{
  "python": "3.12.13",
  "torch": "2.15.0",
  "transformers": "5.17.0",
  "numpy": "2.5.4",
  "model": "Qwen/Qwen2.5-0.5B-Instruct",
  "revision": "7ae557604adf67be50417f59c2c2f167def9a775",
  "packing": {
    "original_bytes": 4194304,
    "packed_bytes": 524288,
    "scale_bytes": 65536,
    "persistent_bytes": 589824,
    "dequantized_bytes": 4194304
  },
  "cache": [
    {
      "batch": 1,
      "tokens": 128,
      "tensor_bytes": 1572864,
      "first_key_shape": [
        1,
        2,
        128,
        64
      ],
      "dtype": "torch.float16"
    },
    {
      "batch": 1,
      "tokens": 512,
      "tensor_bytes": 6291456,
      "first_key_shape": [
        1,
        2,
        512,
        64
      ],
      "dtype": "torch.float16"
    },
    {
      "batch": 2,
      "tokens": 512,
      "tensor_bytes": 12582912,
      "first_key_shape": [
        2,
        2,
        512,
        64
      ],
      "dtype": "torch.float16"
    }
  ],
  "config": {
    "layers": 24,
    "kv_heads": 2,
    "head_dim": 64,
    "fp16_bytes_per_token": 12288
  },
  "projection_only": {
    "batch": 8,
    "tokens": 32768,
    "fp16_cache_bytes": 3221225472
  }
}