scripts/render.py

#!/usr/bin/env python3
"""Render evaluation videos from a trained checkpoint.

Usage:
    python render.py --checkpoint checkpoints/halfcheetah_6x1/ours_s0.pt \
                     --env mamujoco_HalfCheetah_6x1 \
                     --output_dir videos/ \
                     --n_episodes 3
"""

import argparse
import sys
from pathlib import Path

import numpy as np
import torch


MAMUJOCO_ENVS = {
    "mamujoco_HalfCheetah_6x1": ("HalfCheetah-v2", "6x1"),
    "mamujoco_Humanoid_9_8": ("Humanoid-v2", "9|8"),
    "mamujoco_ManySegmentSwimmer_6x1": ("ManySegmentSwimmer", "6x1"),
}


def parse_args():
    parser = argparse.ArgumentParser(description="Render evaluation videos")
    parser.add_argument("--checkpoint", type=str, required=True,
                        help="Path to .pt checkpoint file")
    parser.add_argument("--env", type=str, required=True,
                        choices=list(MAMUJOCO_ENVS.keys()) + ["harvest_5"],
                        help="Environment name")
    parser.add_argument("--output_dir", type=str, default="videos",
                        help="Directory to save rendered videos")
    parser.add_argument("--n_episodes", type=int, default=3,
                        help="Number of episodes to render")
    parser.add_argument("--seed", type=int, default=0,
                        help="Rendering seed")
    parser.add_argument("--fps", type=int, default=30,
                        help="Frames per second for video")
    parser.add_argument("--width", type=int, default=640,
                        help="Video width in pixels")
    parser.add_argument("--height", type=int, default=480,
                        help="Video height in pixels")
    return parser.parse_args()


def make_render_env(env_name, seed=0):
    """Create environment with rendering enabled."""
    if env_name in MAMUJOCO_ENVS:
        scenario, agent_conf = MAMUJOCO_ENVS[env_name]
        try:
            from mappo_lagrangian.envs.safety_ma_mujoco.safety_multiagent_mujoco import MujocoMulti
        except ImportError:
            print("ERROR: mappo_lagrangian package not installed.")
            print("Install from: macpo_base/MAPPO-Lagrangian/")
            sys.exit(1)

        env_args = {
            "scenario": scenario,
            "agent_conf": agent_conf,
            "agent_obsk": 1,
            "episode_limit": 1000,
        }
        env = MujocoMulti(env_args=env_args)
        env.seed(seed)
        return env
    else:
        print(f"Rendering not supported for {env_name}")
        sys.exit(1)


def render_episodes(env, checkpoint, n_episodes, output_dir, fps=30,
                    width=640, height=480):
    """Render episodes and save as videos."""
    try:
        import imageio
    except ImportError:
        print("ERROR: imageio required for video rendering.")
        print("  pip install imageio imageio-ffmpeg")
        sys.exit(1)

    output_dir = Path(output_dir)
    output_dir.mkdir(parents=True, exist_ok=True)

    for ep in range(n_episodes):
        frames = []
        obs = env.reset()
        done = False
        step = 0
        episode_reward = 0.0
        episode_cost = 0.0

        while not done and step < 1000:
            # Render frame
            try:
                frame = env.render(mode="rgb_array")
                if frame is not None:
                    frames.append(frame)
            except Exception as e:
                if step == 0:
                    print(f"  Warning: render failed ({e}). Trying offscreen...")
                    try:
                        frame = env.render(
                            mode="rgb_array",
                            width=width,
                            height=height,
                        )
                        if frame is not None:
                            frames.append(frame)
                    except Exception:
                        print("  Offscreen rendering also failed. Skipping video.")
                        return

            # Template: replace with actual policy forward pass
            n_agents = env.n_agents if hasattr(env, "n_agents") else 1
            action = [env.action_space[i].sample() for i in range(n_agents)]

            obs, rewards, dones, infos = env.step(action)
            episode_reward += sum(rewards) if isinstance(rewards, list) else rewards
            if isinstance(infos, list):
                episode_cost += sum(info.get("cost", 0) for info in infos)
            done = all(dones) if isinstance(dones, list) else dones
            step += 1

        if frames:
            video_path = output_dir / f"episode_{ep:03d}.mp4"
            imageio.mimwrite(str(video_path), frames, fps=fps)
            print(f"  Episode {ep}: {step} steps, reward={episode_reward:.1f}, "
                  f"cost={episode_cost:.2f} -> {video_path}")
        else:
            print(f"  Episode {ep}: no frames captured")


def main():
    args = parse_args()

    print(f"Loading checkpoint: {args.checkpoint}")
    ckpt = torch.load(args.checkpoint, map_location="cpu", weights_only=False)

    if "final_metrics" in ckpt:
        m = ckpt["final_metrics"]
        print(f"  Welfare: {m.get('total_welfare', 'N/A')}")
        print(f"  Final rho: {m.get('final_rho', 'N/A')}")

    try:
        env = make_render_env(args.env, seed=args.seed)
    except (ImportError, SystemExit):
        print("Cannot create rendering environment. Exiting.")
        sys.exit(1)

    print(f"\nRendering {args.n_episodes} episodes to {args.output_dir}/")
    render_episodes(
        env, ckpt, args.n_episodes, args.output_dir,
        fps=args.fps, width=args.width, height=args.height,
    )
    print("Done.")


if __name__ == "__main__":
    main()