search

AlignmentResearch / vlmrm

MIT License
36 stars 11 forks source link

image input for RL training? #3

Closed initial-h closed 2 months ago

initial-h commented 2 months ago

Thanks for this interesting work. Since the paper mentions the method is based on SAC with MLP network, is the input for RL algorithm is image or state? Thanks.

Rocamonde commented 2 months ago

Hi! Thanks for the question. The input to the policy network is the state. Using image arrays would normally require convents.

Will close for now, but if you have any other questions, feel free to create a new issue or, alternatively, email me or one of the other authors.

Best Juan