huggingface / huggingface/diffusers
Potential incorrect reshaping in 2D positional embedding
- Vorherrschende Sprache
- Python
- Sterne
- 34.5k
- Forks
- 7.3k
- Ø Merge
- 3 T. 3 Std.
- Gemergte PRs (30 T.)
- 91
Beschreibung
### Describe the bug
Hi there,
I have concerns with this line of code (https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/embeddings.py#L282).
Specifically, `grid_size` is the tuple consisting of the height `H` and width `W` of the image. `grid` computed in L280 should have the shape `2*H*W`, and L282 reshapes it into `2*1*W*H`. The dimensions `W*H` will be later flattened to match the dimensions of the latent.
However, if you continue to `PatchEmbed` (https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/embeddings.py#L549), you will notice that the latent with shape `BCHW` is flattened into `B(H*W)C`, this flattening operation does not seem to match with `grid` in L282. I think this reordering will mess up the ordering of dimensions when being flattened in case `H` and `W` are not equal.
### Reproduction
This potential bug is conceptual and no need for reproduction.
### Logs
```shell
```
### System Info
Current `diffusers` implementation.
### Who can help?
@yiyixuxu @sayakpaul
Beitragsleitfaden
Rechercherichtung
Start in src/diffusers/models/embeddings.py at the grid construction around line 280 and the PatchEmbed implementation around line 549. Compare the positional grid ordering with the BCHW-to-B(H*W)C flattening for unequal height and width, then confirm whether the ordering matches and add a regression test or correction if it does not.
Vom Indexierungsmodell aus dem Issue-Text verfasst.
Bewertung
- Tech-Stack
- python, pytorch
- Bereich
- machine-learning
- Issue-Typ
- Bug
- Schwierigkeit
- 4/5
- Geschätzter Aufwand
- 3-5 Tage
- Aktivitätsstatus
- Veraltet
- Klarheit
- Größtenteils klar
- Anfängerfreundlichkeit
- 35/100