{ "input_size": [ 224, 224 ], "in_chans": 3, "embed_dim": 96, "num_heads": 1, "num_classes": 1000, "stages": [ 1, 2, 11, 2 ], "q_pool": 3, "q_stride": [ 2, 2 ], "mask_unit_size": [ 8, 8 ], "mask_unit_attn": [ true, true, false, false ], "dim_mul": 2.0, "head_mul": 2.0, "patch_kernel": [ 7, 7 ], "patch_stride": [ 4, 4 ], "patch_padding": [ 3, 3 ], "mlp_ratio": 4.0, "drop_path_rate": 0.0, "norm_layer": "LayerNorm", "head_dropout": 0.0, "head_init_scale": 0.001, "sep_pos_embed": false }