Hello,
Maybe I missed it, but I'm looking for the code that was used to train the single task diffusion policy (in section 5.4). The paper mentioned using a Diffusion Policy trained with the original diffusion policy approach, and lists some of the hyper-parameters used. I'm wondering if the exact code for the DSRL fine tuning of this diffusion policy are supplied? I've been trying to replicate DSRL on a custom task (specifically one that uses image/hybrid conditioning), and I haven't been able to get it to work with two different base diffusion policies (one trained via DPPO's repo, then using this repo modified to support image observations, and a second starting from the RealStanford DiffusionPolicy repository, and then adding in the DSRL functionality on top). For the sake of ensuring it's not an implementation error, I'm wondering if you have published the exact code used for this experiment.
Thank you,
Hello,
Maybe I missed it, but I'm looking for the code that was used to train the single task diffusion policy (in section 5.4). The paper mentioned using a Diffusion Policy trained with the original diffusion policy approach, and lists some of the hyper-parameters used. I'm wondering if the exact code for the DSRL fine tuning of this diffusion policy are supplied? I've been trying to replicate DSRL on a custom task (specifically one that uses image/hybrid conditioning), and I haven't been able to get it to work with two different base diffusion policies (one trained via DPPO's repo, then using this repo modified to support image observations, and a second starting from the RealStanford DiffusionPolicy repository, and then adding in the DSRL functionality on top). For the sake of ensuring it's not an implementation error, I'm wondering if you have published the exact code used for this experiment.
Thank you,