_type: GroundingDINOConfig
decoder_cfg:
  _type: DecoderConfig
  layer_cfg:
    _type: DecoderLayerConfig
    attention_dropout: 0
    cls_emb_type: mm_grounding_dino
    dropout: 0
    emb_dim: 256
    ffn_dim: 2048
    num_heads: 8
    num_levels: 4
    num_points: 4
    use_fusion: true
    use_legacy_pos: false
  num_layers: 6
  share_bbox_head: false
  share_cls_head: false
  share_norm: false
  share_pos_head: false
encoder_cfg:
  _type: CrossScaleEncoderConfig
  layer_cfg:
    _type: CrossScaleEncoderLayerConfig
    attention_dropout: 0.0
    dropout: 0.0
    droppath: 0.0
    emb_dim: 256
    emb_dim_fusion: 256
    ffn_dim: 2048
    ffn_dim_txt: 1024
    num_heads: 8
    num_heads_fusion: 4
    num_heads_txt: 4
    num_levels: 4
    num_points: 4
    use_fusion: true
    use_txt_self_att: false
  num_layers: 6
img_backbone_cfg:
  _type: ImageBackboneConfig
  backbone_kwargs:
    out_features:
    - stage2
    - stage3
    - stage4
  cfg_str: microsoft/swin-tiny-patch4-window7-224
  emb_dim: 256
  num_extra_down_projs: 1
  num_extra_up_projs: 0
  provider: transformers_auto
  use_legacy_pos: false
  use_pretrained: false
img_mean:
- 0.485
- 0.456
- 0.406
img_std:
- 0.229
- 0.224
- 0.225
query_selector_cfg:
  _type: QuerySelectorConfig
  cls_emb_type: mm_grounding_dino
  emb_dim: 256
  num_queries: 900
txt_backbone_cfg:
  _type: TextBackboneConfig
  backbone_kwargs: {}
  cfg_str: bert-base-uncased
  emb_dim: 256
  provider: transformers_auto
  use_pooling: true
  use_pretrained: false