torch-stack

Install PyTorch ecosystem with automatic version compatibility via extras.

Motivation

PyTorch's ecosystem has complex version dependencies between torch, torchvision, torchaudio, and torchtext. Finding compatible versions is tedious and error-prone - this is the "PyTorch version hell."

Before torch-stack:

# Manual version hunting and compatibility checking
pip install torch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0

With torch-stack:

# Just specify torch version, get compatible ecosystem automatically
pip install torch-stack[vision,audio]==2.1.0

torch-stack solves this by:

• 🎯 Extras notation: torch-stack[vision]==2.1.0 installs compatible versions
• 📦 No manual lookup: Automatically resolves ecosystem compatibility
• 🔄 Zero configuration: Works out of the box with correct version mapping

Installation

Basic Installation

# Base torch only
pip install torch-stack==2.1.0

# With torchvision (gets compatible 0.16.0)
pip install torch-stack[vision]==2.1.0

# With audio and vision
pip install torch-stack[vision,audio]==2.1.0

# Everything
pip install torch-stack[all]==2.1.0

CPU/CUDA Specific Installation

For specific hardware targets, use PyTorch's index URLs:

# CPU-only installation
pip install torch-stack[all]==2.1.0 --extra-index-url https://download.pytorch.org/whl/cpu

# CUDA 11.8
pip install torch-stack[all]==2.1.0 --extra-index-url https://download.pytorch.org/whl/cu118

# CUDA 12.1
pip install torch-stack[all]==2.1.0 --extra-index-url https://download.pytorch.org/whl/cu121

Available Index URLs

• CPU: https://download.pytorch.org/whl/cpu
• CUDA 11.8: https://download.pytorch.org/whl/cu118
• CUDA 12.1: https://download.pytorch.org/whl/cu121
• ROCm 5.6: https://download.pytorch.org/whl/rocm5.6

Note: torch-stack resolves version compatibility; the index URL determines CPU/GPU variant.

Usage

Simple Installation

Replace your manual PyTorch installs:

# Instead of researching compatible versions:
pip install torch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0

# Just use:
pip install torch-stack[vision,audio]==2.1.0

In requirements.txt

# Before: manual compatibility management
torch==2.1.0
torchvision==0.16.0
torchaudio==2.1.0

# After: automatic compatibility
torch-stack[vision,audio]==2.1.0

In pyproject.toml

[project]
dependencies = [
    "torch-stack[vision,audio]==2.1.0"
]

Available Extras

• vision - Installs compatible torchvision
• audio - Installs compatible torchaudio
• text - Installs compatible torchtext

Python API (for package maintainers)

from torch_stack.resolver import VersionResolver

# Get compatible versions programmatically
torch_ver = "2.1.0"
vision_ver = VersionResolver.torchvision(torch_ver)  # "0.16.0"
audio_ver = VersionResolver.torchaudio(torch_ver)  # "2.1.0"

Supported Versions

• PyTorch 1.8+ through latest
• Handles version exceptions and edge cases
• CPU and CUDA installation support

No more PyTorch version hell 🎉