurm_logs + +JOB_NAME="rlvr_${PRECISION_MODE}_s${SEED}" + +echo "Submitting job: $JOB_NAME" +echo " Precision: $PRECISION_MODE" +echo " Seed: $SEED" + +JOB_ID=$(sbatch \ + --job-name="$JOB_NAME" \ + --export=ALL,PRECISION_MODE="$PRECISION_MODE",SEED="$SEED" \ + scripts/slurm_train.sh | awk '{print $4}') + +echo "" +echo "Submitted! Job ID: $JOB_ID" +echo "" +echo "Monitor with: squeue -j $JOB_ID" +echo "View output: tail -f results/slurm_logs/${JOB_NAME}_${JOB_ID}.out" +echo "Cancel: scancel $JOB_ID" + |
