{ "act_layer": "nn.GELU", "architectures": [ "Dinov2VisionTransformer" ], "block_chunks": 0, "depth": 12, "drop_path_rate": 0.0, "drop_path_uniform": false, "dtype": "float32", "embed_dim": 768, "ffn_bias": true, "ffn_layer": "mlp", "img_size": 518, "in_chans": 3, "init_values": 1e-05, "interpolate_antialias": false, "interpolate_offset": 0.1, "mlp_ratio": 4.0, "num_heads": 12, "num_register_tokens": 4, "output_indices": [ 2, 5, 8, 11 ], "patch_size": 14, "proj_bias": true, "qkv_bias": true, "transformers_version": "4.57.3" }