update

examples/nx_clean_unet/run.sh

@@ -12,10 +12,10 @@ sh run.sh --stage 3 --stop_stage 3 --system_version centos --file_folder_name fi
 --noise_dir "/data/tianxing/HuggingDatasets/nx_noise/data/noise" \
 --speech_dir "/data/tianxing/HuggingDatasets/aishell/data_aishell/wav/train"
 
-sh run.sh --stage 2 --stop_stage 2 --system_version centos --file_folder_name file_dir --final_model_name mpnet-aishell-20250224 \
+sh run.sh --stage 1 --stop_stage 2 --system_version centos --file_folder_name file_dir --final_model_name nx-clean-unet-aishell-20250228 \
 --noise_dir "/data/tianxing/HuggingDatasets/nx_noise/data/noise" \
 --speech_dir "/data/tianxing/HuggingDatasets/aishell/data_aishell/wav/train" \
---max_epochs 100 --max_count 10000
+--max_epochs 100
 
 
 sh run.sh --stage 1 --stop_stage 2 --system_version centos --file_folder_name file_dir --final_model_name mpnet-nx-speech-20250224 \

examples/nx_clean_unet/yaml/config.yaml

@@ -6,14 +6,14 @@ n_fft: 512
 win_size: 200
 hop_size: 80
 
-down_sampling_num_layers: 5
+down_sampling_num_layers: 6
 down_sampling_in_channels: 1
 down_sampling_hidden_channels: 64
 down_sampling_kernel_size: 4
 down_sampling_stride: 2
 
-tsfm_hidden_size: 512
-tsfm_attention_heads: 8
+tsfm_hidden_size: 2048
+tsfm_attention_heads: 12
 tsfm_num_blocks: 6
 tsfm_dropout_rate: 0.1
 tsfm_max_length: 1024

toolbox/torchaudio/models/nx_clean_unet/modeling_nx_clean_unet.py

@@ -329,4 +329,4 @@ def main():
 
 
 if __name__ == "__main__":
-    main()
+    main2()