debug:pytorch

What problem does it solve?

Debug PyTorch issues systematically. Use when encountering tensor errors, CUDA out of memory errors, gradient problems like NaN loss or exploding gradients, shape mismatches between layers, device conflicts between CPU and GPU, autograd graph issues, DataLoader problems, dtype mismatches, or training instabilities in deep learning workflows.

Core Features & Use Cases

• Systematic debugging: Structured steps to locate root causes across model code, data pipelines, and training loops.
• Error coverage: Handles CUDA OOM, NaN/Inf losses, shape mismatches, device placement, inplace operations, and DataLoader stability.
• Use Case: For a model failing with NaN loss, apply this Skill to enable anomaly detection, inspect logit shapes, verify devices, and stabilize training.

Quick Start

Use the PyTorch debugging skill to reproduce a failing training step with a minimal repro.