Hey and thanks for a great repository.
While setting the code up I found that some of the dependencies were missing from the readme.
Specifically, these are scikit-image, h5py and ml_collections.
Additionally, in more recent versions of scikit-image, compare_psnr is moved and renamed from skimage.measure.compare_psnr to skimage.metrics.peak_signal_noise_ratio.
In more recent versions of pytorch it also seems that the .cuda(async = ...) keyword is replaced with .cuda(non_blocking = ... ) since async is a reserved keyword in python >= 3.7.
I see two ways to fix the issues; either by being more specific in terms of compatible package versions or, my preferred solution, updating the codebase to new keywords and functions.
I'd gladly submit a pull request with the mentioned changes if you'd like me to.
I also made a dockerfile for setting up an environment with the requirements if that should be of any interest.
Hey and thanks for a great repository.
While setting the code up I found that some of the dependencies were missing from the readme.
Specifically, these are
scikit-image,h5pyandml_collections.Additionally, in more recent versions of
scikit-image,compare_psnris moved and renamed fromskimage.measure.compare_psnrtoskimage.metrics.peak_signal_noise_ratio.In more recent versions of
pytorchit also seems that the.cuda(async = ...)keyword is replaced with.cuda(non_blocking = ... )sinceasyncis a reserved keyword inpython >= 3.7.I see two ways to fix the issues; either by being more specific in terms of compatible package versions or, my preferred solution, updating the codebase to new keywords and functions.
I'd gladly submit a pull request with the mentioned changes if you'd like me to.
I also made a
dockerfilefor setting up an environment with the requirements if that should be of any interest.