Skip to main content

Validate & Benchmark

Use the smallest test that can actually prove the claim:

ClaimMinimum meaningful evidence
Parser/config behaviorFocused unit or contract test
Exact checkpoint buildsSuccessful build with exact model revision/config
Task correctnessModel-owned E2E input, oracle, comparator, and thresholds
Target compatibilityFull build/load/run on that target and software cohort
PerformanceFixed workload, warmup, repeated measurement, timing boundary, and quality gate

Start repository validation with ownership checks:

python3 tools/model_ci.py validate

Then run the exact E2E manifest or focused test named by the model change. A skipped GPU preflight is not a pass, and a documentation build is not model parity evidence.

For performance, record model/revision, bundle checksum/config, execution path, hardware/software cohort, complete command, input, warmup, measured iterations, metric boundary, result, and task-quality result.

Use Testing Reference and Benchmarking Reference for command details. The Validation and Benchmarking Tutorial is the course-style lab.