code/slurm_logs/eval_paper_one_config.sbatch

#!/bin/bash
#SBATCH --job-name=vl
#SBATCH --partition=gpu-large
#SBATCH --qos=batch-long
#SBATCH --gres=gpu:1
#SBATCH --cpus-per-task=4
#SBATCH --mem=120G
#SBATCH --time=06:00:00
#SBATCH --output=/weka/s225250685/mats-tist/slurm_logs/eval_paper_%j.out

# Parallel eval of ONE paper-format config — selected by $LABEL env var:
#   LABEL=COLLAB | INDEP
# Output written to paper_${LABEL}_par_passAt8_bird_dev.jsonl to avoid race with 88933.

set -u; cd /weka/s225250685/mats-tist
set -a; source .env; set +a
export HF_HOME=/weka/s225250685/Huggingface HF_HUB_CACHE=/weka/s225250685/Huggingface/hub
export DB_EXEC_API_DISABLE=1 PYTHONNOUSERSITE=1 NO_PROXY=localhost,127.0.0.1
export PYTHONPATH=/weka/s225250685/mats-tist TOKENIZERS_PARALLELISM=false
PY=/weka/s225250685/conda-envs/handbook/bin/python
VLLM=/weka/s225250685/conda-envs/handbook/bin/vllm
AH=/weka/s225250685/mats-tist/alignment-handbook

PLANNER=$AH/output/sft-planner-3B-griffith-v4
FIXER=$AH/output/sft-fixer-llama1b-griffith-v5

case "$LABEL" in
    COLLAB)
        VAL_SEL=$AH/output/orpo-val-sel-collab-paper
        VAL_COND=$AH/output/orpo-val-cond-collab-paper
        ;;
    INDEP)
        VAL_SEL=$AH/output/orpo-val-sel-indep-paper
        VAL_COND=$AH/output/orpo-val-cond-indep-paper
        ;;
    *) echo "FATAL: unknown LABEL=$LABEL"; exit 1 ;;
esac

kill_vllm() { pkill -9 -f "vllm serve" 2>/dev/null || true; sleep 5; }
trap kill_vllm EXIT
wait_url() { for i in {1..180}; do curl --noproxy '*' -fs "$1" >/dev/null 2>&1 && return 0; sleep 5; done; }

echo "[$(date)] $LABEL pipeline booting on GPU..."
$VLLM serve "$PLANNER" --served-model-name planner --port 8100 \
    --dtype bfloat16 --gpu-memory-utilization 0.30 --enforce-eager --max-model-len 8192 \
    > /tmp/eval_p_$$ 2>&1 &
wait_url http://localhost:8100/v1/models && echo "  planner ready"
$VLLM serve "$VAL_SEL" --served-model-name validator_sel --port 8101 \
    --dtype bfloat16 --gpu-memory-utilization 0.12 --enforce-eager --max-model-len 6144 \
    > /tmp/eval_vs_$$ 2>&1 &
wait_url http://localhost:8101/v1/models && echo "  val-sel ready"
$VLLM serve "$VAL_COND" --served-model-name validator_cond --port 8104 \
    --dtype bfloat16 --gpu-memory-utilization 0.12 --enforce-eager --max-model-len 6144 \
    > /tmp/eval_vc_$$ 2>&1 &
wait_url http://localhost:8104/v1/models && echo "  val-cond ready"
$VLLM serve "$FIXER" --served-model-name fixer --port 8102 \
    --dtype bfloat16 --gpu-memory-utilization 0.12 --enforce-eager --max-model-len 6144 \
    > /tmp/eval_f_$$ 2>&1 &
wait_url http://localhost:8102/v1/models && echo "  fixer ready"

OUT=eval_results/paper_${LABEL}_par_passAt8_bird_dev.jsonl
rm -f "$OUT"
$PY scripts/run_pipeline_rollouts.py \
    --input_file data/sft_bird_with_evidence_dev_text2sql.json \
    --output_file "$OUT" \
    --planner_host http://localhost:8100 --planner_format qwen \
    --validator_host none \
    --validator_sel_host http://localhost:8101 \
    --validator_cond_host http://localhost:8104 \
    --fixer_host http://localhost:8102 \
    --fixer_gate_exec_ok \
    --griffith_prompts \
    --K 8 --K_val 1 --K_fix 1 \
    --temperature 1.0 --top_p 0.9 \
    --max_planner_tokens 1024 --max_validator_tokens 384 --max_fixer_tokens 512 \
    --max_questions -1 --n_threads 4

$PY scripts/compute_bestofn_metrics.py "$OUT" paper_${LABEL}_par_passAt8

echo "[$(date)] $LABEL DONE"