NVIDIA Profile a Model
Choose the evidence level before running:
- For one model or testcase, use
trtmc-bench. - For a checked-in release comparison, use
tools/perf_matrix.py.
The public benchmark boundary is public_task_call_wall. It excludes bundle
building, process startup, warmup, report generation, and bundle loading.
Current public tooling does not provide layer attribution; do not infer a
kernel or layer bottleneck from Task API wall time alone.
Before timing
Run the owning family correctness testcase on the exact bundle. Record the repository and checkpoint revisions, GPU and driver, runtime root, family DSO, operation, input, seed, warmups, iterations, and synchronization boundary.
For a focused run:
trtmc-bench run --model <model> --case <case> \
--bundle <bundle> --no-build \
--warmup 3 --iterations 10 --runtime-root <runtime-root> \
--output <result-dir>
Use the same bundle kind, hardware, input, seed, warmups, iterations, and runtime policy for before/after comparisons. If any differ, report the confounder instead of one causal speedup percentage.
For release evidence, first check and then run the exact entry:
python3 tools/perf_matrix.py check <suite> \
--environment <environment> --entry <entry>
python3 tools/perf_matrix.py run <suite> \
--environment <environment> --entry <entry>
Lead the report with correctness and evidence level. Include exact commands, p50 and other suite-owned statistics, measurement scope, limitations, and any target not run.