depthformer_base_encoder_encoder_meta.json

{
  "model_type": "depthformer_base_encoder",
  "inputs": [
    {
      "name": "context_tokens",
      "shape": [
        1,
        1006
      ],
      "dtype": "int32"
    },
    {
      "name": "style_tokens",
      "shape": [
        1,
        6
      ],
      "dtype": "int32"
    }
  ],
  "output_name": "encoder_hidden_states",
  "output_shape": [
    1,
    1006,
    768
  ],
  "model_config": {
    "embed_dim": 768,
    "num_heads": 12,
    "num_encoder_layers": 12,
    "mlp_dim": 2048
  },
  "token_config": {
    "context_length": 1006,
    "style_rvq_depth": 6,
    "rvq_codebook_size": 1024
  },
  "opset_version": 18,
  "ir_version": 8,
  "precision": "fp16"
}