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

Add training on custom openpi datasets

Browse Source 
Cleanup instructions

clean up doc

pass linter

updates

Add test
This commit is contained in:
Michael Equi
2024-12-22 19:19:54 +00:00
parent 385780ecc3
commit 9da84a2f7f
9 changed files with 75 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
examples/aloha_real/README.md
View File

@@ -54,20 +54,4 @@ While we strongly recommend fine-tuning the model to your own data to adapt it t
## Training on your own Aloha dataset
OpenPI suppports training on data collected in the default aloha hdf5 format. To do so you must first convert the data to the huggingface format. We include `scripts/aloha_hd5.py` to help you do this. Once the dataset is converted, add a new `TrainConfig` to `src/openpi/training/configs.py` and replace repo id with the id assigned to your dataset during conversion.
```python
TrainConfig(
name=<your-config-name>,
data=LeRobotAlohaDataConfig(
repo_id=<your-repo-id>,
delta_action_mask=[True] * 6 + [False] + [True] * 6 + [False],
),
),
```
Run the training script:
```bash
uv run scripts/train.py <your-config-name>
```
OpenPI suppports training on data collected in the default aloha hdf5 format using the `scripts/aloha_hd5.py` conversion script. Once the dataset is converted, add a new `TrainConfig` to `src/openpi/training/configs.py` (see the `aloha_static_cups_open` example config) and replace repo id with the id assigned to your dataset during conversion. Before training on a new dataset, you must first compute the norm stats using `scripts/compute_norm_stats.py`.