compression_manifest.json

{
  "calibration_kind": "fused_inputs_embeds",
  "calibration_samples": 256,
  "calibration_source": "outputs/monet-100k/calibration-fused-clean/fused_calibration.pt",
  "compressor_commit": "8cec0acc1931de6f8f73257151ab7007c14dbf4e",
  "compressor_repo": "/data/llm-compressor",
  "device": "cuda:0",
  "dtype": "torch.float32",
  "max_seq_length": 160,
  "method": "gptq-int4",
  "notes": "Re-run scripts/evaluate.py before trusting task quality.",
  "source_model": "outputs/monet-100k/reference-bf16-clean"
}