How to Install and run doctr pytorch build despite having tensorflow installed in my environment

[prithivi1]

I Need to install doctr package in my python environment, where both tensorflow and pytorch is installed. I would like to use the pytorch version. How can set that.

By default it is loading the tensorflow version since it is checked at first in models.utils.init.py

from doctr.file_utils import is_tf_available, is_torch_available

if is_tf_available():
    from .tensorflow import *
elif is_torch_available():
    from .pytorch import *  # type: ignore[assignment]

How can I skip this and import pytorch

[felixdittrich92]

Hi @prithivi1 👋,

You can set an env variable:
(NOTE: set this on top of your code before importing any doctr components)

import os
os.environ["USE_TORCH"] = "1"

or run your script with it:

USE_TORCH=1 python3 path/to/your_script.py

Same works with TensorFlow where USE_TF=0 would disable it and =1 enable :)

[prithivi1]

Thanks @felixdittrich92