mask_rcnn_vit_large_hrfpn_cae_1x_coco.yml

_BASE_: [
  './mask_rcnn_vit_base_hrfpn_cae_1x_coco.yml'
]

weights: output/mask_rcnn_vit_large_hrfpn_cae_1x_coco/model_final


depth: &depth 24
dim: &dim 1024
use_fused_allreduce_gradients: &use_checkpoint True

VisionTransformer:
  img_size: [800, 1344]
  embed_dim: *dim
  depth: *depth
  num_heads: 16
  drop_path_rate: 0.25
  out_indices: [7, 11, 15, 23]
  use_checkpoint: *use_checkpoint
  pretrained: https://bj.bcebos.com/v1/paddledet/models/pretrained/vit_large_cae_pretrained.pdparams

HRFPN:
  in_channels: [*dim, *dim, *dim, *dim]

OptimizerBuilder:
  optimizer:
    layer_decay: 0.9
    weight_decay: 0.02
    num_layers: *depth