milesial comments

Results 62 comments of


                                            milesial

Why is the loss combination of CrossEntropy and Dice used?

Above answers are correct

When I change the dataset, the code of 'loss = criterion(masks_pred, true_masks)' of train.py put error

Closing this, it should be fixed in the latest master.

Questions about 'predict.py''

`-i` and `-o` expect input and output individual images, not folders. You could just call this script in a bash or python loop over all your images.

Foreach gradient clipping

I added tests and used `_group_tensors_by_device_and_dtype`, let's see what CI says

Foreach gradient clipping

@janeyx99 CI is green. I tweaked the import in your util file to avoid import race issues.

@janeyx99 I rebased but the signature change is breaking torch XLA since the patching there expects the old signature ``` Traceback (most recent call last): File "/var/lib/jenkins/workspace/xla/test/../../test/test_view_ops.py", line 15, in...

Foreach gradient clipping

I don't think I can do it from 2) since I don't have write access to the xla repo, I only have a fork and xla.txt can't be in a...

Foreach gradient clipping

Thanks, just did. Let's see what CI thinks of it

Foreach gradient clipping

@wonjoolee95 I rebased this MR. Once it's green you can merge the XLA MR. Then I'll update the pin on this MR and force merge.