Profile ExecuTorch model execution to collect latency and operator timing metrics.
npx skills add https://github.com/huahai2022/execu-xnnpack --skill profile-huahai2022
Please help me install this Agent Skill. Skill: profile Source: https://github.com/huahai2022/execu-xnnpack/tree/main/.claude/skills/profile Command: npx skills add https://github.com/huahai2022/execu-xnnpack --skill profile-huahai2022
Profile ExecuTorch model execution to collect performance metrics, helping you measure latency and operator timing for on-device inference.
Run the profile tool on a loaded model to start collecting etdump and timing data.
High-intent search queries and answers about installing and using this skill.
To profile on-device performance, you enable ETDump during model load and run a forward pass to capture per-layer execution times and structured operator timing metrics for ExecuTorch.
ETDump is a data collection mechanism enabled during model load to gather deterministic profiling data. It captures detailed timing data during execution, allowing you to debug slow models and measure precise latency across devices.
Yes, you can use the Inspector to visualize and compare timing results across runs and devices. This helps analyze collected ETDump metrics for performance optimization and benchmarking during on-device inference.
You need to profile on-device model execution when benchmarking inference speed, debugging slow models, or optimizing performance across devices. It provides precise latency measurements and operator timing data for structured analysis.
MsabihProfile ExecuTorch model execution with per-operator timings and ETDump export.
pytorchProfile ExecuTorch model execution with etdump and Inspector for operator timing.
ArogyaReddyEvaluate trading strategies with walk-forward analysis and realistic transaction costs.
aaronachermannInstruments Python and TypeScript apps with MLflow tracing for LLM calls and retrievals.
