README.md

---
extra_gated_prompt: |-
  By accessing TabPFN, you agree to:
      1. Not use the model in ways that could harm individuals or communities
      2. Comply with all applicable laws and regulations
      3. Properly cite the model and its creators in any resulting publications
      4. Report any discovered vulnerabilities or safety concerns to Prior Labs
extra_gated_fields:
  Organization:
    type: text
    required: true
    description: Company or institution you represent
  Role:
    type: text
    required: true
    description: Your role in the organization
  Country:
    type: country
    required: true
    description: Country where you or your organization is based
  Intended Use:
    type: select
    required: true
    options:
    - Academic Research
    - Education/Teaching
    - Commercial Evaluation
    - Non-profit Use
    - Personal Learning
    - label: Other
      value: other
    description: Primary intended use of TabPFN
  Industry:
    type: select
    required: true
    options:
    - Healthcare/Life Sciences
    - Financial Services
    - Technology
    - Education
    - Manufacturing
    - Research Institution
    - label: Other
      value: other
    description: Your industry sector
  Dataset Size:
    type: select
    required: true
    options:
    - <1000 rows
    - 1000-10000 rows
    - 10000-100000 rows
    - '>100000 rows'
    description: Typical size of datasets you plan to use
  License Agreement:
    type: checkbox
    required: true
    label: >-
      I agree to the terms of the non-commercial license for research and
      evaluation
  Contact Permission:
    type: checkbox
    required: false
    label: Prior Labs may contact me about my use case and provide support (optional)
pipeline_tag: tabular-classification
---

# Model Card for TabPFN-v2

TabPFN is a transformer-based foundation model for tabular data that leverages prior-data based learning to achieve strong performance on small tabular datasets without requiring task-specific training.

## Model Details

### Model Description

TabPFN is a novel approach to tabular data modeling that uses transformer architectures combined with prior knowledge injection to create a foundation model specifically designed for tabular data tasks.

- **Developed by:** Prior Labs
- **Model type:** Transformer-based foundation model for tabular data
- **Language(s):** Python
- **License:** Dual licensing - Open source for research/non-commercial use
- **Finetuned from model:** Custom architecture, trained from scratch

### Model Sources

- **Repository:** https://github.com/priorlabs/tabpfn
- **Paper:** [More Information Needed]
- **Demo:** Available via API access

## Uses

### Direct Use

TabPFN can be directly used for:
- Classification tasks on small to medium-sized tabular datasets
- Automated machine learning workflows
- Quick prototyping and baseline model creation
- Transfer learning applications for tabular data

### Downstream Use

The model can be used as:
- A feature extractor for downstream tasks
- A foundation for transfer learning on domain-specific tabular data
- A component in automated ML pipelines
- A baseline model for benchmarking

### Out-of-Scope Use

- The model is not designed for:
  - Very large datasets (currently optimized for smaller datasets)
  - Non-tabular data formats
  - Time series forecasting
  - Direct regression tasks

## Bias, Risks, and Limitations

- Performance may vary based on dataset size and characteristics
- Model behavior heavily depends on the quality and representativeness of training data
- May not perform optimally on highly imbalanced datasets
- Resource intensive for very large datasets

### Recommendations

- Use on datasets with clear structure and well-defined features
- Validate model outputs especially for sensitive applications
- Consider dataset size limitations when applying the model
- Monitor performance across different subgroups in the data

## How to Get Started with the Model

```python
from tabpfn import TabPFNClassifier

# Initialize model
classifier = TabPFNClassifier()

# Fit and predict
classifier.fit(X_train, y_train)
predictions = classifier.predict(X_test)
```

## Training Details

### Training Data

[More Information Needed]

### Training Procedure

#### Training Hyperparameters

- **Training regime:** Mixed precision training

## Evaluation

### Testing Data, Factors & Metrics

#### Metrics

- Classification accuracy
- F1 score
- ROC-AUC
- Precision-Recall curves

### Results

[More Information Needed]

## Environmental Impact

- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]

## Technical Specifications

### Model Architecture and Objective

TabPFN uses a transformer-based architecture specifically designed for tabular data processing, with modifications to handle varying input sizes and feature types.

### Compute Infrastructure

#### Hardware

Recommended minimum specifications:
- CPU: Modern multi-core processor
- RAM: 16GB+
- GPU: Optional, CPU inference supported

#### Software

- Python 3.7+
- Key dependencies: PyTorch, NumPy, Pandas

## Model Card Contact

For more information, contact Prior Labs.