Feb 12, 2024 · We adapt pre-trained large vision models conditioned on specific downstream tasks in the context of multi-task policy learning.

We introduce task- conditioned adapters that do not require finetuning any pre- trained weights, combined with a single policy trained with behavior cloning and ...

In this work, we adapt pre-trained large vision models conditioned on specific downstream tasks in the context of multi-task policy learning.

We introduce a set of task-conditioned visual adapters that can be inserted inside a pre-trained visual Transformer [33] -based backbone. The task is ...

We introduce task- conditioned adapters that do not require finetuning any pre- trained weights, combined with a single policy trained with behavior cloning and ...

We introduce task-conditioned adapters that do not require finetuning any pretrained weights, combined with a single policy trained with behavior cloning.

We have generated expert demonstrations for 15 new MetaWorld tasks using TD-MPC2 single-task checkpoints. Few-shot expert demonstrations can be downloaded here.

This work introduces task-conditioned adapters that do not require finetuning any pretrained weights, combined with a single policy trained with behavior ...

Task-conditioned adapters to flexibly modulate features from a pre-trained vision model. •A single multi-task policy solving tasks with different embodiments ...

May 6, 2024 · The paper explores a method for adapting visual features in a neural network to perform multiple tasks simultaneously.