EN

← Multitask DiT

Toyota Research Institute

LeRobot Multitask DiT

LeRobot Multitask DiT

CLIP-conditioned diffusion transformer policy for learning multiple tasks.

Specifications

Architecture components: Diffusion action decoder
Backbone: CLIP vision-language embeddings with diffusion transformer
Frameworks: LeRobot / PyTorch
Input modalities: Language instruction
Input modalities: RGB camera images
Input modalities: Robot state / proprioception
Languages: English
Model type: Robot policy
Output action space: Continuous multitask robot action chunks
Parameter count: 0.45B parameters
Pipeline task: Robotics

Lineage

Unknown

Architecture & I/O

trajectory · continuous

Unknown

Continuous multitask robot action chunks

Source · Hugging Face LeRobot

No published default checkpoint.

Resources

documentation · LeRobot multi_task_dit documentation

Submit info