quantize_run.json

{
  "source": "vllm-sr/d3-flash@13dc1cc8ef23caefa765907e1d764751e5629f7d",
  "bits": 8,
  "group_size": 64,
  "method": "round-to-nearest (no calibration)",
  "quantized": "language model only",
  "tool": "scripts/decision_bakeoff/quantize_readout.py (Midium)"
}