_target_: text_recognizer.networks.vqvae.pixelcnn.Decoder
out_channels: 1 
hidden_dim: 8
channels_multipliers: [8, 2, 1]
dropout_rate: 0.25