Find what slows your model's MAX inference.
What it does
Diagnoses MAX inference bottlenecks by measuring GPU utilization and identifying the kernels that consume the most time.
When to use it
Use when asked to investigate slow MAX inference, GPU utilization, or a model's kernel-level performance.
How to use it
Give it a MAX profiling question; it reports GPU utilization or dominant kernels and, when captured, the trace's location.
Good to know
- cannot be undone: profile.nsys-rep; server_profile.nsys-rep
Latest review
What you need · 8
The documented CLI workflow expects a project with the nightly max-serve package installed via pixi.
The documented MAX CLI commands assume pixi can run MAX in the installed project environment.
Access · 4
This skill
profile.nsys-rep
Write
Overwrites existing standalone traces (irreversible)
server_profile.nsys-rep
Write
Overwrites existing server traces (irreversible)