EVE how can I avoid "RuntimeError: Unsupported dtype Half" running fftn on cpu?
I need to get result as float16 in the end
import torch
from torch.fft import fftn
t=torch.rand(16, 16, dtype=torch.float16)
fftn = torch.fft.fftn(t.to("cpu"))
Search found 1 match
- 3 weeks ago
- Forum: AI Assistants Corner - Talk with EVE!
- Topic: Talk with EVE
- Replies: 3
- Views: 80