New tabpfn model

[jackjii79]

https://github.com/h2oai/h2oai/issues/34827

Due to the complexity of TabPFN, automation regression is skipped, manual testing results show down below

[Copilot]

Pull request overview

This PR adds new TabPFN (Prior-data Fitted Networks) model functionality to Driverless AI, introducing both an outlier detection transformer and a supervised model for classification and regression tasks. The implementation includes GPU-accelerated inference, SHAP-based explanations, and support for datasets with up to 10,000 samples.

Key changes:

• Implements TabPFN-based unsupervised outlier detection with density estimation via permutation sampling
• Adds TabPFN supervised model with support for regression, binary, and multi-class classification (including many-class scenarios via codebook-based decomposition)
• Includes SAGE/Shapley value computation for global feature importance and local explanations

Reviewed changes

Copilot reviewed 1 out of 1 changed files in this pull request and generated 20 comments.

File	Description
transformers/outliers/tabpfn_outlier.py	Implements TabPFN-based outlier detection transformer that calculates negative log probability scores using conditional probability estimation across feature permutations
models/algorithms/tabpfn_model.py	Implements TabPFN supervised model with regression/classification support, many-class handling via codebook decomposition, and SHAP-based explanations using SAGE simulation

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[jackjii79]

Testing results:

• TabPFNModel
  ✅ Classification
  ✅ Regression