Bugfix/equiv padding weightedmse

[ritvikvasan]

What does this PR do?

This PR adds a temporary fix to vae/image_encoder where Convolution class expects padding values to be tuple or int (cannot pass None). The main issue here is that ResBlock isn't returning the correct number of vector features that need to be used to compute the pose.

cyto-dl/cyto_dl/models/vae/image_encoder.py

Line 165 in dc30155

y_pose = y_pose[:, :, [2, 0, 1]]

Here y_pose needs to be a 3D vector, but resblocks return multidimensional vectors.

For now the fix is to not use ResBlock and set num_res_units to 0.

Also added a weighted mse loss to try to focus the SDF models on the zero level set region.

cyto-dl/cyto_dl/nn/losses/weighted_mse_loss.py

Line 19 in dc30155

def forward(self, input: Tensor, target: Tensor) -> Tensor:

Also includes some additional args to PC VAE

Fixes #<issue_number>

Before submitting

• Did you make sure title is self-explanatory and the description concisely explains the PR?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you list all the breaking changes introduced by this pull request?
• Did you test your PR locally with pytest command?
• Did you run pre-commit hooks with pre-commit run -a command?

Did you have fun?

Make sure you had fun coding 🙃

[benjijamorris]

Looks good! Could you remove the commented out code?

[ritvikvasan]

will get to this soon