model_name: "clean_unet" channels_input: 1 channels_output: 1 channels_h: 64 max_h: 768 encoder_n_layers: 8 kernel_size: 4 stride: 2 tsfm_n_layers: 5 tsfm_n_head: 8 tsfm_d_model: 512 tsfm_d_inner: 2048