hangX/lerobot
1
0
Fork 0
forked from tangger/lerobot
Code Pull Requests Activity

Improve docs pi (#2110)

Browse Source 
* Improve docs and add numpy to pi install requirments

* fix formatting

* update command

* remvoe numpy dep
This commit is contained in:
Pepijn
2025-10-03 12:06:18 +02:00
committed by GitHub
parent 5c8dd883be
commit a4bed41132
2 changed files with 15 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
docs/source/pi0.mdx
View File

@@ -49,7 +49,7 @@ policy.type=pi0
For training π₀, you can use the standard LeRobot training script with the appropriate configuration: For training π₀, you can use the standard LeRobot training script with the appropriate configuration:
```bash ```bash
python src/lerobot/scripts/train.py \ python src/lerobot/scripts/lerobot_train.py \
--dataset.repo_id=your_dataset \ --dataset.repo_id=your_dataset \
--policy.type=pi0 \ --policy.type=pi0 \
--output_dir=./outputs/pi0_training \ --output_dir=./outputs/pi0_training \

19
docs/source/pi05.mdx
View File

@@ -51,13 +51,13 @@ policy.type=pi05
Here's a complete training command for finetuning the base π₀.₅ model on your own dataset: Here's a complete training command for finetuning the base π₀.₅ model on your own dataset:
```bash ```bash
python src/lerobot/scripts/train.py \ python src/lerobot/scripts/lerobot_train.py\
--dataset.repo_id=your_dataset \ --dataset.repo_id=your_dataset \
--policy.type=pi05 \ --policy.type=pi05 \
--output_dir=./outputs/pi0_training \ --output_dir=./outputs/pi05_training \
--job_name=pi0_training \ --job_name=pi05_training \
--policy.repo_id=lerobot/pi05_base \ --policy.repo_id=your_repo_id \
--policy.pretrained_path=your_repo_id \ --policy.pretrained_path=lerobot/pi05_base \
--policy.compile_model=true \ --policy.compile_model=true \
--policy.gradient_checkpointing=true \ --policy.gradient_checkpointing=true \
--wandb.enable=true \ --wandb.enable=true \
@@ -77,6 +77,15 @@ python src/lerobot/scripts/train.py \
- [lerobot/pi05_base](https://huggingface.co/lerobot/pi05_base) - [lerobot/pi05_base](https://huggingface.co/lerobot/pi05_base)
- [lerobot/pi05_libero](https://huggingface.co/lerobot/pi05_libero) (specifically trained on the Libero dataset) - [lerobot/pi05_libero](https://huggingface.co/lerobot/pi05_libero) (specifically trained on the Libero dataset)
If your dataset is not converted with `quantiles`, you can convert it with the following command:
```bash
python src/lerobot/datasets/v30/augment_dataset_quantile_stats.py \
--repo-id=your_dataset \
```
Or train pi05 with this normalization mapping: `--policy.normalization_mapping='{"ACTION": "MEAN_STD", "STATE": "MEAN_STD", "VISUAL": "IDENTITY"}'`
## Performance Results ## Performance Results
### Libero Benchmark Results ### Libero Benchmark Results