update

examples/vm_sound_classification/conv2d_classifier.yaml

@@ -18,21 +18,21 @@ spec_augment_param:
 conv2d_block_param_list:
   - batch_norm: true
     in_channels: 1
-    out_channels: 32
+    out_channels: 16
     kernel_size: 3
     stride: 1
     dilation: 3
     activation: relu
     dropout: 0.1
-  - in_channels: 32
-    out_channels: 32
+  - in_channels: 16
+    out_channels: 16
     kernel_size: 5
     stride: 2
     dilation: 3
     activation: relu
     dropout: 0.1
-  - in_channels: 32
-    out_channels: 32
+  - in_channels: 16
+    out_channels: 16
     kernel_size: 3
     stride: 1
     dilation: 2
@@ -40,7 +40,7 @@ conv2d_block_param_list:
     dropout: 0.1
 
 cls_head_param:
-  input_dim: 864
+  input_dim: 432
   num_layers: 2
   hidden_dims:
     - 128

examples/vm_sound_classification/run.sh

@@ -12,7 +12,7 @@ sh run.sh --stage 2 --stop_stage 2 --system_version windows --file_folder_name f
 E:/Users/tianx/HuggingDatasets/vm_sound_classification/data/wav_finished/id-ID/wav_finished/*/*.wav" \
 --label_plan 4
 
-sh run.sh --stage 0 --stop_stage 5 --system_version centos --file_folder_name file_dir --final_model_name vm_sound_classification2-ch32 \
+sh run.sh --stage 0 --stop_stage 5 --system_version centos --file_folder_name file_dir --final_model_name vm_sound_classification2-ch16 \
 --filename_patterns "/data/tianxing/PycharmProjects/datasets/voicemail/*/wav_finished/*/*.wav" --label_plan 2
 
 "