{
  "schema": 1,
  "name": "lfm2-5-2-6b",
  "family": "llm",
  "source": "LiquidAI/LFM2.5-2.6B",
  "target": {
    "backend": "coreml",
    "min_os": "macOS15"
  },
  "arch": {
    "hidden": 2048,
    "q_heads": 32,
    "kv_heads": 8,
    "head_dim": 64,
    "layers": 30,
    "vocab": 128000,
    "eps": 1e-05
  },
  "rope": {
    "strategy": "standard_1d",
    "theta": 10000000.0
  },
  "layers_schedule": {
    "schedule": "hybrid",
    "attn_idx": [
      2,
      5,
      9,
      13,
      17,
      21,
      24,
      27
    ],
    "conv_idx": [
      0,
      1,
      3,
      4,
      6,
      7,
      8,
      10,
      11,
      12,
      14,
      15,
      16,
      18,
      19,
      20,
      22,
      23,
      25,
      26,
      28,
      29
    ],
    "conv_kernel": 3
  },
  "artifacts": {
    "graphs": {
      "chunk0": {
        "package": "lfm2_5_2_6b_decode_chunk0.mlpackage",
        "compiled": "lfm2_5_2_6b_decode_chunk0.mlmodelc",
        "seq": 1,
        "ctx": 512,
        "precision": "fp16",
        "compute_units": "cpuAndNeuralEngine",
        "stateful": false
      },
      "chunk1": {
        "package": "lfm2_5_2_6b_decode_chunk1.mlpackage",
        "compiled": "lfm2_5_2_6b_decode_chunk1.mlmodelc",
        "seq": 1,
        "ctx": 512,
        "precision": "fp16",
        "compute_units": "cpuAndNeuralEngine",
        "stateful": false
      },
      "chunk2": {
        "package": "lfm2_5_2_6b_decode_chunk2.mlpackage",
        "compiled": "lfm2_5_2_6b_decode_chunk2.mlmodelc",
        "seq": 1,
        "ctx": 512,
        "precision": "fp16",
        "compute_units": "cpuAndNeuralEngine",
        "stateful": false
      },
      "chunk3": {
        "package": "lfm2_5_2_6b_decode_chunk3.mlpackage",
        "compiled": "lfm2_5_2_6b_decode_chunk3.mlmodelc",
        "seq": 1,
        "ctx": 512,
        "precision": "fp16",
        "compute_units": "cpuAndNeuralEngine",
        "stateful": false
      },
      "lmhead": {
        "package": "lfm2_5_2_6b_lmhead.mlpackage",
        "compiled": "lfm2_5_2_6b_lmhead.mlmodelc",
        "seq": 1,
        "ctx": 512,
        "precision": "fp16",
        "compute_units": "cpuAndNeuralEngine",
        "stateful": false
      }
    },
    "host_weights": {
      "embed": "lfm2_5_2_6b_embed_f16.bin"
    },
    "tokenizer": "tokenizer.json"
  },
  "lmhead": {
    "slices": 8,
    "slice_width": 16000,
    "vocab": 128000,
    "argmax_in_graph": false
  },
  "host_params": {
    "max_new": 64,
    "mask_neg": -40000.0
  },
  "success": {
    "parity": "greedy_exact",
    "cosine_min": 0.999,
    "min_tok_per_s": 4.0
  },
  "chat": {
    "prefix": "<|startoftext|><|im_start|>user\n",
    "suffix": "<|im_end|>\n<|im_start|>assistant\n<think>"
  }
}