Update class_weights to support list[float]

[robmarkcole]

Close #2060

Could be cleaner to deprecate tensor support, not sure if there is a process to this?

Validated with

trainer:
  max_epochs: 1
model:
  class_path: ClassificationTask
  init_args:
    model: 'resnet18'
    in_channels: 13
    num_classes: 10
    class_weights:
      - 1.0
      - 1.0
      - 1.0
      - 1.0
      - 1.0
      - 1.0
      - 1.0
      - 1.0
      - 1.0
      - 1.0
data:
  class_path: EuroSAT100DataModule
  init_args:
    batch_size: 8
  dict_kwargs:
    download: true

[robmarkcole]

Not sure why docs fail to build

[adamjstewart]

Could use https://docs.python.org/3/library/warnings.html#warnings.deprecated or explicitly raise https://docs.python.org/3/library/exceptions.html#DeprecationWarning for deprecation.

I'm not sure why the doc tests are failing either, but you're not the only PR with failing tests. Will try to investigate this later if it keeps happening.

[adamjstewart]

Restarting the docs build fixed the other PR, hoping this is just a transient issue.

[robmarkcole]

Hmm did this change break this test?

FAILED tests/models/test_copernicusfm.py::TestCopernicusFMBase::test_copernicusfm_variable - AssertionError: The input tensor is not within the configured range `[0.001, 510000000.0]`.

[adamjstewart]

Nope, this is because I used random numbers which are sometimes outside the valid range. Will submit a fix. If you rerun it should only fail 1% of the time.

[robmarkcole]

Is there a rerun button somewhere? @adamjstewart

[adamjstewart]

Yes, just hit it. I don't think you have access, but I'm happy to rerun until CI is happy.

[robmarkcole]

If you could - I dont see the button 🥹

[robmarkcole]

@adamjstewart thoughts?

[adamjstewart]

Let's discuss more on #2060