{
  "inspected_at": "2026-10-27T21:15:51.516392+00:00",
  "python": "3.14.4",
  "versions": {
    "numpy": "2.3.5",
    "torch": "2.13.0",
    "transformers": "5.12.1",
    "safetensors": "0.8.0",
    "sentencepiece": "0.2.1"
  },
  "llama_cpp_commit": "620d2225cc436ab0b001e3200e590e3b286ef9ab",
  "gguf_py_version": "0.19.0 (from pinned source)",
  "llama_cpp_version": "version: 0.7.0 (build 11669, commit 620d2225c)\nbuilt with AppleClang 21.0.0.21000334 for Darwin arm64",
  "quantizer_sha256": "291b594d3265eb3764f842abd5f4f1f7a7b4f5b4b39124ac475e8d645a284dc2",
  "metadata_overrides": {
    "general.basename": "Llama",
    "general.finetune": "Instruct",
    "general.source.url": "https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct/tree/0cb88a4f764b7a12671c53f0838cd831a0843b95"
  },
  "source": {
    "model": "meta-llama/Llama-3.2-3B-Instruct",
    "revision": "0cb88a4f764b7a12671c53f0838cd831a0843b95",
    "config": {
      "architectures": [
        "LlamaForCausalLM"
      ],
      "attention_bias": false,
      "attention_dropout": 0.0,
      "bos_token_id": 128000,
      "eos_token_id": [
        128001,
        128008,
        128009
      ],
      "head_dim": 128,
      "hidden_act": "silu",
      "hidden_size": 3072,
      "initializer_range": 0.02,
      "intermediate_size": 8192,
      "max_position_embeddings": 131072,
      "mlp_bias": false,
      "model_type": "llama",
      "num_attention_heads": 24,
      "num_hidden_layers": 28,
      "num_key_value_heads": 8,
      "pretraining_tp": 1,
      "rms_norm_eps": 1e-05,
      "rope_scaling": {
        "factor": 32.0,
        "high_freq_factor": 4.0,
        "low_freq_factor": 1.0,
        "original_max_position_embeddings": 8192,
        "rope_type": "llama3"
      },
      "rope_theta": 500000.0,
      "tie_word_embeddings": true,
      "torch_dtype": "bfloat16",
      "transformers_version": "4.45.0.dev0",
      "use_cache": true,
      "vocab_size": 128256
    },
    "shards": [
      {
        "file": "model-00001-of-00002.safetensors",
        "file_bytes": 4965799096,
        "sha256": "13cbd6d16e927a0c5bad54102514e6e18b4a47b3a6eb911e39d678d328d19f55",
        "header_bytes": 21680,
        "data_start": 21688,
        "data_bytes": 4965777408,
        "tensor_count": 187,
        "parameter_count": 2482888704,
        "metadata": {
          "format": "pt"
        }
      },
      {
        "file": "model-00002-of-00002.safetensors",
        "file_bytes": 1459729952,
        "sha256": "7b770216613ac5c34d7c54bdff1fa616bc4e338a9d0b20af6303e48c295ee23c",
        "header_bytes": 7704,
        "data_start": 7712,
        "data_bytes": 1459722240,
        "tensor_count": 67,
        "parameter_count": 729861120,
        "metadata": {
          "format": "pt"
        }
      }
    ],
    "tensor_count": 254,
    "parameter_count": 3212749824,
    "file_bytes": 6425529048,
    "data_bytes": 6425499648,
    "dtype_counts": {
      "BF16": 254
    },
    "operational_files": [
      {
        "file": "config.json",
        "file_bytes": 878,
        "sha256": "39fb36dc5416f445ebc4e71cb71fbcf6727e80a35836d8ba1a1474c318467b7a"
      },
      {
        "file": "generation_config.json",
        "file_bytes": 189,
        "sha256": "88effbb63300dbbc7390143fbbdd9d9fa50587b37e8bfd16c8c90d4970a74a36"
      },
      {
        "file": "model.safetensors.index.json",
        "file_bytes": 20919,
        "sha256": "0a57c30afc07c81610ae7fe013bea3b9d50365f9a8145ae98e8714220f3d7c7c"
      },
      {
        "file": "special_tokens_map.json",
        "file_bytes": 296,
        "sha256": "6f38c73729248f6c127296386e3cdde96e254636cc58b4169d3fd32328d9a8ec"
      },
      {
        "file": "tokenizer.json",
        "file_bytes": 9085657,
        "sha256": "79e3e522635f3171300913bb421464a87de6222182a0570b9b2ccba2a964b2b4"
      },
      {
        "file": "tokenizer_config.json",
        "file_bytes": 54528,
        "sha256": "9823dcfdc1121869029da45192238e85cf44f0b232a6d9dc20e4fe6f4242a14e"
      }
    ],
    "examples": {
      "model.embed_tokens.weight": {
        "dtype": "BF16",
        "shape": [
          128256,
          3072
        ],
        "data_offsets": [
          0,
          788004864
        ],
        "file": "model-00001-of-00002.safetensors",
        "data_start": 21688
      },
      "model.layers.0.input_layernorm.weight": {
        "dtype": "BF16",
        "shape": [
          3072
        ],
        "data_offsets": [
          788004864,
          788011008
        ],
        "file": "model-00001-of-00002.safetensors",
        "data_start": 21688
      },
      "model.layers.0.self_attn.q_proj.weight": {
        "dtype": "BF16",
        "shape": [
          3072,
          3072
        ],
        "data_offsets": [
          964177920,
          983052288
        ],
        "file": "model-00001-of-00002.safetensors",
        "data_start": 21688
      },
      "model.layers.0.self_attn.k_proj.weight": {
        "dtype": "BF16",
        "shape": [
          1024,
          3072
        ],
        "data_offsets": [
          939012096,
          945303552
        ],
        "file": "model-00001-of-00002.safetensors",
        "data_start": 21688
      },
      "model.layers.0.mlp.gate_proj.weight": {
        "dtype": "BF16",
        "shape": [
          8192,
          3072
        ],
        "data_offsets": [
          838342656,
          888674304
        ],
        "file": "model-00001-of-00002.safetensors",
        "data_start": 21688
      }
    },
    "lm_head_stored": false
  },
  "gguf": {
    "bf16": {
      "file": "Llama-3.2-3B-Instruct-BF16.gguf",
      "sha256": "4eb637e2e395dae8ac2e2cc85f559207684fd1dc8c76185369e5a0d72bba3445",
      "file_bytes": 6433688160,
      "version": 3,
      "tensor_count": 255,
      "stored_value_count": 3212749888,
      "metadata_count": 37,
      "metadata": {
        "general.architecture": "llama",
        "general.type": "model",
        "general.sampling.top_p": 0.8999999761581421,
        "general.sampling.temp": 0.6000000238418579,
        "general.name": "Llama 3.2 3B Instruct",
        "general.finetune": "Instruct",
        "general.basename": "Llama",
        "general.size_label": "3.2B",
        "general.license": "llama3.2",
        "general.source.url": "https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct/tree/0cb88a4f764b7a12671c53f0838cd831a0843b95",
        "llama.block_count": 28,
        "llama.context_length": 131072,
        "llama.embedding_length": 3072,
        "llama.feed_forward_length": 8192,
        "llama.attention.head_count": 24,
        "llama.attention.head_count_kv": 8,
        "llama.rope.freq_base": 500000.0,
        "llama.attention.layer_norm_rms_epsilon": 9.999999747378752e-06,
        "llama.attention.key_length": 128,
        "llama.attention.value_length": 128,
        "general.file_type": 32,
        "llama.vocab_size": 128256,
        "llama.rope.dimension_count": 128,
        "general.quantization_version": 2,
        "tokenizer.ggml.model": "gpt2",
        "tokenizer.ggml.pre": "llama-bpe",
        "tokenizer.ggml.bos_token_id": 128000,
        "tokenizer.ggml.eos_token_id": 128009,
        "tokenizer.ggml.add_bos_token": true,
        "tokenizer.ggml.add_eos_token": false,
        "tokenizer.ggml.add_sep_token": false
      },
      "metadata_end": 7822906,
      "tensor_table_bytes": 15130,
      "header_padding_bytes": 12,
      "data_start": 7838048,
      "alignment": 32,
      "tokenizer_metadata_bytes": 7821550,
      "token_count": 128256,
      "merge_count": 280147,
      "chat_template_present": true,
      "tensor_bytes": 6425850112,
      "padding_bytes": 0,
      "dtype_breakdown": {
        "F32": {
          "tensors": 58,
          "values": 175168,
          "bytes": 700672
        },
        "BF16": {
          "tensors": 197,
          "values": 3212574720,
          "bytes": 6425149440
        }
      },
      "examples": {
        "rope_freqs.weight": {
          "name": "rope_freqs.weight",
          "shape": [
            64
          ],
          "type": "F32",
          "values": 64,
          "bytes": 256,
          "relative_offset": 0,
          "absolute_offset": 7838048
        },
        "token_embd.weight": {
          "name": "token_embd.weight",
          "shape": [
            3072,
            128256
          ],
          "type": "BF16",
          "values": 394002432,
          "bytes": 788004864,
          "relative_offset": 256,
          "absolute_offset": 7838304
        },
        "blk.0.attn_norm.weight": {
          "name": "blk.0.attn_norm.weight",
          "shape": [
            3072
          ],
          "type": "F32",
          "values": 3072,
          "bytes": 12288,
          "relative_offset": 788005120,
          "absolute_offset": 795843168
        },
        "blk.0.ffn_gate.weight": {
          "name": "blk.0.ffn_gate.weight",
          "shape": [
            3072,
            8192
          ],
          "type": "BF16",
          "values": 25165824,
          "bytes": 50331648,
          "relative_offset": 838349056,
          "absolute_offset": 846187104
        },
        "blk.0.attn_k.weight": {
          "name": "blk.0.attn_k.weight",
          "shape": [
            3072,
            1024
          ],
          "type": "BF16",
          "values": 3145728,
          "bytes": 6291456,
          "relative_offset": 939024640,
          "absolute_offset": 946862688
        },
        "blk.0.attn_q.weight": {
          "name": "blk.0.attn_q.weight",
          "shape": [
            3072,
            3072
          ],
          "type": "BF16",
          "values": 9437184,
          "bytes": 18874368,
          "relative_offset": 964190464,
          "absolute_offset": 972028512
        }
      },
      "tensor_bits_per_source_parameter": 16.00087268295186,
      "file_bits_per_source_parameter": 16.02039003955759,
      "inventory_file": "bf16-tensors.json"
    },
    "q4_0": {
      "file": "Llama-3.2-3B-Instruct-Q4_0.gguf",
      "sha256": "f961b04ff043fe953113571eda531696d84c338b1c38364ac5e8ef7a37d13c6b",
      "file_bytes": 1917190752,
      "version": 3,
      "tensor_count": 255,
      "stored_value_count": 3212749888,
      "metadata_count": 37,
      "metadata": {
        "general.architecture": "llama",
        "general.type": "model",
        "general.sampling.top_p": 0.8999999761581421,
        "general.sampling.temp": 0.6000000238418579,
        "general.name": "Llama 3.2 3B Instruct",
        "general.finetune": "Instruct",
        "general.basename": "Llama",
        "general.size_label": "3.2B",
        "general.license": "llama3.2",
        "general.source.url": "https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct/tree/0cb88a4f764b7a12671c53f0838cd831a0843b95",
        "llama.block_count": 28,
        "llama.context_length": 131072,
        "llama.embedding_length": 3072,
        "llama.feed_forward_length": 8192,
        "llama.attention.head_count": 24,
        "llama.attention.head_count_kv": 8,
        "llama.rope.freq_base": 500000.0,
        "llama.attention.layer_norm_rms_epsilon": 9.999999747378752e-06,
        "llama.attention.key_length": 128,
        "llama.attention.value_length": 128,
        "llama.vocab_size": 128256,
        "llama.rope.dimension_count": 128,
        "tokenizer.ggml.model": "gpt2",
        "tokenizer.ggml.pre": "llama-bpe",
        "tokenizer.ggml.bos_token_id": 128000,
        "tokenizer.ggml.eos_token_id": 128009,
        "tokenizer.ggml.add_bos_token": true,
        "tokenizer.ggml.add_eos_token": false,
        "tokenizer.ggml.add_sep_token": false,
        "general.quantization_version": 2,
        "general.file_type": 2
      },
      "metadata_end": 7822906,
      "tensor_table_bytes": 15130,
      "header_padding_bytes": 12,
      "data_start": 7838048,
      "alignment": 32,
      "tokenizer_metadata_bytes": 7821550,
      "token_count": 128256,
      "merge_count": 280147,
      "chat_template_present": true,
      "tensor_bytes": 1909352704,
      "padding_bytes": 0,
      "dtype_breakdown": {
        "F32": {
          "tensors": 58,
          "values": 175168,
          "bytes": 700672
        },
        "Q6_K": {
          "tensors": 1,
          "values": 394002432,
          "bytes": 323205120
        },
        "Q4_0": {
          "tensors": 196,
          "values": 2818572288,
          "bytes": 1585446912
        }
      },
      "examples": {
        "rope_freqs.weight": {
          "name": "rope_freqs.weight",
          "shape": [
            64
          ],
          "type": "F32",
          "values": 64,
          "bytes": 256,
          "relative_offset": 12288,
          "absolute_offset": 7850336
        },
        "token_embd.weight": {
          "name": "token_embd.weight",
          "shape": [
            3072,
            128256
          ],
          "type": "Q6_K",
          "values": 394002432,
          "bytes": 323205120,
          "relative_offset": 12544,
          "absolute_offset": 7850592
        },
        "blk.0.attn_k.weight": {
          "name": "blk.0.attn_k.weight",
          "shape": [
            3072,
            1024
          ],
          "type": "Q4_0",
          "values": 3145728,
          "bytes": 1769472,
          "relative_offset": 323217664,
          "absolute_offset": 331055712
        },
        "blk.0.attn_norm.weight": {
          "name": "blk.0.attn_norm.weight",
          "shape": [
            3072
          ],
          "type": "F32",
          "values": 3072,
          "bytes": 12288,
          "relative_offset": 324987136,
          "absolute_offset": 332825184
        },
        "blk.0.attn_q.weight": {
          "name": "blk.0.attn_q.weight",
          "shape": [
            3072,
            3072
          ],
          "type": "Q4_0",
          "values": 9437184,
          "bytes": 5308416,
          "relative_offset": 330307840,
          "absolute_offset": 338145888
        },
        "blk.0.ffn_gate.weight": {
          "name": "blk.0.ffn_gate.weight",
          "shape": [
            3072,
            8192
          ],
          "type": "Q4_0",
          "values": 25165824,
          "bytes": 14155776,
          "relative_offset": 351541504,
          "absolute_offset": 359379552
        }
      },
      "tensor_bits_per_source_parameter": 4.754438555375048,
      "file_bits_per_source_parameter": 4.773955911980777,
      "inventory_file": "q4_0-tensors.json"
    },
    "q4_0_embeddings_q8_0": {
      "file": "Llama-3.2-3B-Instruct-Q4_0-embeddings-Q8_0.gguf",
      "sha256": "03f0f2ab2452e42873084ecc63113caad9e1ed1359097f9c25f776a2a5d74fcc",
      "file_bytes": 2012613216,
      "version": 3,
      "tensor_count": 255,
      "stored_value_count": 3212749888,
      "metadata_count": 37,
      "metadata": {
        "general.architecture": "llama",
        "general.type": "model",
        "general.sampling.top_p": 0.8999999761581421,
        "general.sampling.temp": 0.6000000238418579,
        "general.name": "Llama 3.2 3B Instruct",
        "general.finetune": "Instruct",
        "general.basename": "Llama",
        "general.size_label": "3.2B",
        "general.license": "llama3.2",
        "general.source.url": "https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct/tree/0cb88a4f764b7a12671c53f0838cd831a0843b95",
        "llama.block_count": 28,
        "llama.context_length": 131072,
        "llama.embedding_length": 3072,
        "llama.feed_forward_length": 8192,
        "llama.attention.head_count": 24,
        "llama.attention.head_count_kv": 8,
        "llama.rope.freq_base": 500000.0,
        "llama.attention.layer_norm_rms_epsilon": 9.999999747378752e-06,
        "llama.attention.key_length": 128,
        "llama.attention.value_length": 128,
        "llama.vocab_size": 128256,
        "llama.rope.dimension_count": 128,
        "tokenizer.ggml.model": "gpt2",
        "tokenizer.ggml.pre": "llama-bpe",
        "tokenizer.ggml.bos_token_id": 128000,
        "tokenizer.ggml.eos_token_id": 128009,
        "tokenizer.ggml.add_bos_token": true,
        "tokenizer.ggml.add_eos_token": false,
        "tokenizer.ggml.add_sep_token": false,
        "general.quantization_version": 2,
        "general.file_type": 2
      },
      "metadata_end": 7822906,
      "tensor_table_bytes": 15130,
      "header_padding_bytes": 12,
      "data_start": 7838048,
      "alignment": 32,
      "tokenizer_metadata_bytes": 7821550,
      "token_count": 128256,
      "merge_count": 280147,
      "chat_template_present": true,
      "tensor_bytes": 2004775168,
      "padding_bytes": 0,
      "dtype_breakdown": {
        "F32": {
          "tensors": 58,
          "values": 175168,
          "bytes": 700672
        },
        "Q8_0": {
          "tensors": 1,
          "values": 394002432,
          "bytes": 418627584
        },
        "Q4_0": {
          "tensors": 196,
          "values": 2818572288,
          "bytes": 1585446912
        }
      },
      "examples": {
        "rope_freqs.weight": {
          "name": "rope_freqs.weight",
          "shape": [
            64
          ],
          "type": "F32",
          "values": 64,
          "bytes": 256,
          "relative_offset": 12288,
          "absolute_offset": 7850336
        },
        "token_embd.weight": {
          "name": "token_embd.weight",
          "shape": [
            3072,
            128256
          ],
          "type": "Q8_0",
          "values": 394002432,
          "bytes": 418627584,
          "relative_offset": 12544,
          "absolute_offset": 7850592
        },
        "blk.0.attn_k.weight": {
          "name": "blk.0.attn_k.weight",
          "shape": [
            3072,
            1024
          ],
          "type": "Q4_0",
          "values": 3145728,
          "bytes": 1769472,
          "relative_offset": 418640128,
          "absolute_offset": 426478176
        },
        "blk.0.attn_norm.weight": {
          "name": "blk.0.attn_norm.weight",
          "shape": [
            3072
          ],
          "type": "F32",
          "values": 3072,
          "bytes": 12288,
          "relative_offset": 420409600,
          "absolute_offset": 428247648
        },
        "blk.0.attn_q.weight": {
          "name": "blk.0.attn_q.weight",
          "shape": [
            3072,
            3072
          ],
          "type": "Q4_0",
          "values": 9437184,
          "bytes": 5308416,
          "relative_offset": 425730304,
          "absolute_offset": 433568352
        },
        "blk.0.ffn_gate.weight": {
          "name": "blk.0.ffn_gate.weight",
          "shape": [
            3072,
            8192
          ],
          "type": "Q4_0",
          "values": 25165824,
          "bytes": 14155776,
          "relative_offset": 446963968,
          "absolute_offset": 454802016
        }
      },
      "tensor_bits_per_source_parameter": 4.992048003299494,
      "file_bits_per_source_parameter": 5.011565359905222,
      "inventory_file": "q4_0_embeddings_q8_0-tensors.json"
    }
  },
  "conversion_verification": {
    "source_tensors_checked": 254,
    "all_exact_after_documented_transforms": true,
    "generated_tensor": {
      "name": "rope_freqs.weight",
      "values": [
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.234937071800232,
        1.6513292789459229,
        2.2766594886779785,
        3.2922627925872803,
        5.171597003936768,
        9.666730880737305,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0,
        32.0
      ],
      "exact_match": true
    },
    "transform_counts": {
      "identity": 141,
      "identity; BF16 to F32": 57,
      "RoPE row permutation": 56
    }
  },
  "q4_0_block": {
    "tensor": "blk.0.ffn_gate.weight",
    "block_index": 0,
    "file_offset": 359379552,
    "hex_bytes": "e0 1b e7 07 5d 55 8d d7 55 e9 f4 93 5c db bc 77 b3 60",
    "scale": 0.00384521484375,
    "codes": [
      7,
      7,
      13,
      5,
      13,
      7,
      5,
      9,
      4,
      3,
      12,
      11,
      12,
      7,
      3,
      0,
      14,
      0,
      5,
      5,
      8,
      13,
      5,
      14,
      15,
      9,
      5,
      13,
      11,
      7,
      11,
      6
    ],
    "original_bf16_values": [
      -0.002960205078125,
      -0.004791259765625,
      0.0198974609375,
      -0.0108642578125,
      0.020751953125,
      -0.00250244140625,
      -0.01318359375,
      0.0047607421875,
      -0.01495361328125,
      -0.02001953125,
      0.015869140625,
      0.0133056640625,
      0.0137939453125,
      -0.0035552978515625,
      -0.018798828125,
      -0.03076171875,
      0.0233154296875,
      -0.029296875,
      -0.01171875,
      -0.01171875,
      4.124641418457031e-05,
      0.01953125,
      -0.01220703125,
      0.0230712890625,
      0.02978515625,
      0.002166748046875,
      -0.01080322265625,
      0.017333984375,
      0.0103759765625,
      -0.002166748046875,
      0.01153564453125,
      -0.0093994140625
    ],
    "reconstructed_values": [
      -0.00384521484375,
      -0.00384521484375,
      0.01922607421875,
      -0.01153564453125,
      0.01922607421875,
      -0.00384521484375,
      -0.01153564453125,
      0.00384521484375,
      -0.015380859375,
      -0.01922607421875,
      0.015380859375,
      0.01153564453125,
      0.015380859375,
      -0.00384521484375,
      -0.01922607421875,
      -0.03076171875,
      0.0230712890625,
      -0.03076171875,
      -0.01153564453125,
      -0.01153564453125,
      0.0,
      0.01922607421875,
      -0.01153564453125,
      0.0230712890625,
      0.02691650390625,
      0.00384521484375,
      -0.01153564453125,
      0.01922607421875,
      0.01153564453125,
      -0.00384521484375,
      0.01153564453125,
      -0.0076904296875
    ],
    "mean_absolute_error": 0.0009125247597694397,
    "max_absolute_error": 0.00286865234375,
    "manual_decode_matches_gguf": true
  },
  "embedding_override_changes_only_embedding_tensor": true,
  "commands": [
    "python convert_hf_to_gguf.py MODEL_DIR --outtype bf16 --model-name 'Llama 3.2 3B Instruct' --metadata metadata.json --outfile WORK_DIR/Llama-3.2-3B-Instruct-BF16.gguf",
    "llama-quantize WORK_DIR/Llama-3.2-3B-Instruct-BF16.gguf WORK_DIR/Llama-3.2-3B-Instruct-Q4_0.gguf Q4_0 8",
    "llama-quantize --token-embedding-type q8_0 WORK_DIR/Llama-3.2-3B-Instruct-BF16.gguf WORK_DIR/Llama-3.2-3B-Instruct-Q4_0-embeddings-Q8_0.gguf Q4_0 8"
  ],
  "limits": "File inspection and conversion checks only. One quantized block does not measure model quality, and no inference performance claim is made."
}