config.json

{
  "vision_tower": "gatedeltanet_siglip_2cross_history",
  "pretrain_encoder": "/home/pubw/proj/encoder/playground/models/siglip2-base-patch16-224",
  "num_temporal_layers": 4,
  "temporal_layer_indices": [
    -1,
    -2,
    -3,
    -4
  ],
  "conv_kernel_size": 4,
  "state_expansion_factor": 1,
  "temporal_state_init_mode": "learn",
  "temporal_mlp_ratio": 2,
  "select_layer": -2,
  "state_select_layer": -2,
  "vision_output_hidden_states": true
}