mlspaces to lerobot script#52
Conversation
| quat_xyzw = np.concatenate([pose7[..., 4:7], pose7[..., 3:4]], axis=-1) | ||
| rotmat = Rotation.from_quat(quat_xyzw).as_matrix() |
There was a problem hiding this comment.
can just do rotmat = Rotation.from_quat(pose7[..., 3:], scalar_first=True), which is a bit clearer
|
|
||
| dataset = LeRobotDataset.create( | ||
| repo_id=args.repo_id, | ||
| fps=POLICY_FPS, |
There was a problem hiding this comment.
should probably read this from the data instead of hardcoding
|
|
||
| decord.bridge.set_bridge("native") | ||
|
|
||
| MLSPACES_GRIPPER_MAX_POS = 0.824033 |
There was a problem hiding this comment.
we should add documentation that says this is for the droid data? i.e. would require modification for the rby1 data
| @@ -0,0 +1,513 @@ | |||
| """Download and extract molmobot-data. | |||
There was a problem hiding this comment.
maybe rename this to be more descriptive, like bulk_download_mb_data.py or something? just so it's clear what data this is for.
|
|
||
| ```bash | ||
| DATA=./mbdata/FrankaPickOmniCamConfig/part0/train | ||
| python scripts/repair_video_paths.py "$DATA" |
There was a problem hiding this comment.
This shouldn't be necessary - is it?
There was a problem hiding this comment.
it is part of the instructions in hugging face
There was a problem hiding this comment.
Oh we should probably amend that then, the video paths have already been repaired before uploading.
2 participants
No description provided.