My environment lists and some bugs fixed for reference

[qingtianwu]

scikit-image 0.19.2 pypi_0 pypi

tensorboardx 1.2 pypi_0 pypi
torch 1.8.1+cu101 pypi_0 pypi
torchaudio 0.8.1 pypi_0 pypi
torchgeometry 0.1.2 pypi_0 pypi
torchtext 0.9.0 pypi_0 pypi
torchvision 0.9.1+cu101 pypi_0 pypi
torchviz 0.0.1 pypi_0 pypi
albumentations 0.5.2 pypi_0 pypi

And fix some bugs:
1、 When classfication,
def forward(self, x):
x = self.encoder(x)[0] -----------------> x = self.encoder(x)[-1]
2、 # df['ImageId'], df['ClassId'] = zip(*df['ImageId_ClassId'].str.split('_'))
The new kaggle dataset has fixed the new heads of the table.