# Profile

> Profile ExecuTorch model execution. Use when measuring performance, analyzing operator timing, or debugging slow models.

- Skill: `pytorch/profile` (Agent Skill)
- Install (CLI): `npx skillmds@latest add pytorch/profile`
- Raw SKILL.md: https://api.skillmd.com/api/skills/pytorch/profile/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: pytorch (https://skillmd.com/u/pytorch)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/pytorch/profile

---


# Profile

## 1. Enable ETDump when loading
```python
program = runtime.load_program("model.pte", enable_etdump=True, debug_buffer_size=int(1e7))
```

## 2. Execute and save
```python
outputs = program.load_method("forward").execute(inputs)
program.write_etdump_result_to_file("etdump.etdp", "debug.bin")
```

## 3. Analyze with Inspector
```python
from executorch.devtools import Inspector
inspector = Inspector(etrecord="model.etrecord", etdump_path="etdump.etdp")
inspector.print_data_tabular()
```