Pytorch Patterns

This is a collection of PyTorch idioms that will save you from common mistakes like hardcoding CUDA calls or forgetting to set model.train() and model.eval(). It covers the full stack: device-agnostic code, reproducibility setup with proper seeding, shape-annotated forward passes, training loops with mixed precision and gradient clipping, and DataLoader configurations that actually use your CPU cores. The checkpoint patterns include optimizer state, which people forget surprisingly often. If you're writing PyTorch from scratch or reviewing someone else's training code, this gives you the patterns that separate working code from production-ready code.

skills.sh installs: 5.8K
GitHub stars: 300.3K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add affaan-m/ecc --skill pytorch-patterns --agent claude-code

Installs into .claude/skills of the current project.

Files

PyTorch Development Patterns

Idiomatic PyTorch patterns and best practices for building robust, efficient, and reproducible deep learning applications.

When to Activate

Writing new PyTorch models or training scripts
Reviewing deep learning code
Debugging training loops or data pipelines
Optimizing GPU memory usage or training speed
Setting up reproducible experiments

Core Principles

1. Device-Agnostic Code

Always write code that works on both CPU and GPU without hardcoding devices.

# Good: Device-agnostic
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = MyModel().to(device)
data = data.to(device)

# Bad: Hardcoded device
model = MyModel().cuda()  # Crashes if no GPU
data = data.cuda()

2. Reproducibility First

Set all random seeds for reproducible results.

# Good: Full reproducibility setup
def set_seed(seed: int = 42) -> None:
    torch.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)
    np.random.seed(seed)
    random.seed(seed)
    torch.backends.cudnn.deterministic = True
    torch.backends.cudnn.benchmark = False

SKILL.md on GitHub