update

examples/nx_mpnet/run.sh

@@ -3,7 +3,7 @@
 : <<'END'
 
 
-sh run.sh --stage 1 --stop_stage 2 --system_version centos --file_folder_name file_dir --final_model_name nx-mpnet-aishell-20250224 \
+sh run.sh --stage 2 --stop_stage 2 --system_version centos --file_folder_name file_dir --final_model_name nx-mpnet-aishell-20250224 \
 --noise_dir "/data/tianxing/HuggingDatasets/nx_noise/data/noise" \
 --speech_dir "/data/tianxing/HuggingDatasets/aishell/data_aishell/wav/train" \
 --max_epochs 100

toolbox/torchaudio/models/nx_mpnet/configuration_nx_mpnet.py

@@ -11,8 +11,8 @@ class NXMPNetConfig(PretrainedConfig):
                  sample_rate: int = 8000,
                  segment_size: int = 16000,
                  n_fft: int = 512,
-                 win_length: int = 200,
-                 hop_length: int = 80,
+                 win_size: int = 200,
+                 hop_size: int = 80,
 
                  dense_num_blocks: int = 4,
                  dense_hidden_size: int = 64,
@@ -29,8 +29,8 @@ class NXMPNetConfig(PretrainedConfig):
                  tsfm_dropout_rate: float = 0.0,
                  tsfm_max_time_relative_position: int = 1024,
                  tsfm_max_freq_relative_position: int = 128,
-                 tsfm_chunk_size: int = 4,
-                 tsfm_num_left_chunks: int = 128,
+                 tsfm_chunk_size: int = 1,
+                 tsfm_num_left_chunks: int = 64,
                  tsfm_num_right_chunks: int = 2,
 
                  discriminator_dim: int = 32,
@@ -51,8 +51,8 @@ class NXMPNetConfig(PretrainedConfig):
         self.sample_rate = sample_rate
         self.segment_size = segment_size
         self.n_fft = n_fft
-        self.win_length = win_length
-        self.hop_length = hop_length
+        self.win_size = win_size
+        self.hop_size = hop_size
 
         self.dense_num_blocks = dense_num_blocks
         self.dense_hidden_size = dense_hidden_size