#!/bin/bash
#SBATCH --job-name=dppo_robomimic_full
#SBATCH --account=hk-project-p0022232
#SBATCH --partition=accelerated
#SBATCH --gres=gpu:1
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=40
#SBATCH --time=08:00:00
#SBATCH --mem=32G
#SBATCH --output=logs/dppo_robomimic_full_%j.out
#SBATCH --error=logs/dppo_robomimic_full_%j.err

module load devel/cuda/12.4

# MuJoCo environment for robomimic
export MUJOCO_PY_MUJOCO_PATH=$HOME/.mujoco/mujoco210
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.mujoco/mujoco210/bin:/usr/lib/nvidia
export MUJOCO_GL=egl

# Environment variables
export WANDB_MODE=online
export DPPO_WANDB_ENTITY=${DPPO_WANDB_ENTITY:-"dominik_roth"}
export DPPO_DATA_DIR=${DPPO_DATA_DIR:-$SLURM_SUBMIT_DIR/data}
export DPPO_LOG_DIR=${DPPO_LOG_DIR:-$SLURM_SUBMIT_DIR/log}

# Parse command line arguments
TASK=${1:-lift}  # lift, can, square, transport
MODE=${2:-finetune}  # pretrain or finetune

cd $SLURM_SUBMIT_DIR
source .venv/bin/activate

# Apply HoReKa MuJoCo compilation fix
echo "Applying HoReKa MuJoCo compilation fix..."
python -c "exec(open('fix_mujoco_compilation.py').read()); apply_mujoco_fix(); import mujoco_py; print('MuJoCo ready!')"

echo "Starting Robomimic $TASK $MODE experiment..."
echo "Job ID: $SLURM_JOB_ID"

# Select config based on mode
if [ "$MODE" = "pretrain" ]; then
    CONFIG_DIR="cfg/robomimic/pretrain/${TASK}"
    CONFIG_NAME="pre_diffusion_mlp"
elif [ "$MODE" = "finetune" ]; then
    CONFIG_DIR="cfg/robomimic/finetune/${TASK}"
    CONFIG_NAME="ft_ppo_diffusion_mlp"
else
    echo "Invalid mode: $MODE"
    exit 1
fi

# Run experiment
python script/run.py \
    --config-name=$CONFIG_NAME \
    --config-dir=$CONFIG_DIR \
    wandb=${WANDB_MODE:-null}

echo "Experiment completed!"