search

atong01 / conditional-flow-matching

TorchCFM: a Conditional Flow Matching library
https://arxiv.org/abs/2302.00482
MIT License
1.25k stars 101 forks source link

Added the ForestFlow example #66

Closed AlexiaJM closed 1 year ago

AlexiaJM commented 1 year ago

This adds the ForestFlow example. Let me know if the format is okay. I left the pip install for the special library requirements (except for torchcfm, which is assumed installed).

review-notebook-app[bot] commented 1 year ago

Check out this pull request on  ReviewNB

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

kilianFatras commented 1 year ago

Thank you for your contribution @AlexiaJM! Indeed, the requirements are directly in the notebook. In all other sections of the library, we have a requirement.txt file that users have to install to run specific parts of the library. I would prefer some consistency and adding a requirements.txt file. @atong01 @josephdviviano what is your opinion?

Also, note that I have just added a test to the forest-flow branch. Everything passed on my side. Note that I still have to add tests for the guidance functions.

atong01 commented 1 year ago

We should add optional dependencies. This would make testing easier. Ideally the syntax would be pip install torchcfm pip install torchcfm[forest] pip install torchcfm[image] pip install torchcfm[singlecell]

AlexiaJM commented 1 year ago

I removed the installs from the notebook. I incremented the torchcfm package version to be able to add some optional requirements. You can manually test it using pip install .[examples].

I put everything related to examples into 'examples', but feel free to organize it differently.

josephdviviano commented 1 year ago

Sorry I'm late to this. I think it's easiest if you group all optional deps under .[examples] as already mentioned, just to make maintenance of your pyproject.toml easier. We do something similar in torchgfn (we have an optional .[scripts] install flag).