Open
Conversation
There were small mismatch of `next_latents_mean`, `next_latents` and `log_prob` between train and inference, the causes are: 1. Different precision of `next_latents` were used for `log_prob` computation: During inference, it was actually used as `float32` while during training, `next_latents` were passed directly into the `scheduler.step` as `bfloat16`. Casting `next_latents` to the actual `input_dtype` will fix the issue. 2. Small precision difference between `timesteps` and `sigmas`. Passing `next_t` for `forward` function during sampling fixes it.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
There were small mismatch of
next_latents_mean,next_latentsandlog_probbetween train and inference, the causes are:Different precision of
next_latentswere used forlog_probcomputation: During inference, it was actually used asfloat32while during training,next_latentswere passed directly into thescheduler.stepasbfloat16. Castingnext_latentsto the actualinput_dtypewill fix the issue.Small precision difference between
timestepsandsigmas. Passingnext_tforforwardfunction during sampling fixes it.