{
  "architecture": "llama",
  "artifact_id": "hermes3-llama31",
  "auxiliary_tokenizer": {
    "metadata": "tokenizer/metadata.json",
    "path": "tokenizer"
  },
  "comparison_policy": {
    "cached_vs_cache_free": {
      "atol": 1e-06,
      "rtol": 1e-05
    },
    "primary_value": "logits",
    "reload": {
      "atol": 1e-07,
      "rtol": 1e-06
    },
    "require_finite": true
  },
  "format": {
    "container": "huggingface-safetensors",
    "dtype": "float32",
    "layout": "single-file",
    "package_path": "package"
  },
  "packages": {
    "gguf_q4_k_m": {
      "metadata": "gguf/metadata.json",
      "model": "gguf/model-Q4_K_M.gguf",
      "reference": "gguf/reference.json"
    },
    "safetensors_f32": {
      "config": "package/config.json",
      "model": "package/model.safetensors",
      "sha256": "22d5f7094e33c1e50eb689b20f56094239f0df30a23b17ef508ce1a6b49c81dc",
      "size_bytes": 347488
    }
  },
  "reference_runtime": {
    "config_class": "LlamaConfig",
    "implementation": "huggingface-transformers",
    "model_class": "LlamaForCausalLM",
    "trust_remote_code": false,
    "version": "4.55.0"
  },
  "schema_version": 1,
  "semantic_features": {
    "explicit_head_dim": false,
    "non_null_rope_scaling": true,
    "query_width_matches_hidden_size": true
  },
  "source": {
    "config_sha256": "7e189a0da074da6a878d5a276336caf3d62c57138f461bc6239402639f1bc6f5",
    "repository": "NousResearch/Hermes-3-Llama-3.1-8B",
    "revision": "896ea440e5a9e6070e3d8a2774daf2b481ab425b"
  },
  "validation": "POSTCHECK_OK"
}