llama-guard-3-8b-gptq-4bit

llama-guard-3-8b-gptq-4bit

Reproducible GPTQ 4-bit quantization of meta-llama/Llama-Guard-3-8B at revision 7327bd9f6efbbe6101dc6cc4736302b3cbb6e425.

Reproducibility

• Pipeline version: 2.0.0
• Quantizer: GPTQModel==7.1.0
• Source dtype: bfloat16
• Calibration data: 500 frozen C4 samples (250 English + 250 Portuguese)
• Calibration SHA-256: 65ea2fdb18470289314defc0fb94e27d4877b62143569615a1ce01745607ce1f
• C4 revision: 1588ec454efa1a09f29cd18ddd04fe05fc8653a2
• Random seed: 42

Quantization configuration

{
  "group_size": 128,
  "sym": true,
  "desc_act": false,
  "true_sequential": true,
  "lm_head": false,
  "pack_dtype": "int32",
  "format": "gptq",
  "damp_percent": 0.05,
  "damp_auto_increment": 0.01,
  "act_group_aware": true,
  "static_groups": false,
  "mse": 0.0,
  "hessian": {
    "chunk_size": null,
    "chunk_bytes": null,
    "staging_dtype": "float32"
  },
  "bits": 4
}

Calibration call

{
  "batch_size": 4,
  "calibration_sort": "desc",
  "calibration_concat_size": null,
  "calibration_data_min_length": 10,
  "calibration_concat_separator": null
}

The complete runtime environment, resolved checkpoint metadata, and file hashes are stored in runtime_environment.json, quantization_manifest.json, and SHA256SUMS.

Versioning

The checkpoint that preceded this reproducible build is preserved under the v1-original tag. Consumers that require an immutable dependency should pin a tag or commit SHA instead of relying on main.