search

pharmapsychotic / clip-interrogator

Image to prompt with BLIP and CLIP
MIT License
2.58k stars 429 forks source link

bug 0.5.4 #87

Open ucasiggcas opened 11 months ago

ucasiggcas commented 11 months ago

Traceback (most recent call last): File "/bin/image2text.py", line 5, in from clip_interrogator import Config, Interrogator, list_caption_models, list_clip_models File "/lib/python3.9/site-packages/clip_interrogator/init.py", line 1, in from .clip_interrogator import Interrogator, Config File "python3.9/site-packages/clip_interrogator/clip_interrogator.py", line 30, in class Config: File "/lib/python3.9/site-packages/clip_interrogator/clip_interrogator.py", line 53, in Config device: str = ("mps" if torch.backends.mps.is_available() else "cuda" if torch.cuda.is_available() else "cpu") AttributeError: module 'torch.backends' has no attribute 'mps'

import torch torch.version '1.11.0'

pharmapsychotic commented 10 months ago

ah sorry looks like mps backend was added in pytorch 1.13, I'll update the requirement to reflect the minimum version