lerobot/docs/source/evaluation.mdx

# Evaluation

`lerobot-eval` runs a trained policy on a simulation benchmark and reports success rate, reward, and (optionally) episode videos. It handles environment creation, batched rollouts, and metric aggregation automatically.

## Quick start

Evaluate a Hub-hosted policy on LIBERO:

```bash
lerobot-eval \
    --policy.path=pepijn223/smolvla_libero \
    --env.type=libero \
    --env.task=libero_spatial \
    --eval.n_episodes=10 \
    --policy.device=cuda
```

Evaluate a local checkpoint:

```bash
lerobot-eval \
    --policy.path=outputs/train/act_pusht/checkpoints/005000/pretrained_model \
    --env.type=pusht \
    --eval.n_episodes=10
```

`batch_size` defaults to **auto** (based on CPU cores). The script picks the right number of parallel environments for your machine.

## Key flags

| Flag | Default | Description |
|---|---|---|
| `--policy.path` | required | Hub repo ID or local path to a pretrained model |
| `--env.type` | required | Benchmark name (`pusht`, `libero`, `metaworld`, etc.) |
| `--env.task` | varies | Task or suite name (e.g. `libero_spatial`, `libero_10`) |
| `--eval.n_episodes` | `50` | Total episodes to run (across all tasks) |
| `--eval.batch_size` | `0` (auto) | Number of parallel environments. `0` = auto-tune from CPU cores |
| `--eval.use_async_envs` | `true` | Use `AsyncVectorEnv` (parallel stepping). Auto-downgrades to sync when `batch_size=1` |
| `--policy.device` | `cuda` | Inference device |
| `--policy.use_amp` | `false` | Mixed-precision inference (saves VRAM, faster on Ampere+) |
| `--seed` | `1000` | Random seed for reproducibility |
| `--output_dir` | auto-generated | Where to write results and videos |

### Environment-specific flags

Some benchmarks accept additional flags through `--env.*`:

```bash
# LIBERO: map simulator camera names to policy feature names
--env.camera_name_mapping='{"agentview_image": "camera1", "robot0_eye_in_hand_image": "camera2"}'

# Fill unused camera slots with zeros
--policy.empty_cameras=1
```

See each benchmark's documentation ([LIBERO](libero), [Meta-World](metaworld)) for benchmark-specific flags.

## How batch_size works

`batch_size` controls how many environments run in parallel within a single `VectorEnv`:

| `batch_size` | Behavior |
|---|---|
| `0` (default) | Auto-tune: `floor(cpu_cores × 0.7)`, capped by `n_episodes` and `64` |
| `1` | Single environment, synchronous. Useful for debugging |
| `N` | N environments step in parallel via `AsyncVectorEnv` |

When `batch_size > 1` and `use_async_envs=true`, each environment runs in its own subprocess via Gymnasium's `AsyncVectorEnv`. This parallelizes the simulation stepping (the main bottleneck), while the policy runs a single batched forward pass on GPU.

**Example:** On a 16-core machine with `n_episodes=100`:
- Auto batch_size = `floor(16 × 0.7)` = `11`
- 11 environments step simultaneously → ~11× faster than sequential

## Performance

### AsyncVectorEnv (default)

`AsyncVectorEnv` spawns one subprocess per environment. Each subprocess has its own simulator instance. While the policy computes actions on GPU, all environments step in parallel on CPU:

```
GPU:  [inference]....[inference]....[inference]....
CPU:  [step × N]....................[step × N]......
      ↑ parallel                   ↑ parallel
```

For GPU-based simulators (LIBERO, Meta-World), the environments use **lazy initialization**: the GPU/EGL context is created inside the worker subprocess on first `reset()`, not in the parent process. This avoids `EGL_BAD_CONTEXT` crashes from inheriting stale GPU handles across `fork()`.

### Lazy task loading

For multi-task benchmarks (e.g. LIBERO with 10 tasks), environments are wrapped in `_LazyAsyncVectorEnv` which defers worker creation until the task is actually evaluated. This keeps peak process count = `batch_size` instead of `n_tasks × batch_size`. After each task completes, workers are closed to free resources.

### Tuning for speed

| Situation | Recommendation |
|---|---|
| Slow eval, low GPU utilization | Increase `batch_size` (or leave at auto) |
| Out of memory (system RAM) | Decrease `batch_size` |
| Out of GPU memory | Decrease `batch_size`, or use `--policy.use_amp=true` |
| Debugging / single-stepping | `--eval.batch_size=1 --eval.use_async_envs=false` |

## Output

Results are written to `output_dir` (default: `outputs/eval/<date>/<time>_<job_name>/`):

- `eval_info.json` — full metrics: per-episode, per-task, per-group, and overall aggregates
- `videos/` — episode recordings (when `--eval.n_episodes_to_render > 0`)

### Metrics

| Metric | Description |
|---|---|
| `pc_success` | Success rate (%). Based on `info["is_success"]` from the environment |
| `avg_sum_reward` | Mean cumulative reward per episode |
| `avg_max_reward` | Mean peak reward per episode |
| `n_episodes` | Total episodes evaluated |
| `eval_s` | Total wall-clock time |
| `eval_ep_s` | Mean wall-clock time per episode |

## Multi-task evaluation

For benchmarks with multiple tasks (LIBERO suites, Meta-World MT50), `lerobot-eval` automatically:

1. Creates environments for all tasks in the selected suite(s)
2. Evaluates each task sequentially (one task's workers at a time)
3. Aggregates metrics per-task, per-group (suite), and overall

```bash
# Evaluate all 10 tasks in libero_spatial
lerobot-eval \
    --policy.path=pepijn223/smolvla_libero \
    --env.type=libero \
    --env.task=libero_spatial \
    --eval.n_episodes=10

# Evaluate multiple suites
lerobot-eval \
    --policy.path=pepijn223/smolvla_libero \
    --env.type=libero \
    --env.task="libero_spatial,libero_object" \
    --eval.n_episodes=10
```

## API usage

You can call the eval functions directly from Python:

```python
from lerobot.envs.factory import make_env
from lerobot.policies.factory import make_policy
from lerobot.scripts.lerobot_eval import eval_policy

envs = make_env(env_cfg, n_envs=10)
policy = make_policy(cfg=policy_cfg, env_cfg=env_cfg)

metrics = eval_policy(
    env=envs["libero_spatial"][0],
    policy=policy,
    n_episodes=10,
)
print(metrics["pc_success"])
```