Skip to content

Latest commit

 

History

History
27 lines (23 loc) · 573 Bytes

File metadata and controls

27 lines (23 loc) · 573 Bytes

Experiment Runbook

Purpose

Every meaningful run should leave behind structured evidence.

Required run sequence

  1. Validate repo
  2. Check backend health
  3. Run model benchmarks
  4. Score outputs
  5. Attempt promotion
  6. Run quantization-retention report
  7. Record experiment metadata

Minimum experiment metadata

  • model_name
  • adapter
  • backend endpoint
  • prompt_profile
  • benchmark input file
  • scored output file
  • promotion decision
  • quantization report path
  • notes

Rule

No checkpoint should be described as improved unless an experiment record exists.