Diffusion Steps and Truncation Settings of TransfuserDP

[cheng1559]

Hello author, I am currently trying to reproduce the results of TransfuserDP as presented in your paper, and I have encountered some issues during the process.

In your code, the trajectory is truncated to 50 out of 1000 steps during training when adding noise. If I would like to apply a vanilla Diffusion Policy instead, how should I appropriately set the total number of diffusion steps? Also, is it available to include a truncation step?

I would greatly appreciate your guidance on this. Thank you!