{
  "schema_version": "1",
  "slug": "qwen3-4b-awq",
  "version": "0.1.0",
  "title": "Qwen3-4B · AWQ W4A16",
  "summary": "A pinned AWQ W4A16 recipe targeting compressed-tensors export and vLLM inference. Dataset selection remains an explicit planning blocker.",
  "recipe_digest": "d375e4452dc0b86bae059c6f7b5d38d8ae3d40f1c31216156169c07aa259194f",
  "recipe_file_sha256": "3b6ccf40e164c895e31463a1ab1933ba6a886e13356a52ff8770ea7f965f00bc",
  "recipe_path": "/registry/v1/recipes/qwen3-4b-awq/0.1.0/recipe.yaml",
  "validator": {
    "package": "lazybrick",
    "version": "0.0.1",
    "command": "fingerprint"
  },
  "lifecycle": {
    "status": "planned",
    "summary": "The execution recipe is designed, but immutable calibration and evaluation datasets have not been selected."
  },
  "validation": {
    "status": "blocked",
    "summary": "Planning fails closed until both dataset references are pinned.",
    "blockers": [
      {
        "path": "calibration.dataset",
        "reason_code": "mutable_reference",
        "summary": "Calibration dataset URI and revision are placeholders."
      },
      {
        "path": "evaluation.dataset",
        "reason_code": "mutable_reference",
        "summary": "Evaluation dataset URI and revision are placeholders."
      }
    ]
  },
  "artifact": {
    "status": "none",
    "summary": "No quantized artifact has been produced or published."
  },
  "evidence": {
    "status": "not_measured",
    "summary": "No accuracy, latency, throughput, memory, or hardware evidence bundle exists."
  },
  "model": {
    "id": "Qwen/Qwen3-4B",
    "family": "Qwen3",
    "revision": "1cfa9a7208912126459214e8b04321603b3df60c"
  },
  "method": {
    "id": "awq",
    "label": "AWQ",
    "weight_bits": 4,
    "activation_bits": 16,
    "group_size": 128
  },
  "runtime": {
    "export_format": "compressed-tensors",
    "inference": "vLLM"
  },
  "target": {
    "accelerator_family": "NVIDIA CUDA",
    "minimum_memory_gib": 40
  },
  "tags": ["Qwen3", "AWQ", "W4A16", "vLLM", "compressed-tensors"],
  "usage": {
    "cli": "lazybrick validate recipe.yaml",
    "python": "from lazybrick import load_recipe\n\nrecipe = load_recipe(\"recipe.yaml\")\nprint(recipe.digest)"
  }
}