Profile ExecuTorch model execution with etdump and Inspector for operator timing.
npx skills add https://github.com/pytorch/executorch --skill profile-pytorch
Please help me install this Agent Skill. Skill: profile Source: https://github.com/pytorch/executorch/tree/main/.claude/skills/profile Command: npx skills add https://github.com/pytorch/executorch --skill profile-pytorch
Profile ExecuTorch model execution to measure performance, analyze operator timing, and identify bottlenecks during on-device inference.
Run a profiling session by loading your model with etdump enabled and inspect results with the Inspector.
High-intent search queries and answers about installing and using this skill.
To profile ExecuTorch model inference latency on mobile devices, load your model with etdump enabled and inspect the generated instrumentation results to measure on-device execution time and identify bottlenecks.
Operator timing analysis for on-device execution measures the execution time of individual layers or operations within a model, enabling developers to pinpoint slow operators during mobile and embedded inference workflows.
Debug slow operator performance in ExecuTorch by enabling etdump during model load to collect timing data, then use the Inspector tool to analyze the operator distribution across layers and pinpoint execution bottlenecks.
Yes, you can measure throughput and latency for embedded device inference by profiling model execution, which quantifies performance metrics and presents operator timing results via the Inspector tool.
Yes, you need etdump enabled during model load to collect execution time metrics, as it generates the instrumentation data required by the Inspector tool to present timing insights and identify performance bottlenecks.
huahai2022Profile ExecuTorch model execution to collect latency and operator timing metrics.
acaprinoProfile Python CPU and memory usage with cProfile, memory_profiler, and line_profiler.
MsabihProfile ExecuTorch model execution with per-operator timings and ETDump export.
charlesjones-devIdentify performance bottlenecks in software applications using profiling and benchmarks.
