{
  "base": "S:\\Proiecte\\qwen-mash\\models\\qwen35-4b-mtp\\Qwen3.5-4B-Q8_0.gguf",
  "base_validation": {
    "architecture": "qwen35",
    "backbone_block_count": 32,
    "block_indices": [
      0,
      1,
      2,
      3,
      4,
      5,
      6,
      7,
      8,
      9,
      10,
      11,
      12,
      13,
      14,
      15,
      16,
      17,
      18,
      19,
      20,
      21,
      22,
      23,
      24,
      25,
      26,
      27,
      28,
      29,
      30,
      31,
      32
    ],
    "hidden_size": 2560,
    "metadata_block_count": 33,
    "mtp_present": true,
    "q8_tensor_count": 257,
    "required_q8_tensors": {
      "blk.0.attn_qkv.weight": 8,
      "blk.0.ffn_gate.weight": 8
    },
    "tensor_count": 441
  },
  "composition": "base tensors + per_layer_token_embd.weight only",
  "dry_run": false,
  "flash_validation": {
    "ignored_incompatible_ple_tensor_count": 6,
    "ignored_incompatible_ple_tensor_names": [
      "blk.1.ple_conv1d.weight",
      "blk.1.ple_key.weight",
      "blk.1.ple_norm_conv.weight",
      "blk.1.ple_norm_key.weight",
      "blk.1.ple_norm_query.weight",
      "blk.1.ple_value.weight"
    ],
    "metadata_source": "S:\\Proiecte\\qwen-mash\\models\\flash-next\\UD-Q3_K_XL\\Qwen3.8-Flash-Next-UD-Q3_K_XL-00001-of-00003.gguf",
    "table": {
      "ggml_type": 20,
      "name": "per_layer_token_embd.weight",
      "payload_bytes": 28800138240,
      "shape": [
        160,
        320001536
      ]
    },
    "table_source": "S:\\Proiecte\\qwen-mash\\models\\flash-next\\UD-Q3_K_XL\\Qwen3.8-Flash-Next-UD-Q3_K_XL-00002-of-00003.gguf",
    "tokenizer": {
      "base": {
        "bos_token_id": 248044,
        "eos_token_id": 248046,
        "vocab_size": 248320
      },
      "flash": {
        "bos_token_id": 248044,
        "eos_token_id": 248046,
        "ple_eos_token_id": 248044,
        "vocab_size": 248320
      }
    }
  },
  "metadata_added": [
    "qwen35.ple.layers",
    "qwen35.ple.ngram_size",
    "qwen35.ple.heads_per_ngram",
    "qwen35.ple.conv_kernel",
    "qwen35.ple.eos_token_id",
    "qwen35.embedding_length_per_layer_input",
    "qwen35.ple.layer_multipliers",
    "qwen35.ple.head_offsets",
    "qwen35.ple.head_vocab_sizes",
    "qwen35.ple.image_token_id",
    "qwen35.ngram.gate",
    "qwen35.ngram.layer",
    "qwen35.ngram.source_architecture"
  ],
  "output": "S:\\Proiecte\\qwen-mash\\models\\qwen35-4b-mtp\\Qwen3.5-4B-Q8_0-FlashNgram-MTP.gguf",
  "output_bytes": 33410719936,
  "output_bytes_estimate": 33410719936,
  "output_data_start": 10970304,
  "split": {
    "count": 3,
    "indices": [
      1,
      2,
      3
    ],
    "stem": "Qwen3.8-Flash-Next-UD-Q3_K_XL"
  }
}