Datasets:
Formats:
parquet
Languages:
English
Size:
10M - 100M
Tags:
biology
chemistry
drug-discovery
clinical-trials
protein-protein-interaction
gene-essentiality
License:
| # Submit CT LLM Gemini jobs SEQUENTIALLY to avoid 250 RPD contention. | |
| # | |
| # The free tier only allows 250 requests/day. Running 16 jobs concurrently | |
| # means each gets ~15 RPD (useless). Sequential gives each job the full | |
| # 250 RPD quota. The benchmark script auto-resumes from existing predictions. | |
| # | |
| # Order: smallest tasks first (L3=160 → L4=400 → L2=400 → L1=900) | |
| # Expected timeline at 250 RPD: ~30 days total | |
| # | |
| # Usage: | |
| # bash slurm/submit_ct_gemini_sequential.sh [--dry-run] | |
| set -euo pipefail | |
| SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)" | |
| SBATCH="/opt/ohpc/pub/software/slurm/24.05.2/bin/sbatch" | |
| SLURM_TMPL="${SCRIPT_DIR}/run_ct_llm_gemini.slurm" | |
| DRY_RUN=false | |
| if [[ "${1:-}" == "--dry-run" ]]; then | |
| DRY_RUN=true | |
| echo "=== DRY RUN ===" | |
| fi | |
| MODEL="gemini-2.5-flash" | |
| # Job list: smallest tasks first for fastest partial results | |
| # Format: TASK CONFIG FS | |
| JOBS=( | |
| "ct-l3 zero-shot 0" | |
| "ct-l3 3-shot 0" | |
| "ct-l3 3-shot 1" | |
| "ct-l3 3-shot 2" | |
| "ct-l4 zero-shot 0" | |
| "ct-l4 3-shot 0" | |
| "ct-l4 3-shot 1" | |
| "ct-l4 3-shot 2" | |
| "ct-l2 zero-shot 0" | |
| "ct-l2 3-shot 0" | |
| "ct-l2 3-shot 1" | |
| "ct-l2 3-shot 2" | |
| "ct-l1 zero-shot 0" | |
| "ct-l1 3-shot 0" | |
| "ct-l1 3-shot 1" | |
| "ct-l1 3-shot 2" | |
| ) | |
| PREV_JOBID="" | |
| TOTAL=0 | |
| for entry in "${JOBS[@]}"; do | |
| read -r TASK CONFIG FS <<< "$entry" | |
| JOB_NAME="ct_${TASK}_${MODEL}_${CONFIG}_fs${FS}" | |
| TOTAL=$((TOTAL + 1)) | |
| if $DRY_RUN; then | |
| if [[ -n "$PREV_JOBID" ]]; then | |
| echo " [DRY] $JOB_NAME (after job $PREV_JOBID)" | |
| else | |
| echo " [DRY] $JOB_NAME (first job)" | |
| fi | |
| PREV_JOBID="DRY_${TOTAL}" | |
| else | |
| DEP_FLAG="" | |
| if [[ -n "$PREV_JOBID" ]]; then | |
| DEP_FLAG="--dependency=afterany:${PREV_JOBID}" | |
| fi | |
| JOBID=$($SBATCH \ | |
| --job-name="$JOB_NAME" \ | |
| --parsable \ | |
| $DEP_FLAG \ | |
| --export="ALL,TASK=$TASK,MODEL=$MODEL,CONFIG=$CONFIG,FS=$FS" \ | |
| "$SLURM_TMPL") | |
| echo " Submitted: $JOB_NAME → $JOBID (dep: ${PREV_JOBID:-none})" | |
| PREV_JOBID="$JOBID" | |
| fi | |
| done | |
| echo "" | |
| echo "Total jobs: $TOTAL (sequential chain)" | |
| if $DRY_RUN; then | |
| echo "(Dry run — no jobs submitted)" | |
| echo "Estimated time at 250 RPD: ~30 days" | |
| else | |
| echo "Chain started. Each job runs after the previous completes." | |
| echo "Monitor: squeue -u \$USER | grep gemini" | |
| fi | |