Neither CUDA nor MPS are available - defaulting to CPU. Note: This module is much faster with a GPU.
isspid opened this issue · 2 comments
While running ocrmypdf with EasyOCR as an engine I get the following output. When using EasyOCR not as part of ocrmypdf I have actually not had any problems with using the GPU.
Start processing 11 pages concurrently _sync.py:259
1 redoing OCR _pipeline.py:309
2 redoing OCR _pipeline.py:309
3 redoing OCR _pipeline.py:309
4 redoing OCR _pipeline.py:309
5 redoing OCR _pipeline.py:309
6 redoing OCR _pipeline.py:309
7 redoing OCR _pipeline.py:309
8 redoing OCR _pipeline.py:309
9 redoing OCR _pipeline.py:309
10 redoing OCR _pipeline.py:309
11 redoing OCR _pipeline.py:309
11 Neither CUDA nor MPS are available - defaulting to CPU. Note: This module is much faster with a GPU. easyocr.py:80
4 Neither CUDA nor MPS are available - defaulting to CPU. Note: This module is much faster with a GPU. easyocr.py:80
It seems that actually when I run EasyOCR in this environment, the GPU also doesn't get used. In another environment the GPU does get used. Could it be that the way pytorch was installed is the incorrect one?
Created a new conda environment with python 3.10. First installed pytorch with this command:
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117 which I got from the official pytorch website.
Then I installed ocrmypd and finally I ran: pip install git+https://github.com/ocrmypdf/OCRmyPDF-EasyOCR.git.
Now everything is working fine.