{
  "architecture" : {
    "attentionLayers" : 28,
    "class" : "dense",
    "totalLayers" : 28
  },
  "capabilities" : [
    "chat",
    "tools"
  ],
  "driver" : "causal-lm",
  "engine" : {
    "allowed" : [
      "pipelined",
      "sequential"
    ],
    "expectFrequentReshapes" : true,
    "kind" : "pipelined"
  },
  "functions" : {
    "main" : {
      "role" : "step"
    }
  },
  "inputs" : {
    "input_ids" : {
      "bind" : "tokens",
      "dtype" : "int32"
    },
    "position_ids" : {
      "bind" : "positions",
      "dtype" : "int32",
      "mode" : "running"
    }
  },
  "limits" : {
    "contextMax" : {
      "macos" : 4096
    },
    "maxOutputDefault" : 1024,
    "prefillChunk" : 512,
    "prefillChunkThreshold" : 1024
  },
  "memory" : {
    "bytesPerToken" : 1545688993,
    "kvBytesPerToken" : 28672,
    "requiresIncreasedMemoryLimit" : false,
    "weightsBytes" : 1545688993
  },
  "outputs" : {
    "logits" : {
      "dtype" : "float16",
      "role" : "logits",
      "seq" : "dynamic"
    }
  },
  "provenance" : {
    "exportedOn" : {
      "date" : "2026-09-23",
      "os" : "26A428"
    },
    "exporter" : {
      "coreai-core" : "1.0.0b2",
      "coreai-opt" : "0.2.1",
      "coreai-torch" : "0.4.2",
      "torch" : "2.11.0"
    },
    "files" : {
      "metadata.json" : "sha256:afa5f0ff71d348b30ce8f6b0e6bb19850a89b7f75b4745e512ee7a82d2f7a908",
      "qwen2_5_coder_1_5b.aimodel/main.hash" : "sha256:9b8cd7a47be7784d8cd5867c0373772d99b651274cfe646808631020d6c40388",
      "qwen2_5_coder_1_5b.aimodel/main.mlirb" : "sha256:132cc44a656d8404bea0a1e849db1ae7f755208a06c2d5bbdbc5c1bcac023c70",
      "qwen2_5_coder_1_5b.aimodel/metadata.json" : "sha256:5a45101a3ba4551db484c7e94c31bb4d5cdf89ea7f0fde3e06dd9b854e89c502",
      "tokenizer/generation_config.json" : "sha256:1a628a5775bc69cde01c6749a531150ca4d3189652c618a174f7077923acf3b1",
      "tokenizer/merges.txt" : "sha256:599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3",
      "tokenizer/tokenizer.json" : "sha256:c0382117ea329cdf097041132f6d735924b697924d6f6fc3945713e96ce87539",
      "tokenizer/tokenizer_config.json" : "sha256:959e7f1d9a1b7641a6d6ce05ca97b75c7894fcb66cbe5a040406458fb1128ee4",
      "tokenizer/vocab.json" : "sha256:ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910"
    },
    "forge" : "0.1.0",
    "recipe" : "qwen2.5-coder-1.5b@1",
    "source" : {
      "hfRepo" : "Qwen/Qwen2.5-Coder-1.5B-Instruct",
      "revision" : "2e1fd397ee46e1388853d2af2c993145b0f1098a"
    }
  },
  "quantization" : {
    "notes" : "weight-only PTQ on the torch module (coreai-opt eager Quantizer); exceptions resolve first-match in declaration order, then default; modules: int8-linear-perchannel ×141",
    "scheme" : "int8-linear-perchannel",
    "weightBits" : 8
  },
  "schema" : 1,
  "states" : {
    "keyCache" : {
      "axis" : 3,
      "initial" : 256,
      "kind" : "kvGrowing"
    },
    "valueCache" : {
      "axis" : 3,
      "initial" : 256,
      "kind" : "kvGrowing"
    }
  },
  "target" : {
    "arch" : "any",
    "backend" : "coreai",
    "compiled" : false,
    "compute" : "gpu",
    "minOS" : "27.0",
    "platform" : "macos"
  },
  "text" : {
    "chatTemplate" : "embedded",
    "tokenizer" : "tokenizer/",
    "toolDialect" : "hermes"
  }
}