Installation¶

Requirements¶

Python: 3.8 or higher
PyTorch: 1.0 or higher
NumPy: For numerical operations
SciPy: For statistical functions
scikit-learn: For K-means initialization and metrics

Installation Methods¶

From PyPI (Recommended)¶

The easiest way to install tgmm is using pip:

pip install tgmm

This will install the latest stable release along with all required dependencies.

From Source (Development)¶

For the latest features or if you want to contribute:

# Clone the repository
git clone https://github.com/adriansousapoza/TorchGMM.git
cd TorchGMM

# Install in editable mode
pip install -e .

The -e flag installs in editable mode, so changes to the source code are immediately available.

With GPU Support¶

tgmm automatically uses CUDA if available. To enable GPU acceleration:

Install CUDA-enabled PyTorch following the official PyTorch installation guide

Verify CUDA availability:

import torch
print(f"CUDA available: {torch.cuda.is_available()}")
print(f"CUDA device: {torch.cuda.get_device_name(0)}")

Use GPU in tgmm:

from tgmm import GaussianMixture

gmm = GaussianMixture(n_components=3, n_features=2, device='cuda')
gmm.fit(X)  # X will be automatically moved to GPU

Optional Dependencies¶

For running tutorials and visualization:

pip install jupyter matplotlib seaborn

For building documentation:

pip install mkdocs mkdocs-material mkdocstrings[python] mkdocs-jupyter

Verification¶

Verify your installation:

import tgmm
print(f"tgmm version: {tgmm.__version__}")

# Test basic functionality
from tgmm import GaussianMixture
import torch

X = torch.randn(100, 2)
gmm = GaussianMixture(n_components=3, n_features=2)
gmm.fit(X)
labels = gmm.predict(X)
print(f"Fitted GMM with {gmm.n_components} components")
print(f"Converged: {gmm.converged_}")

Troubleshooting¶

Import Errors¶

If you encounter import errors:

# Ensure all dependencies are installed
pip install torch numpy scipy scikit-learn

# Reinstall tgmm
pip uninstall tgmm
pip install tgmm

CUDA Errors¶

If you have CUDA-related issues:

Check PyTorch CUDA installation:

import torch
print(torch.version.cuda)  # Should show CUDA version

Fall back to CPU:

gmm = GaussianMixture(n_components=3, n_features=2, device='cpu')

Version Conflicts¶

If you have version conflicts with dependencies:

# Create a fresh virtual environment
python -m venv tgmm_env
source tgmm_env/bin/activate  # On Windows: tgmm_env\Scripts\activate

# Install tgmm
pip install tgmm

Next Steps¶

Next steps: Check out the Quick Start Guide or explore the Tutorials. - API Reference - Complete API documentation