tangger/lerobot
1
0
Fork 1
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

update README

Browse Source
This commit is contained in:
Alexander Soare
2024-03-26 16:28:16 +00:00
parent 1ed0110900
commit be4441c7ff
3 changed files with 23 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
README.md
View File

@@ -135,11 +135,7 @@ hydra.run.dir=outputs/visualize_dataset/example
### Evaluate a pretrained policy ### Evaluate a pretrained policy
You can import our environment class, download pretrained policies from the HuggingFace hub, and use our rollout utilities with rendering: Check out [example 2](./examples/2_evaluate_pretrained_policy.py) to see how you can load a pretrained policy from HuggingFace hub, load up the corresponding environment and model, and run an evaluation.
```python
""" Copy pasted from `examples/2_evaluate_pretrained_policy.py`
# TODO
```
Or you can achieve the same result by executing our script from the command line: Or you can achieve the same result by executing our script from the command line:
```bash ```bash
@@ -150,7 +146,7 @@ eval_episodes=10 \
hydra.run.dir=outputs/eval/example_hub hydra.run.dir=outputs/eval/example_hub
``` ```
After launching training of your own policy, you can also re-evaluate the checkpoints with: After training your own policy, you can also re-evaluate the checkpoints with:
```bash ```bash
python lerobot/scripts/eval.py \ python lerobot/scripts/eval.py \
--config PATH/TO/FOLDER/config.yaml \ --config PATH/TO/FOLDER/config.yaml \
@@ -163,19 +159,9 @@ See `python lerobot/scripts/eval.py --help` for more instructions.
### Train your own policy ### Train your own policy
You can import our dataset, environment, policy classes, and use our training utilities (if some data is missing, it will be automatically downloaded from HuggingFace hub): You can import our dataset, environment, policy classes, and use our training utilities (if some data is missing, it will be automatically downloaded from HuggingFace hub): check out [example 3](./examples/3_train_policy.py). After you run this, you may want to revisit [example 2](./examples/2_evaluate_pretrained_policy.py) to evaluate your training output!
```python
""" Copy pasted from `examples/3_train_policy.py`
# TODO
```
Or you can achieve the same result by executing our script from the command line: In general, you can use our training script to easily train any policy on any environment:
```bash
python lerobot/scripts/train.py \
hydra.run.dir=outputs/train/example
```
You can easily train any policy on any environment:
```bash ```bash
python lerobot/scripts/train.py \ python lerobot/scripts/train.py \
env=aloha \ env=aloha \

2
examples/2_evaluate_pretrained_policy.py
View File

@@ -14,7 +14,7 @@ from lerobot.scripts.eval import eval
hub_id = "lerobot/diffusion_policy_pusht_image" hub_id = "lerobot/diffusion_policy_pusht_image"
folder = Path(snapshot_download(hub_id)) folder = Path(snapshot_download(hub_id))
# OR uncomment the following to evaluate a policy from the local outputs/train folder. # OR uncomment the following to evaluate a policy from the local outputs/train folder.
folder = Path("outputs/train/example_pusht_diffusion") # folder = Path("outputs/train/example_pusht_diffusion")
config_path = folder / "config.yaml" config_path = folder / "config.yaml"
weights_path = folder / "model.pt" weights_path = folder / "model.pt"

22
tests/test_examples.py
View File

@@ -46,11 +46,25 @@ def test_examples_3_and_2():
with open(path, "r") as file: with open(path, "r") as file:
file_contents = file.read() file_contents = file.read()
# Do less evals and use CPU. # Do less evals, use CPU, and use the local model.
file_contents = _find_and_replace( file_contents = _find_and_replace(
file_contents, file_contents,
['"eval_episodes=10"', '"rollout_batch_size=10"', '"device=cuda"'], [
['"eval_episodes=1"', '"rollout_batch_size=1"','"device=cpu"'], '"eval_episodes=10"',
'"rollout_batch_size=10"',
'"device=cuda"',
'# folder = Path("outputs/train/example_pusht_diffusion")',
'hub_id = "lerobot/diffusion_policy_pusht_image"',
"folder = Path(snapshot_download(hub_id)",
],
[
'"eval_episodes=1"',
'"rollout_batch_size=1"',
'"device=cpu"',
'folder = Path("outputs/train/example_pusht_diffusion")',
"",
"",
],
) )
assert Path(f"outputs/train/example_pusht_diffusion").exists() assert Path(f"outputs/train/example_pusht_diffusion").exists()