#!/bin/bash
# Run planner-INDEP and planner-COLLAB ORPO at max_steps=600 (full schedule).
# GPU 0 sequentially: INDEP then COLLAB.
# Goal: check if the collab-vs-indep gap surfaces with more training.
set -e
cd /home/datht/mats-sql-tist/alignment-handbook

LOG=/tmp/collab_orpo_long.log

echo "[1/2] ORPO planner INDEPENDENT (max_steps=600)..." | tee $LOG
CUDA_VISIBLE_DEVICES=0 PYTHONPATH=src/ ACCELERATE_LOG_LEVEL=info \
    /home/datht/anaconda3/envs/mats/bin/accelerate launch \
    --main_process_port 29570 \
    --config_file recipes/accelerate_configs/single_gpu0_local.yaml \
    scripts/run_orpo.py recipes/qwen-coder-0.5b-bird/orpo-planner-independent.yaml 2>&1 | tee -a $LOG

echo "[2/2] ORPO planner COLLABORATIVE (max_steps=600)..." | tee -a $LOG
CUDA_VISIBLE_DEVICES=0 PYTHONPATH=src/ ACCELERATE_LOG_LEVEL=info \
    /home/datht/anaconda3/envs/mats/bin/accelerate launch \
    --main_process_port 29571 \
    --config_file recipes/accelerate_configs/single_gpu0_local.yaml \
    scripts/run_orpo.py recipes/qwen-coder-0.5b-bird/orpo-planner-collaborative.yaml 2>&1 | tee -a $LOG

echo "DONE_COLLAB_ORPO_LONG" | tee -a $LOG