[TabArena] Benchmarking TabFlex; Verifying our Implemenetation

[LennartPurucker]

Heyho,

We are working on benchmarking TabFlex on TabArena. I have implemented and benchmarked a first version of your model in this pull request: autogluon/tabrepo#171

There was little room for error in implementing the TICL. Please let me know if you find any problems!
Moreover, could you verify if I used the correct checkpoint (from the TICL code)?

I'm happy to discuss this further here or in the PR about how to maximize the performance of your model.
Feel free to test the implementation yourself or propose changes I should incorporate.

Once everyone is happy with the code, I can rerun the config and add its results to the public leaderboard.

Thank you!

[LennartPurucker]

Small ping @amueller @yzeng58

[LennartPurucker]

The newest paper also includes a regression result. Is the target discretization as in the paper also public? And should one use TabFlex for regression, given the results?
The main branch only has classification (https://github.com/microsoft/ticl/blob/main/ticl/prediction/tabflex.py#L9)

We are planning to add the current results we have to TabFlex for a future version of the leaderboard and would love to get your feedback to make sure it is represented correctly!