Installation are different depending what are we trying to do:
For baskerville we can't use the usual python -m pip install . as torch and tochvision need to be loaded independently.
Locally and in tests we need a full installation equivalent to python -m pip install .
Napari plugin conflicts with full installation because requirements for pydantic are different.
Suggestion:
python -m pip install . should have a minimal installation needed for everything, but it doesn't have all dependencies.
python -m pip install ."[baskerville]" will have the needed libraries for running in baskervile (which need to be complemented with torch and tochvision modules).
python -m pip install ."[all]" should have everything to train the model locally.
python -m pip install ."[napari]" with have the base version of dependencies from 1. plus the version of pydantic that is compatible with napari.
python -m pip install ."[tests]" should have everything to train the model locally and run tests.
Installation are different depending what are we trying to do:
python -m pip install .as torch and tochvision need to be loaded independently.python -m pip install .Suggestion:
python -m pip install .should have a minimal installation needed for everything, but it doesn't have all dependencies.python -m pip install ."[baskerville]"will have the needed libraries for running in baskervile (which need to be complemented with torch and tochvision modules).python -m pip install ."[all]"should have everything to train the model locally.python -m pip install ."[napari]"with have the base version of dependencies from 1. plus the version of pydantic that is compatible with napari.python -m pip install ."[tests]"should have everything to train the model locally and run tests.