Installation

Installation#

Requirements#

  • Python 3.10 or higher

  • PyTorch 2.0 or higher

  • NumPy

  • scikit-learn

  • scipy

Install from PyPI#

Note

PyPI release coming soon!

Once released, you can install incerto using pip:

pip install incerto

Install from Source#

For development or to get the latest features:

git clone https://github.com/steverab/incerto.git
cd incerto
pip install -e .

Development Installation#

To install with development dependencies (for contributing):

git clone https://github.com/steverab/incerto.git
cd incerto
pip install -e ".[dev]"

Or if using uv:

git clone https://github.com/steverab/incerto.git
cd incerto
uv pip install -e .

Verify Installation#

To verify that incerto is installed correctly:

import incerto
print(incerto.__version__)

# Test imports
from incerto import calibration, ood, conformal, sp, llm, bayesian, active, shift
print("All modules imported successfully!")

GPU Support#

incerto automatically uses CUDA if PyTorch is installed with GPU support:

import torch
print(f"CUDA available: {torch.cuda.is_available()}")

Optional Dependencies#

Some features may require additional packages:

  • transformers: For LLM uncertainty methods

  • matplotlib: For visualization utilities

  • rich: For pretty output formatting

These are included in the base installation.

Troubleshooting#

ImportError: No module named ‘incerto’

Make sure you installed the package and are not in the source directory

PyTorch version conflicts

Ensure PyTorch 2.0+ is installed: pip install --upgrade torch

CUDA errors

Check PyTorch CUDA compatibility: torch.cuda.is_available()

Contents