Leakpro-wide device and mps support
Aberta
enhancement
priority - 4
- Linguagem predominante
- Python
- Estrelas
- 23
- Forks
- 28
- Merge médio
- 4d 12h
- PRs com merge (30d)
- 5
Descrição
# Feature
## Desired Behavior / Functionality
Add support for mps, the cuda equivalent for mac
Instead of restating ' "cuda:0" if torch.cuda.is_available() else "cpu" ' lets have a import called device, `from ... import device` which in its definition handles coda/cpu/mps
## What Needs to Be Done
Create device import.
Replace all statically set devices with this
## How Can It Be Tested
A test should include creating a tensor, on the cpu, importing the device module and moving the tensor to the device then checking the device again.
Guia de contribuição
Avaliação
Esta issue ainda não foi avaliada.