HCAI-Lab/w2-consensus-deepdive-unlearning-artifacts / social-data-attribution-w2 /scripts /slurm /eda /eda_array.sbatch
| #SBATCH --cpus-per-task=8 | |
| #SBATCH --mem=32G | |
| #SBATCH --time=12:00:00 | |
| #SBATCH --partition=cpu-small | |
| #SBATCH --array=0-0 | |
| #SBATCH --output=logs/dolma_eda/%A_%a.out | |
| #SBATCH --error=logs/dolma_eda/%A_%a.err | |
| set -euo pipefail | |
| mkdir -p logs/dolma_eda | |
| PROJECT_ROOT=${PROJECT_ROOT:-${SLURM_SUBMIT_DIR:-$PWD}} | |
| cd "$PROJECT_ROOT" | |
| PYTHONEXEC=${PYTHONEXEC:-.venv/bin/python} | |
| PYTHONPATH=${PYTHONPATH:-$PROJECT_ROOT/src} | |
| export PYTHONPATH | |
| MANIFEST=${MANIFEST:?} | |
| OUTDIR=${OUTDIR:-runs/dolma_enriched/eda} | |
| WORKERS_DIR=${WORKERS_DIR:-$OUTDIR/workers} | |
| VALUE_FIELDS=${VALUE_FIELDS:-"lang warc_content_type cc_dump"} | |
| TOP_K=${TOP_K:-50} | |
| JOINT_TOP_K=${JOINT_TOP_K:-200} | |
| HEATMAP_K=${HEATMAP_K:-25} | |
| mkdir -p "$WORKERS_DIR" | |
| LINE=$(sed -n "$((SLURM_ARRAY_TASK_ID+1))p" "$MANIFEST") | |
| if [[ -z "$LINE" ]]; then | |
| echo "No manifest entry for task $SLURM_ARRAY_TASK_ID" | |
| exit 1 | |
| fi | |
| "$PYTHONEXEC" -m dolma.eda.core \ | |
| --mode worker \ | |
| --manifest "$MANIFEST" \ | |
| --manifest-index "$SLURM_ARRAY_TASK_ID" \ | |
| --output-dir "$OUTDIR" \ | |
| --worker-output-dir "$WORKERS_DIR/$SLURM_ARRAY_TASK_ID" \ | |
| --value-fields $VALUE_FIELDS \ | |
| --top-k "$TOP_K" \ | |
| --joint-top-k "$JOINT_TOP_K" \ | |
| --heatmap-k "$HEATMAP_K" \ | |
| --no-report \ | |
| --no-write-png \ | |
| --resume | |
Xet Storage Details
- Size:
- 1.25 kB
- Xet hash:
- 6deeecb2814b83bfe88f5074cbca670e95f8667865205968c63ebd655aa666c0
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.