multi-GPU VAE Fix for Cosmos 3#13924
Merged
Merged
Conversation
Under sharded placement (device_map="balanced"), vae.encode() runs on the VAE's own device while the mean/inv_std buffers were pinned to x.device, causing a cross-device RuntimeError. Compute raw_mu first, then pin the normalization buffers to its device so all tensors share one device.
|
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. |
Member
|
Thanks for your PR! This needs a bit of library-wise thinking. I'd suggest shipping the VAE fixes in this PR. We can maintain a script in our docs for the time being to show how this can be used. |
33ea463 to
9f099d4
Compare
Contributor
|
Hi @atharvajoshi10, thanks for the PR! It does not appear to link an issue it fixes. If this PR addresses an existing issue, please add a closing keyword (e.g. |
sayakpaul
approved these changes
Jun 23, 2026
DN6
pushed a commit
that referenced
this pull request
Jul 1, 2026
fix(cosmos3): pin VAE latent norm buffers to encode output device Under sharded placement (device_map="balanced"), vae.encode() runs on the VAE's own device while the mean/inv_std buffers were pinned to x.device, causing a cross-device RuntimeError. Compute raw_mu first, then pin the normalization buffers to its device so all tensors share one device. Co-authored-by: Atharva Joshi <[email protected]> Co-authored-by: Sayak Paul <[email protected]>
JingyaHuang
pushed a commit
to JingyaHuang/diffusers
that referenced
this pull request
Jul 2, 2026
fix(cosmos3): pin VAE latent norm buffers to encode output device Under sharded placement (device_map="balanced"), vae.encode() runs on the VAE's own device while the mean/inv_std buffers were pinned to x.device, causing a cross-device RuntimeError. Compute raw_mu first, then pin the normalization buffers to its device so all tensors share one device. Co-authored-by: Atharva Joshi <[email protected]> Co-authored-by: Sayak Paul <[email protected]>
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
4 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.
multi-GPU VAE fix
What this does
Fixes a cross-device error that surfaced when running Cosmos 3 under sharded (
device_map) placement.Fix VAE latent-norm device mismatch under sharded placement
Under
device_map="balanced",vae.encode()runs on the VAE's own device while themean/inv_stdnormalization buffers were pinned tox.device, causing a cross-deviceRuntimeError. Now computesraw_mufirst and pins the normalization buffers to its device so all tensors share one device.