Apply torch.compile to PyTorch models for accelerated training and inference.
npx skills add https://github.com/tylertitsworth/skills --skill torch-compile
Please help me install this Agent Skill. Skill: torch-compile Source: https://github.com/tylertitsworth/skills/tree/main/torch-compile Command: npx skills add https://github.com/tylertitsworth/skills --skill torch-compile
PyTorch model compilation with torch.compile to accelerate training and inference, optimize graphs, reduce Python overhead, and tune performance via Inductor and backends.
Run torch.compile on your PyTorch model to enable end-to-end acceleration.
High-intent search queries and answers about installing and using this skill.
Use torch.compile to optimize PyTorch models by reducing Python overhead and accelerating training and inference on GPUs and CPUs through configurable modes and backends.
Graph breaks disrupt PyTorch compilation when encountering unsupported operations, which you can diagnose by tuning Inductor options and comparing backends to optimize the execution graph.
Yes, torch.compile supports GPU and CPU workloads handling dynamic shapes, allowing you to prepare optimized models for deployment with configurable shapes and caching.
Autotune Inductor options and backends for PyTorch performance optimization via a configurable dictionary, selecting modes and backends to reduce Python overhead and tune execution graphs.
Avoid using torch.compile when your PyTorch model operations consistently trigger graph breaks or when dynamic shapes cannot be easily configured, preventing effective Inductor optimization.
aayushsoamStandardize PyTorch training with device-agnostic and reproducible patterns.
NousResearchRun LLM inference on CPUs, Apple Silicon, and AMD GPUs with GGUF quantization.
AquiveMediaAnalyze Tideways profiling data to locate PHP performance bottlenecks.
SnakeORefactor PyTorch code with modular nn.Module, torch.compile, AMP, and optimized DataLoader.
