rl/scripts/train.py

#!/usr/bin/env python3
"""Train V2D-G1-TableObject-v0 with the abstract PPO trainer.

Must be launched through Isaac Lab (AppLauncher) so Sim starts first:

  cd simulation && source .venv/bin/activate
  ../rl/train.sh --headless --num_envs 64
"""

from __future__ import annotations

import argparse
import os
import sys
from datetime import datetime
from pathlib import Path

from isaaclab.app import AppLauncher

parser = argparse.ArgumentParser(description="Train G1 table-object RL (algorithm stub).")
parser.add_argument("--task", type=str, default="V2D-G1-TableObject-v0")
parser.add_argument("--num_envs", type=int, default=None)
parser.add_argument("--max_iterations", type=int, default=None)
parser.add_argument("--seed", type=int, default=42)
AppLauncher.add_app_launcher_args(parser)
args_cli = parser.parse_args()

app_launcher = AppLauncher(args_cli)
simulation_app = app_launcher.app

import gymnasium as gym  # noqa: E402
from isaaclab_rl.rsl_rl import RslRlVecEnvWrapper  # noqa: E402
from isaaclab_tasks.utils import parse_env_cfg  # noqa: E402

import v2d_sim  # noqa: E402, F401
from v2d_rl.algorithms.ppo import RslRlPpoTrainer  # noqa: E402


def main() -> None:
    env_cfg = parse_env_cfg(args_cli.task, device=args_cli.device, num_envs=args_cli.num_envs)
    env = gym.make(args_cli.task, cfg=env_cfg)
    spec = gym.spec(args_cli.task)
    agent_cfg_entry = spec.kwargs["rsl_rl_cfg_entry_point"]
    mod_name, cls_name = agent_cfg_entry.rsplit(":", 1)
    import importlib

    agent_cfg = getattr(importlib.import_module(mod_name), cls_name)()
    if args_cli.max_iterations is not None:
        agent_cfg.max_iterations = args_cli.max_iterations
    agent_cfg.seed = args_cli.seed

    env = RslRlVecEnvWrapper(env)
    log_root = Path(os.environ.get("RUNS_DIR", "runs")) / "rl" / args_cli.task
    log_dir = str(log_root / datetime.now().strftime("%Y-%m-%d_%H-%M-%S"))
    Path(log_dir).mkdir(parents=True, exist_ok=True)
    print(f"[v2d-rl] task={args_cli.task}  log={log_dir}", flush=True)
    trainer = RslRlPpoTrainer(agent_cfg)
    trainer.train(env, log_dir=log_dir)
    env.close()


if __name__ == "__main__":
    try:
        main()
    finally:
        simulation_app.close()
        sys.exit(0)