feat(Train): enable buckets with streaming dataset (#4312)

This commit is contained in:
Steven Palma
2026-08-03 16:00:48 +02:00
committed by GitHub
parent f1efa588b8
commit e867359d09
4 changed files with 57 additions and 1 deletions
+16
View File
@@ -142,6 +142,22 @@ repo_id = "yaak-ai/L2D-v3"
dataset = StreamingLeRobotDataset(repo_id) # streams directly from the Hub
```
Datasets stored in an [HF Storage Bucket](https://huggingface.co/docs/hub/storage-buckets) (`hf://buckets/`) can be streamed the same way by passing `repo_type="bucket"`:
```python
dataset = StreamingLeRobotDataset("my-org/my-bucket", repo_type="bucket")
```
Both options are available in `lerobot-train` through `--dataset.streaming=true`, and `--dataset.repo_type=bucket` to stream from a bucket instead of a Hub dataset repo:
```bash
lerobot-train \
--dataset.repo_id=my-org/my-bucket \
--dataset.repo_type=bucket \
--dataset.streaming=true \
...
```
<div style="display:flex; justify-content:center; gap:12px; flex-wrap:wrap;">
<figure style="margin:0; text-align:center;">
<img