huggingface / huggingface/diffusers
HunyuanVideoPieeline can't run on multi-GPU
- Vorherrschende Sprache
- Python
- Sterne
- 34.5k
- Forks
- 7.3k
- Ø Merge
- 3 T. 3 Std.
- Gemergte PRs (30 T.)
- 91
Beschreibung
### Describe the bug
With the code provided in the doc, HunyuanVideoPipeline raises "Expected all tensors to be on the same device" error on multi-GPU platform.
### Reproduction
With the code provided in the doc, HunyuanVideoPipeline raises "Expected all tensors to be on the same device" error on multi-GPU platform.
### Logs
```shell
```
### System Info
8*A100
### Who can help?
_No response_
Beitragsleitfaden
Rechercherichtung
Start with the documented HunyuanVideoPipeline example and reproduce the reported device mismatch on the stated 8×A100 multi-GPU setup. Trace where tensors are placed across devices; done means the documented pipeline runs without the “Expected all tensors to be on the same device” error.
Vom Indexierungsmodell aus dem Issue-Text verfasst.
Bewertung
- Tech-Stack
- python, pytorch
- Bereich
- distributed-systems, machine-learning
- Issue-Typ
- Bug
- Schwierigkeit
- 4/5
- Geschätzter Aufwand
- 3-5 Tage
- Aktivitätsstatus
- Veraltet
- Klarheit
- Muss geklärt werden
- Anfängerfreundlichkeit
- 30/100