efficiency_report.json

{
  "repo_name": "EfficiencyLatency-ModelRepo",
  "policy_name": "efficiency_latency_v1",
  "selected_checkpoint": "step_700",
  "weighted_overall": 0.739,
  "latency_ms": 860,
  "efficiency_score": 0.859,
  "quality_floor": 0.730,
  "eligible_checkpoint_count": 5,
  "rejected_checkpoint_count": 5,
  "selection_reason": "Selected step_700 because it meets the quality floor and has the highest efficiency_score among quality-floor checkpoints (0.859); step_1000 has a higher weighted overall score (0.752) but a lower efficiency score (0.611) because of its 1230 ms latency.",
  "readback_expected_files": [
    "README.md",
    "audit/efficiency_screen.csv",
    "config.json",
    "efficiency_report.json",
    "figures/fig1.png",
    "figures/fig2.png",
    "figures/fig3.png",
    "pytorch_model.bin"
  ]
}