UniPC Multistep fix tensor dtype/device on order=3#7532
Merged
yiyixuxu merged 3 commits intohuggingface:mainfrom Apr 3, 2024
Merged
UniPC Multistep fix tensor dtype/device on order=3#7532yiyixuxu merged 3 commits intohuggingface:mainfrom
yiyixuxu merged 3 commits intohuggingface:mainfrom
Conversation
It wasn't catching errs on order==3. Might be excessive?
For completions sake. Probably overkill?
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
noskill
pushed a commit
to noskill/diffusers
that referenced
this pull request
Apr 5, 2024
* UniPC UTs iterate solvers on FP16 It wasn't catching errs on order==3. Might be excessive? * UniPC Multistep fix tensor dtype/device on order=3 * UniPC UTs Add v_pred to fp16 test iter For completions sake. Probably overkill?
sayakpaul
pushed a commit
that referenced
this pull request
Dec 23, 2024
* UniPC UTs iterate solvers on FP16 It wasn't catching errs on order==3. Might be excessive? * UniPC Multistep fix tensor dtype/device on order=3 * UniPC UTs Add v_pred to fp16 test iter For completions sake. Probably overkill?
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
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
As I found in #7517, UniPC with order=3 was erring on step 3 due to one of the tensors not being the correct dtype and device. This PR corrects that and updates the fp16 test to check every combination of order, solver, prediction to ensure they all work against half precision.
I verified working on both FP16 and BF16 on CUDA, and the updated test will correctly fail if you checkout to the first commit 036e33e before I fixed the
rhos_ptypings.The test might be a bit over-comprehensive as it stands. Right now I don't think its possible for all the parameters to create incompatible tensors but in my mind by just checking everything it should catch future refactors.
Should merge cleanly on top of #7531
@yiyixuxu