search

N3PDF / pycompressor

Compression code for PDF replicas.
https://n3pdf.github.io/pycompressor/
GNU General Public License v3.0
1 stars 0 forks source link

Interface GANs within pyCompressor #23

Closed Radonirinaunimi closed 4 years ago

Radonirinaunimi commented 4 years ago

This PR solves #22.

Radonirinaunimi commented 4 years ago

@scarrazza, @scarlehoff,

Everything is now put in place. The following basically summarizes the workflow of the pyCompressor: https://github.com/N3PDF/pycompressor/blob/494db7c06fe6908bb0368e83a662d6d700be210d/src/pycompressor/compressing.py#L36-L58

Both repositories (pycompressor & ganpdfs) can now be made public (if needed) and then I can take care of deploying the first version of the package to pypi and conda (for this some PRs have to be closed). Also, before doing actual hyperopt, I might some machines to run a full test and produce some results.

scarlehoff commented 4 years ago

Thanks! You did a tremendous work. I haven't been looking closely at all the PR (mainly because a lack of time) but it really looks like high-quality code :100: Looks super clean and it's easy to follow and read :)

RE public: I agree I think the first think would be to close some of the PR and I would also add documentation before making the package public. I would at least populate the how-to of both repositories (even if it is just a copy of the readme) and a from-zero-to-result example.

RE conda: conda-forge is out of the question for now for everything that contains tensorflow, but you can use the Zahari conda channel (https://packages.nnpdf.science/conda) since I guess this is a PDF tool after all.

Radonirinaunimi commented 4 years ago

Thanks! You did a tremendous work. I haven't been looking closely at all the PR (mainly because a lack of time) but it really looks like high-quality code 100 Looks super clean and it's easy to follow and read :)

Thanks :sweat_smile: ! I will try to finish at least the foundations asap to start producing some results that can be analysed.

RE public: I agree I think the first think would be to close some of the PR and I would also add documentation before making the package public. I would at least populate the how-to of both repositories (even if it is just a copy of the readme) and a from-zero-to-result example.

Adding a full documentation might take some time, but definitely the HowTo section and the results presentation can be done in very short amount of time.

RE conda: conda-forge is out of the question for now for everything that contains tensorflow, but you can use the Zahari conda channel (https://packages.nnpdf.science/conda) since I guess this is a PDF tool after all.

In this case, from the time being, it might not be worth the tears to go trough this yet.