🚀 PyTorch Installation Assistant

An intelligent, autonomous PyTorch installer that automatically detects your system, GPU, and CUDA configuration to install the optimal PyTorch setup for your hardware.

✨ Features

• 🧠 Intelligent GPU Detection: Automatically detects NVIDIA, AMD, and Apple Silicon GPUs
• 🎯 Smart CUDA Matching: Finds the best PyTorch version for your CUDA installation
• 🤖 Autonomous CUDA Installation: Automatically installs CUDA on Windows using package managers
• 📦 Complete Ecosystem: Installs torch, torchvision, and torchaudio with version compatibility
• 🔄 Fallback Logic: Handles older CUDA versions and compatibility issues gracefully
• 🎮 Hardware-Specific Optimization: Tailored recommendations for different GPU generations
• 🔍 Comprehensive Testing: Post-install verification with tensor operations
• 📊 Detailed Reporting: Shows complete system and package information

⚠️ GPU Compatibility Notice

Testing Status: This installer has been primarily tested on GT 900 series and older GPUs, as well as GTX 10 series cards. While it should work with newer GPU generations (RTX 20/30/40 series), comprehensive testing across all NVIDIA GPU models is ongoing.

If you encounter issues with newer GPUs, please report them via GitHub issues to help improve compatibility.

�️ Installation

Simply download the torch_installer.py script - no additional dependencies required beyond Python's standard library.

# Download the script
curl -O https://raw.githubusercontent.com/coff33ninja/torch-installer/main/torch_installer.py

# Or clone the repository
git clone https://github.com/coff33ninja/torch-installer.git
cd torch-installer/pytorch-installer.git

🚀 Quick Start

Basic Installation

# Automatic installation with smart detection
python torch_installer.py

# CPU-only installation
python torch_installer.py --cpu-only

# Force specific CUDA version
python torch_installer.py --force-cuda cu121

CUDA Auto-Installation (Windows Only)

# Auto-install recommended CUDA version
python torch_installer.py --auto-install-cuda

# Install specific CUDA version
python torch_installer.py --auto-install-cuda --cuda-version 12.1

# Dry-run to see what would be installed
python torch_installer.py --auto-install-cuda --dry-run

� Commnand Reference

Core Installation Commands

Command	Description	Example
python torch_installer.py	Auto-detect and install optimal PyTorch	Basic usage
--cpu-only	Force CPU-only installation	python torch_installer.py --cpu-only
--force-cuda cu121	Force specific CUDA version	python torch_installer.py --force-cuda cu121
--force-reinstall	Reinstall even if PyTorch exists	python torch_installer.py --force-reinstall

CUDA Management (Windows)

Command	Description	Example
--auto-install-cuda	Automatically install CUDA	python torch_installer.py --auto-install-cuda
--cuda-version 12.1	Specify CUDA version to install	python torch_installer.py --auto-install-cuda --cuda-version 12.1

Information & Diagnostics

Command	Description	Example
--gpu-info	Show GPU and CUDA compatibility	python torch_installer.py --gpu-info
--show-versions	Display installed PyTorch ecosystem	python torch_installer.py --show-versions
--show-matching	Demo CUDA version matching logic	python torch_installer.py --show-matching
--list-cuda	List supported CUDA versions	python torch_installer.py --list-cuda

Development & Testing

Command	Description	Example
--dry-run	Show commands without executing	python torch_installer.py --dry-run
--log	Log all output to timestamped file	python torch_installer.py --log

🎮 GPU Support Matrix

NVIDIA GPUs

GPU Generation	Recommended CUDA	PyTorch Support	Performance
RTX 40 Series	CUDA 12.1+	✅ Excellent	🔥🔥🔥🔥🔥
RTX 30 Series	CUDA 12.1+	✅ Excellent	🔥🔥🔥🔥🔥
RTX 20 Series	CUDA 11.8+	✅ Excellent	🔥🔥🔥🔥
GTX 16 Series	CUDA 11.8+	✅ Very Good	🔥🔥🔥🔥
GTX 10 Series	CUDA 11.8+	✅ Good	🔥🔥🔥
GT 700 Series	CUDA 11.8	⚠️ Limited	🔥🔥
Older GPUs	Manual Install	❌ Not Recommended	🔥

Other GPUs

GPU Type	Support	Recommendation
Apple Silicon (M1/M2/M3)	✅ MPS Support	Automatic detection
AMD GPUs	⚠️ ROCm (Linux only)	Manual ROCm installation
Intel GPUs	❌ Not supported	Use CPU-only mode

🔧 Usage Examples

Scenario 1: First-time Installation

# Let the installer detect everything automatically
python torch_installer.py

# Output example:
# 🚀 PyTorch Installation Assistant
# 🎮 Detected GPU: GeForce RTX 3080
# 🚀 Detected CUDA version: 12.1
# 🎯 Installing PyTorch with CUDA 121 wheels
# ✅ PyTorch installation completed successfully!

Scenario 2: Upgrading CUDA and PyTorch

# Auto-install newer CUDA version
python torch_installer.py --auto-install-cuda --cuda-version 12.1

# Then reinstall PyTorch
python torch_installer.py --force-reinstall

Scenario 3: Troubleshooting Installation

# Check current setup
python torch_installer.py --show-versions

# See GPU compatibility
python torch_installer.py --gpu-info

# Test what would be installed
python torch_installer.py --dry-run

Scenario 4: Development Environment

# Install with logging for debugging
python torch_installer.py --log

# Check CUDA matching logic
python torch_installer.py --show-matching

🧠 Intelligent Features

Smart CUDA Version Matching

The installer automatically matches your CUDA version to compatible PyTorch versions:

🔍 Detected CUDA: 11.1
📋 Supported versions: ['121', '118', '117', '116', '113']
⚠️ Fallback match: CUDA 111 -> PyTorch cu113 (oldest supported)
✅ Would install: PyTorch 2.0.1 with CUDA 111
📦 Full package set: torch=2.0.1, torchvision=0.15.2, torchaudio=2.0.2

GPU-Specific Recommendations

For older GPUs:

💡 GPU ACCELERATION UPGRADE GUIDE (GeForce GT 710):
   ⚠️ Your GeForce GT 710 is an older GPU with limited CUDA support
   💡 Recommended: CUDA 11.8 for optimal compatibility
   🤖 AUTOMATIC INSTALLATION AVAILABLE:
   • Run: python torch_installer.py --auto-install-cuda

For modern GPUs:

💡 GPU ACCELERATION UPGRADE GUIDE (GeForce RTX 3080):
   🚀 Your GeForce RTX 3080 supports modern CUDA versions
   ✨ Recommended: CUDA 12.1 for best performance
   🤖 AUTOMATIC INSTALLATION AVAILABLE:
   • Run: python torch_installer.py --auto-install-cuda

🔍 System Information Display

Complete Ecosystem View

python torch_installer.py --show-versions

# Output:
# 📊 Installed PyTorch Ecosystem:
#    🔥 PyTorch: 2.8.0+cu121
#    👁️ TorchVision: 0.23.0+cu121
#    🔊 TorchAudio: 2.8.0+cu121
#    🎯 CUDA Support: True
#    🚀 CUDA Version: 12.1
#    🎮 GPU Count: 1
#    🎮 GPU 0: GeForce RTX 3080

GPU Compatibility Analysis

python torch_installer.py --gpu-info

# Output:
# 🎮 GPU and CUDA Compatibility Information
# 🎮 Detected GPU: GeForce RTX 3080
# 💾 GPU Memory: 10240MB
# 🔍 Detected CUDA: 12.1
# ✅ Latest PyTorch supports your CUDA via cu121

🤖 CUDA Auto-Installation (Windows)

Prerequisites

• Windows 10/11
• NVIDIA GPU with compatible drivers
• Package manager: winget (built-in) or chocolatey

Installation Process

1. Detection: Identifies your GPU model and current CUDA version
2. Recommendation: Suggests optimal CUDA version for your hardware
3. Package Manager Check: Verifies winget or chocolatey availability
4. Version Matching: Finds compatible CUDA version in repositories
5. Installation: Automatically downloads and installs CUDA
6. Verification: Confirms successful installation

Example Output

python torch_installer.py --auto-install-cuda

# 🤖 CUDA Auto-Installation Mode
# 🎮 Detected GPU: GeForce RTX 3080
# 📋 Current CUDA: 11.8
# 🔧 Attempting to install CUDA 12.1 for GeForce RTX 3080
# 📦 Trying winget (Windows Package Manager)...
# ✅ Found CUDA versions in winget: 13.0, 12.9, 12.1...
# 🔧 Installing CUDA 12.1 via winget...
# ✅ Successfully installed CUDA 12.1
# 🔄 Please restart your command prompt and run the installer again

🔧 Advanced Configuration

Environment Variables

• CUDA_HOME: Override CUDA installation path detection
• PYTORCH_CUDA_ALLOC_CONF: Configure CUDA memory allocation

Custom Package Managers

The installer supports:

• winget: Native Windows package manager (recommended)
• chocolatey: Third-party package manager with more versions

Offline Installation

For air-gapped environments:

1. Download PyTorch wheels manually from https://pytorch.org/get-started/locally/
2. Use pip install with local wheel files
3. Run installer with --show-versions to verify

🐛 Troubleshooting

Common Issues

"CUDA not available" after installation

# Check CUDA installation
nvidia-smi

# Verify PyTorch CUDA support
python -c "import torch; print(torch.cuda.is_available())"

# Reinstall with force
python torch_installer.py --force-reinstall

Package manager not found (Windows)

# Install chocolatey
Set-ExecutionPolicy Bypass -Scope Process -Force
iex ((New-Object System.Net.WebClient).DownloadString('https://chocolatey.org/install.ps1'))

# Or update Windows for winget (Windows 10)
# winget is included in Windows 11 by default

Older CUDA version detected

# Check what would be installed
python torch_installer.py --show-matching

# Auto-upgrade CUDA (Windows)
python torch_installer.py --auto-install-cuda

# Or force specific PyTorch version
python torch_installer.py --force-cuda cu118

Debug Mode

# Enable detailed logging
python torch_installer.py --log --dry-run

# Check system compatibility
python torch_installer.py --gpu-info --show-versions

🔄 Update & Maintenance

Updating PyTorch

# Check for updates and reinstall
python torch_installer.py --force-reinstall

# Upgrade to specific version
python torch_installer.py --force-cuda cu121 --force-reinstall

Updating CUDA (Windows)

# Auto-install latest compatible version
python torch_installer.py --auto-install-cuda

# Install specific version
python torch_installer.py --auto-install-cuda --cuda-version 12.1

🤝 Contributing

Reporting Issues

When reporting issues, please include:

# System information
python torch_installer.py --gpu-info --show-versions --log

# Attach the generated log file

Feature Requests

• GPU support for additional vendors
• Package manager support for other platforms
• Integration with conda/mamba environments

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• NVIDIA for CUDA toolkit and GPU drivers
• PyTorch Team for the excellent deep learning framework
• Microsoft for winget package manager
• Chocolatey community for package management on Windows

---

📞 Support

For support and questions:

• 📧 Create an issue on GitHub
• 💬 Join the discussion in GitHub Discussions
• 📖 Check the troubleshooting section above

Happy Deep Learning! 🚀🔥