config.json

{
  "_name_or_path": "runs/train/stage35_BS18/model/vision_tower",
  "architectures": [
    "PS3VisionModel"
  ],
  "class_token": null,
  "drop": 0.0,
  "drop_path": null,
  "dynamic_img_size": true,
  "final_norm": false,
  "hidden_size": 1152,
  "highres_selection_feature": true,
  "highres_selection_module_depth": 3,
  "highres_selection_module_hidden_dim": 512,
  "highres_selection_module_kernel_size": 28,
  "highres_selection_module_out_dim": 512,
  "image_size": 3780,
  "img_size": null,
  "max_select_num": 2560,
  "max_select_num_each_scale": null,
  "min_select_num": 1,
  "model_name": "vit_so400m_patch14_siglip_378",
  "model_type": "ps3_vision_model",
  "patch_drop": null,
  "patch_size": 14,
  "pool": "map",
  "pretrained": false,
  "ps3": true,
  "ps3_scales": [
    378,
    756,
    1512,
    3780
  ],
  "radio": false,
  "radio_adapter_mlp_hidden_dim": null,
  "radio_adapter_mlp_input_dim": null,
  "radio_adapter_mlp_num_inner": null,
  "radio_adapter_mlp_output_dim": null,
  "radio_adapter_mlp_version": null,
  "select_based_on_layer": [
    0,
    9,
    18,
    26
  ],
  "separate_pos_emb": true,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.46.0",
  "vision_tower_name": "vit_so400m_patch14_siglip_378"
}