No way to silence the `litmodels` advertisement

[ruro]

Bug description

Running trainer.fit with default checkpointing currently prints

💡 Tip: For seamless cloud uploads and versioning, try installing [litmodels](https://pypi.org/project/litmodels/) to enable LitModelCheckpoint, which syncs automatically with the Lightning model registry.

This advertisement is displayed even if litmodels is installed, but not used.

I don't need any "cloud uploads and versioning". In fact, I am often working with proprietary models and restricted data. I sincerely hope that this "tip" isn't suggesting that lightning is going to automatically upload anything anywhere (whether I have this litmodels package installed or not).

Please, do one of the following (in order of personal preference):

1. remove this advertisement
2. only show this advertisement when appropriate (perhaps, if you detect that the user does have litmodels installed or is already using some "cloud"-related stuff)
3. provide a way to silence this advertisement

What version are you seeing the problem on?

v2.5

Reproduced in studio

No response

How to reproduce the bug

model = ...
dl_train = ...
dl_valid = ...
trainer = L.Trainer(accelerator="auto", max_epochs=3)
trainer.fit(model, dl_train, dl_valid)


As far as I can tell, the model and dataloaders are irrelevant to this issue.

Error messages and logs

💡 Tip: For seamless cloud uploads and versioning, try installing [litmodels](https://pypi.org/project/litmodels/) to enable LitModelCheckpoint, which syncs automatically with the Lightning model registry.

Environment

Current environment

#- PyTorch Lightning Version (e.g., 2.5.0): 2.5.5
#- PyTorch Version (e.g., 2.5): 2.8.0
#- Python version (e.g., 3.12): 3.12.11
#- OS (e.g., Linux): Linux
#- CUDA/cuDNN version: 12.8
#- GPU models and configuration: any model, default configuration
#- How you installed Lightning(`conda`, `pip`, source): pip

More info

No response

[bhimrazy]

Hi @ruro,
This might be intentional to help promote the package, but in case anyone prefers to filter the tip message, the following snippet seemed to work for me:

import logging

class TipFilter(logging.Filter):
    def filter(self, record):
        return "💡 Tip" not in record.getMessage()

logging.getLogger('lightning.pytorch.utilities.rank_zero').addFilter(TipFilter())

[ruro]

By "silence this advertisement" I meant by configuring the Trainer. For example, if the user is explicitly requesting enable_checkpointing=True or passing a checkpoint callback, then they are probably not interested in your cloud-based checkpointing alternative.

Obviously, any warning or message can be filtered out in a bunch of ways (we are working in a Turing-complete language, after all). But I don't think that it's appropriate to expect that the user will add custom log filtering in order to remove ads from the output of the program.

[deependujha]

Hi @ruro,

litmodels is a new product from the Lightning team, and this tip helps users discover and try it so we can gather early feedback.

We understand some users may not want such messages, especially in controlled environments. You’re right, users shouldn’t need to modify logging just to hide non-essential tips. We’ll revisit how and when these are shown and likely add a Trainer flag or env variable to disable them.

The goal isn’t to advertise but to surface new integrations in a lightweight way, while keeping the experience clean and predictable.

Thanks for the thoughtful feedback!