code/scripts/collab_phase_f_bcd_gpu0only.sh

#!/bin/bash
# Run best-of-8 eval for configs (b), (b'), (d) only — config (a) rollouts already exist.
set -e
cd /home/datht/mats-sql-tist

EVAL_INPUT=data/sft_bird_with_evidence_dev_text2sql.json
[ ! -f "$EVAL_INPUT" ] && EVAL_INPUT=data/sft_bird_dev_with_evidence_text2sql_new.json

V_SFT=alignment-handbook/output/qwen-coder0.5b-bird-validator-sft
F_SFT=alignment-handbook/output/qwen-coder0.5b-bird-fixer-sft
F_ORPO=alignment-handbook/output/qwen-coder0.5b-bird-fixer-orpo
P_INDEP=alignment-handbook/output/qwen-coder0.5b-bird-planner-INDEP-orpo
P_COLLAB=alignment-handbook/output/qwen-coder0.5b-bird-planner-COLLAB-orpo

VLLM=/home/datht/anaconda3/envs/llm/bin/vllm

run_eval_bestof8() {
    local LABEL=$1 P_CKPT=$2 V_CKPT=$3 F_CKPT=$4
    echo ""
    echo "==== $LABEL ===="
    pkill -f "vllm serve" 2>/dev/null || true
    sleep 8

    CUDA_VISIBLE_DEVICES=0 $VLLM serve "$P_CKPT" \
        --served-model-name planner --port 8100 --dtype bfloat16 \
        --gpu-memory-utilization 0.28 --enforce-eager --max-model-len 4096 \
        > /tmp/collab_eval_p.log 2>&1 &
    sleep 3
    CUDA_VISIBLE_DEVICES=0 $VLLM serve "$V_CKPT" \
        --served-model-name validator --port 8101 --dtype bfloat16 \
        --gpu-memory-utilization 0.28 --enforce-eager --max-model-len 4096 \
        > /tmp/collab_eval_v.log 2>&1 &
    sleep 3
    CUDA_VISIBLE_DEVICES=0 $VLLM serve "$F_CKPT" \
        --served-model-name fixer --port 8102 --dtype bfloat16 \
        --gpu-memory-utilization 0.28 --enforce-eager --max-model-len 4096 \
        > /tmp/collab_eval_f.log 2>&1 &

    for url in http://localhost:8100/v1/models http://localhost:8101/v1/models http://localhost:8102/v1/models; do
        for i in {1..120}; do
            curl --noproxy localhost -fs $url >/dev/null 2>&1 && break
            sleep 5
        done
    done

    OUT=eval_results/collab_BoN8_${LABEL}_bird_dev.jsonl
    rm -f "$OUT"
    PYTHONPATH=. /home/datht/anaconda3/envs/mats/bin/python scripts/run_pipeline_rollouts.py \
        --input_file "$EVAL_INPUT" \
        --output_file "$OUT" \
        --planner_host http://localhost:8100 \
        --validator_host http://localhost:8101 \
        --fixer_host http://localhost:8102 \
        --K 8 --K_val 1 --K_fix 1 --temperature 0.7 --top_p 0.9 \
        --max_questions -1 --n_threads 6

    /home/datht/anaconda3/envs/mats/bin/python scripts/compute_bestofn_metrics.py "$OUT" "$LABEL" || echo "metric failed for $LABEL"
}

run_eval_bestof8 b_planner_indep "$P_INDEP"   "$V_SFT" "$F_SFT"
run_eval_bestof8 b_prime_planner_indep_with_fixer_orpo "$P_INDEP" "$V_SFT" "$F_ORPO"
run_eval_bestof8 d_planner_collab_with_fixer_orpo      "$P_COLLAB" "$V_SFT" "$F_ORPO"

pkill -f "vllm serve" 2>/dev/null || true
echo "DONE_BCD_BESTOFN_EVAL"