throughput.json

{
  "modality": "text",
  "model": "sentence-transformers/all-MiniLM-L6-v2",
  "device": "cuda",
  "gpu": "NVIDIA GeForce RTX 5090",
  "embedding_dim": 384,
  "autobatch": {
    "min_batch": 8,
    "max_batch": 512,
    "start_batch": 32,
    "grow_free_fraction": 0.5,
    "shrink_free_fraction": 0.2,
    "grow_factor": 1.5,
    "shrink_factor": 0.5,
    "oom_factor": 0.5
  },
  "runs": [
    {
      "num_workers": 1,
      "pre_run_counts": {
        "pending": 500,
        "in_progress": 0,
        "done": 0,
        "failed": 0,
        "requeued": 0
      },
      "total_processed": 500,
      "wall_seconds": 14.76,
      "throughput": 33.9,
      "note": "small-scale single-worker RTX 5090 run"
    }
  ]
}