Install en_core_web_trf on Windows get conflicting dependencies error

[xxyzz]

How to reproduce the behaviour

Run this command py -m pip install https://github.com/explosion/spacy-models/releases/download/en_core_web_trf-3.4.1/en_core_web_trf-3.4.1-py3-none-any.whl will get the following error from pip:

ERROR: Cannot install en-core-web-trf because these package versions have conflicting dependencies.

The conflict is caused by:
    spacy-transformers 1.1.8 depends on torch>=1.6.0
    spacy-transformers 1.1.7 depends on torch>=1.6.0
    spacy-transformers 1.1.6 depends on torch>=1.6.0
    spacy-transformers 1.1.5 depends on torch>=1.6.0
    spacy-transformers 1.1.4 depends on torch>=1.6.0
ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts
    spacy-transformers 1.1.3 depends on torch>=1.6.0
    spacy-transformers 1.1.2 depends on torch>=1.6.0

To fix this you could try to:
1. loosen the range of package versions you've specified
2. remove package versions to allow pip attempt to solve the dependency conflict

The full log can be found at here: https://github.com/xxyzz/WordDumb/actions/runs/3573895840/jobs/6008499300

I also tried the python -m spacy download en_core_web_trf command but got the same error.

Your Environment

• Operating System: Microsoft Windows Server 2022
• Python Version Used: 3.11.0
• spaCy Version Used: 3.4.3
• Environment Information:

[adrianeboyd]

It looks like torch doesn't fully support python 3.11 yet. I would suggest using python 3.10 instead for now.

[xxyzz]

I just found torch only provides a single 3.11 wheel file for Linux and was about to close the issue. Thanks for your help!

[xxyzz]

Now I get another problem, the torch package for Windows on pypi.org is the CPU build version. This would get AssertionError: Torch not compiled with CUDA enabled error from torch when spaCy loading the trf model. The CUDA build torch package should be installed with this command python -m pip install torch --extra-index-url https://download.pytorch.org/whl/cu117.

[polm]

Glad you figured it out. Torch installs can be a little tricky, it's best to use their install widget to get it set up.