6 Commits

Author	SHA1	Message	Date
Pepijn	ac41cd6672	fix: bypass torchcodec video decoding by pre-reading frames via cv2 ... When torchcodec is installed, qwen-vl-utils ignores the fps parameter and defaults to 24fps if video metadata is missing, causing shape mismatches. Fix by reading video frames directly as PIL images and passing them to the processor, bypassing torchcodec entirely. Made-with: Cursor	2026-03-30 16:03:26 +02:00
Pepijn	9b211a45d6	fix: disable thinking mode in Qwen35VL single-episode fallback path ... The single-episode `segment_skills` method was missing `enable_thinking=False` in `apply_chat_template`, causing the model to output reasoning traces instead of JSON when the batch path fails and falls back to per-episode processing. Made-with: Cursor	2026-03-30 15:31:18 +02:00
root	a6387da464	add license	2026-03-11 23:14:22 +00:00
Jade Choghari	0328b3f4aa	Update src/lerobot/data_processing/data_annotations/vlm_annotations.py ... Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> Signed-off-by: Jade Choghari <chogharijade@gmail.com>	2026-03-11 16:10:37 -07:00
root	819c1b9710	add tests/fixes	2026-03-11 22:49:06 +00:00
root	f0848c6887	add subtasl	2026-03-11 19:51:48 +00:00