1 2 3 4 5
_target_: text_recognizer.networks.convnext.ConvNext dim: 16 dim_mults: [2, 4, 8] depths: [3, 3, 6] downsampling_factors: [[2, 2], [2, 2], [2, 2]]