Skip to content

DPPO with Summed Likelihood (QSM-PG)#13

Open
jlidard wants to merge 7 commits intomainfrom
dppo_sumlikelihood
Open

DPPO with Summed Likelihood (QSM-PG)#13
jlidard wants to merge 7 commits intomainfrom
dppo_sumlikelihood

Conversation

@jlidard
Copy link
Owner

@jlidard jlidard commented Nov 23, 2024

Take expectation over env steps and sum over all denoising steps.


This change is Reviewable

obs_repeat = {
"state": obs["state"].repeat_interleave(self.ft_denoising_steps, dim=0)
}
denoising_inds = denoising_inds.repeat_interleave(chains_prev.shape[0])
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is wrong? obs_repeat now has obs_1, obs_1, obs_1, ... obs_2, obs_2,...., but denoising_inds has 0, 0,0,0,0,,,,...1,1,1,1,1, .....

I think you want repeat instead of repeat_interleave for denoising_inds


# exponentially interpolate between the base and the current clipping value over denoising steps and repeat
t = (denoising_inds.float() / (self.ft_denoising_steps - 1)).to(self.device)
t = t[
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this needs to be modified too then I think

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants