As discussed at Kaggle HPA-Cell-Segmentation dependencies too strict, pip initially failed to install this package, as I have pytorch version 1.5.0, plus some dependency on that version (I no longer have the logs from the install-sorry). I fixed the problem by modifying requirements.txt to use >= instead of == for pytorch & torchvision, and tested (Python 3.8.5, Windows 10 64 bit) with the code from here. Segmentation now runs beautifully.
As discussed at Kaggle HPA-Cell-Segmentation dependencies too strict, pip initially failed to install this package, as I have pytorch version 1.5.0, plus some dependency on that version (I no longer have the logs from the install-sorry). I fixed the problem by modifying requirements.txt to use >= instead of == for pytorch & torchvision, and tested (Python 3.8.5, Windows 10 64 bit) with the code from here. Segmentation now runs beautifully.