{
  "architectures": [
    "DashengModel"
  ],
  "encoder_kwargs": {
    "depth": 32,
    "embed_dim": 1280,
    "num_heads": 16,
    "patch_size": [
      64,
      4
    ],
    "patch_stride": [
      64,
      4
    ],
    "target_length": 1008
  },
  "loss": "BCELoss",
  "model_type": "dasheng",
  "name": "dasheng-0.6B",
  "torch_dtype": "float32",
  "transformers_version": "4.35.2"
}