{
  "architectures": [
    "MonkeyOCRv2ViTAEVisionTransformer"
  ],
  "auto_map": {
    "AutoConfig": "configuration_monkeyocrv2_vitae.MonkeyOCRv2ViTAEEncoderConfig",
    "AutoModel": "modeling_monkeyocrv2_vitae_vision.MonkeyOCRv2ViTAEVisionTransformer"
  },
  "model_type": "monkeyocrv2_vitae_encoder",
  "num_channels": 3,
  "patch_size": 32,
  "temporal_patch_size": 1,
  "stage_dims": [
    64,
    128,
    256,
    512
  ],
  "stage_depths": [
    2,
    2,
    8,
    2
  ],
  "stage_heads": [
    1,
    2,
    4,
    8
  ],
  "window_size": 7,
  "mlp_ratio": 4.0,
  "hidden_size": 1024,
  "rms_norm_eps": 1e-05,
  "use_bias": false,
  "attn_implementation": "sdpa",
  "init_merger_std": 0.02,
  "initializer_range": 0.02,
  "is_causal": false,
  "post_norm": true,
  "gradient_checkpointing": false,
  "nc_groups": [
    1,
    32,
    64,
    128
  ],
  "rc_groups": [
    1,
    16,
    32,
    64
  ],
  "prm_embed_dim": 64,
  "downsample_ratios": [
    4,
    2,
    2,
    2
  ],
  "kernel_sizes": [
    7,
    3,
    3,
    3
  ],
  "rc_tokens_type": [
    "window",
    "window",
    "transformer",
    "transformer"
  ],
  "nc_tokens_type": [
    "window",
    "window",
    "transformer",
    "transformer"
  ],
  "rc_heads": [
    1,
    1,
    2,
    4
  ],
  "rc_embed_dims": [
    64,
    64,
    128,
    256
  ]
}