Skip to content

installation.md

Latest commit

 

History

History
187 lines (122 loc) · 3.58 KB

installation.md

File metadata and controls

187 lines (122 loc) · 3.58 KB

Installation

This guide covers different installation options for ShortKit-ML.

Requirements

  • Python 3.9, 3.10, or 3.11
  • pip or uv package manager
  • PyTorch (torch) and torchvision (included in core install; see PyTorch with CUDA for GPU)

Quick Install

From PyPI (Recommended)

pip install shortcut-detect

From Source

git clone https://github.com/criticaldata/ShortKit-ML.git
cd Shortcut_Detect
pip install -e .

Installation Options

The library has several optional dependency groups for different use cases:

Basic Installation

Core functionality with all detection methods:

pip install shortcut-detect

With Dashboard

For the interactive Gradio web interface:

pip install "shortcut-detect[dashboard]"

With Jupyter Support

For running examples in Jupyter notebooks:

pip install "shortcut-detect[jupyter]"

With Reporting

For PDF/HTML report generation:

pip install "shortcut-detect[reporting]"

With Hugging Face

For embedding generation from HuggingFace models:

pip install "shortcut-detect[hf]"

With VAE

VAE shortcut detection requires torch and torchvision; both are core dependencies and included in the default install. To explicitly install VAE support:

pip install "shortcut-detect[vae]"

Full Installation

Install everything (recommended for development):

pip install "shortcut-detect[all]"

Using uv (Faster)

uv is a fast Python package installer:

# Install uv
curl -LsSf https://astral.sh/uv/install.sh | sh

# Create virtual environment
uv venv --python 3.10
source .venv/bin/activate  # Windows: .venv\Scripts\activate

# Install with all dependencies
uv pip install -e ".[all]"

PyTorch with CUDA

For GPU acceleration (optional):

# CUDA 12.1
pip install torch==2.1.0+cu121 torchvision==0.16.0+cu121 \
  --index-url https://download.pytorch.org/whl/cu121

# CUDA 11.8
pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 \
  --index-url https://download.pytorch.org/whl/cu118

Jupyter Kernel Setup

To use the library in Jupyter notebooks:

# Install Jupyter support
pip install "shortcut-detect[jupyter]"

# Register the kernel
python -m ipykernel install --user --name=shortcut_detect --display-name="Python (Shortcut_Detect)"

# Launch Jupyter
jupyter lab

Verify Installation

import shortcut_detect
print(shortcut_detect.__version__)  # Should print: 0.1.0

# Test imports
from shortcut_detect import ShortcutDetector, HBACDetector, SKLearnProbe
print("All imports successful!")

Troubleshooting

Import Errors

If you get import errors, ensure all dependencies are installed:

pip install -e ".[all]" --force-reinstall

CUDA Issues

If PyTorch doesn't detect your GPU:

import torch
print(torch.cuda.is_available())  # Should be True
print(torch.cuda.device_count())  # Number of GPUs

WeasyPrint Issues (PDF Generation)

WeasyPrint requires system dependencies:

=== "Ubuntu/Debian"

```bash
sudo apt-get install libpango-1.0-0 libpangocairo-1.0-0 libgdk-pixbuf2.0-0
```

=== "macOS"

```bash
brew install pango gdk-pixbuf libffi
```

=== "Windows"

Install GTK3 runtime from [gtk.org](https://www.gtk.org/docs/installations/windows/)

Next Steps