bug fix in from_datasets method

[jakubgalik-digica]

What does this PR do?

Fixes #<issue_number>

Does your PR introduce any breaking changes? If yes, please list them.

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or minor internal changes/refactors)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing, make sure you have read the review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

[rohitgr7]

I don't think it's a bug since LightningDataModule doesn't have batch_size as an argument.

[awaelchli]

We can't assume the datamodule takes a batch_size argument in general. In which library did you find this problem?

[jakubgalik-digica]

What do you mean which library? pytorch lightning library.

[jakubgalik-digica]

I tried to use SemanticSegmentationData.from_dataset method

[jakubgalik-digica]

and i get (sorry about pic of error, instead of str)

[rohitgr7]

good catch, in that case we need to add another **kwargs parameter and check if the existing ones are part of it or not.

[carmocca]

From skimming the flash.DataModule source code, it doesn't look like this method is supposed to be compatible with it, as it already manages creating the DataLoaders etc.

https://github.com/Lightning-AI/lightning-flash/blob/644f2b559f87b40a5b623f5260eee0cf924ea0a5/flash/core/data/new_data_module.py

Thoughts @krshrimali?

[awaelchli]

Flash has an extra parameter **datamodule_kwargs in their special from_* methods. But this is unfortunately not a universal solution, because a parameter like batch_size, which is also used as a parameter by the from_dataset method, can't be passed in "twice" as kwargs. There are some solutions

1. use a proper dict datamodule_kwargs as parameter instead of kwargs. But this would be inconsistent with how Flash handles it.
2. Have datamodule_kwargs as proper kwargs, but do a signature inspection for the init to know whether you need to pass in parameters like batch_size
3. Override the from_datasets method in FlashDatamodule and raise an error that this method is not supported.
4. Something else?

[rohitgr7]

i'd say 2

[awaelchli]

Implemented in #14185

[rohitgr7]

closing in favor of #14185