feat(data): add recipe-driven language supervision (#4182)

* feat(data): add recipe-driven language supervision

* test(collate): expect preserved language columns

* Address PR review feedback

* Address Claude review feedback
This commit is contained in:
Pepijn
2026-08-04 16:48:47 +02:00
committed by GitHub
parent f66e5128ec
commit 64b23178d5
28 changed files with 1062 additions and 79 deletions
+1 -3
View File
@@ -25,7 +25,7 @@ from datasets import Dataset # noqa: E402
from lerobot.datasets.io_utils import (
hf_transform_to_torch,
)
from lerobot.datasets.sampler import EpisodeAwareSampler
from lerobot.datasets.sampler import EpisodeAwareSampler, compute_sampler_state
def calculate_episode_data_index(hf_dataset: Dataset) -> dict[str, torch.Tensor]:
@@ -154,8 +154,6 @@ def test_partial_episode_drop_warns(caplog):
# --- seeded (seed, epoch) shuffling, resume, and state ---
from lerobot.datasets.sampler import compute_sampler_state # noqa: E402
EPISODE_BOUNDS = ([0, 2, 3], [2, 3, 6]) # episodes of 2, 1 and 3 frames