update

README.md

@@ -1,5 +1,5 @@
 ---
-title: VM Sound Classification
+title: NX Denoise
 emoji: 🐢
 colorFrom: purple
 colorTo: blue

examples/spectrum_unet_irm_aishell/run.sh

@@ -12,7 +12,7 @@ sh run.sh --stage 1 --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 3 --system_version centos --file_folder_name file_dir \
+sh run.sh --stage 3 --stop_stage 3 --system_version centos --file_folder_name file_dir \
 --noise_dir "/data/tianxing/HuggingDatasets/nx_noise/data/noise" \
 --speech_dir "/data/tianxing/HuggingDatasets/aishell/data_aishell/wav/train"
 

examples/spectrum_unet_irm_aishell/step_2_train_model.py

@@ -39,7 +39,7 @@ def get_args():
     parser.add_argument("--max_epochs", default=100, type=int)
 
     parser.add_argument("--batch_size", default=64, type=int)
-    parser.add_argument("--learning_rate", default=1e-3, type=float)
+    parser.add_argument("--learning_rate", default=1e-4, type=float)
     parser.add_argument("--num_serialized_models_to_keep", default=10, type=int)
     parser.add_argument("--patience", default=5, type=int)
     parser.add_argument("--serialization_dir", default="serialization_dir", type=str)
@@ -303,7 +303,8 @@ def main():
             snr_loss = snr_mse_loss.forward(lsnr_prediction, snr_db_target)
             if torch.any(torch.isnan(snr_loss)) or torch.any(torch.isinf(snr_loss)):
                 raise AssertionError("nan or inf in snr_loss")
-            loss = irm_loss + 0.1 * snr_loss
+            # loss = irm_loss + 0.1 * snr_loss
+            loss = irm_loss + 0.05 * snr_loss
             # loss = irm_loss
 
             total_loss += loss.item()
@@ -345,7 +346,8 @@ def main():
                 if torch.max(lsnr_prediction) > 1 or torch.min(lsnr_prediction) < 0:
                     raise AssertionError(f"expected lsnr_prediction between 0 and 1.")
                 snr_loss = snr_mse_loss.forward(lsnr_prediction, snr_db_target)
-                loss = irm_loss + 0.1 * snr_loss
+                # loss = irm_loss + 0.1 * snr_loss
+                loss = irm_loss + 0.05 * snr_loss
                 # loss = irm_loss
 
                 total_loss += loss.item()