Reformat notebooks

[ghost]

Description

• Incorporate pip installs into the notebook [TypeError in transforms notebook #851]
• Incorporate dataset download into the notebook
• Add an environmental variable to choose directory
• Use a temporary directory as a fall back
• Remove unused imports
• Apply isort and black formatting
• Replace %-formatting with f-string
• Unify section order, naming, contents
• Clear display for overly long outputs

Status

Ready

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).

[ghost]

There may be a bug in spleen_segmentation_3d_lightning.ipynb.
See "Check best model output with the input image and label".

RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

[gagandaroach]

There may be a bug in spleen_segmentation_3d_lightning.ipynb.
See "Check best model output with the input image and label".

RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

I ran into this problem before. I was able to fix it in the notebook by sending the net to the gpu device. In the last cell, add a net.to(device) line like this

net.eval()
device = torch.device("cuda:0")
net.to(device)
with torch.no_grad():
    for i, val_data in enumerate(net.val_dataloader()):

The weight tensor is on cpu and the incoming tensor is on the gpu. Hard for the computer to multiply those together :)

[wyli]

There may be a bug in spleen_segmentation_3d_lightning.ipynb.
See "Check best model output with the input image and label".

RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

I ran into this problem before. I was able to fix it in the notebook by sending the net to the gpu device. In the last cell, add a net.to(device) line like this

net.eval()
device = torch.device("cuda:0")
net.to(device)
with torch.no_grad():
    for i, val_data in enumerate(net.val_dataloader()):

The weight tensor is on cpu and the incoming tensor is on the gpu. Hard for the computer to multiply those together :)

please could you file an issue about it