Really good work!!!
Here is some questions on how to train the diffusion policy based on the generated data, since I'm currently working on similar work and have come up with some problems. So I'm here for your suggestions.
I wonder how to set the state and action in data when training diffusion policy. Did you just keep them the same across every timestamp or did you use any other techniques in setting them? I'm just not quite sure whether setting the action the same as the state will cause any unexpected problems in training DP.
Really good work!!!
Here is some questions on how to train the diffusion policy based on the generated data, since I'm currently working on similar work and have come up with some problems. So I'm here for your suggestions.
I wonder how to set the state and action in data when training diffusion policy. Did you just keep them the same across every timestamp or did you use any other techniques in setting them? I'm just not quite sure whether setting the action the same as the state will cause any unexpected problems in training DP.