format

examples/speechcommand_v1/s0/conf/mdtc.yaml

@@ -42,7 +42,7 @@ model:
 
 optim: adam
 optim_conf:
-    lr: 0.001
+    lr: 0.0002
     weight_decay: 0.00005
 
 training_config:

examples/speechcommand_v1/s0/local/data_download.sh

@@ -14,7 +14,7 @@ url=http://download.tensorflow.org/data/$file_name
 mkdir -p $data_dir
 if [ ! -f $data_dir/$file_name ]; then
     echo "downloading $url..."
-    wget -O $data_dir/$file_name $url 
+    wget -O $data_dir/$file_name $url
 else
     echo "$file_name exist in $data_dir, skip download it"
 fi

examples/speechcommand_v1/s0/local/prepare_speech_command.py

@@ -1,6 +1,5 @@
 #!/usr/bin/env python
 import os
-import sys
 import argparse
 
 CLASSES = 'unknown, yes, no, up, down, left, right, on, off, stop, go'.split(
@@ -9,14 +8,11 @@ CLASS_TO_IDX = {CLASSES[i]: str(i) for i in range(len(CLASSES))}
 
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(
-        description=
-        'prepare kaldi format file for google speech command dataset ')
+        description='prepare kaldi format file for google speech command')
     parser.add_argument(
         '--wav_list',
         required=True,
-        help=
-        'wave list is a file containts full path of a wav file in google speech command dataset'
-    )
+        help='full path of a wav file in google speech command dataset')
     parser.add_argument('--data_dir',
                         required=True,
                         help='folder to write kaldi format files')

examples/speechcommand_v1/s0/local/split_dataset.py

@@ -1,4 +1,5 @@
-"""Splits the google speech commands into train, validation and test set """
+#!/usr/bin/env python
+'''Splits the google speech commands into train, validation and test set'''
 
 import os
 import shutil

examples/speechcommand_v1/s0/run.sh

@@ -6,7 +6,7 @@
 
 export CUDA_VISIBLE_DEVICES="0"
 
-stage=3
+stage=2
 stop_stage=4
 num_keywords=11
 
@@ -21,7 +21,8 @@ dir=exp/mdtc_debug
 num_average=1
 score_checkpoint=$dir/avg_${num_average}.pt
 
-download_dir=/mnt/mnt-data-3/jingyong.hou/data # your data dir
+# your data dir
+download_dir=/mnt/mnt-data-3/jingyong.hou/data
 speech_command_dir=$download_dir/speech_commands_v1
 . tools/parse_options.sh || exit 1;
 
@@ -41,7 +42,7 @@ if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
     data=data/$x
     mkdir -p $data
     # make wav.scp utt2spk text file
-    find $speech_command_dir/$x -name *.wav | grep -v "_background_noise_" > $data/wav.list  
+    find $speech_command_dir/$x -name *.wav | grep -v "_background_noise_" > $data/wav.list
     python local/prepare_speech_command.py --wav_list=$data/wav.list --data_dir=$data
   done
 fi

kws/bin/test.py

@@ -27,7 +27,6 @@ from kws.dataset.dataset import Dataset
 from kws.model.kws_model import init_model
 from kws.utils.checkpoint import load_checkpoint
 from kws.utils.executor import Executor
-from kws.utils.mask import padding_mask
 
 
 def get_args():

kws/bin/train.py

@@ -220,10 +220,11 @@ def main():
         training_config['epoch'] = epoch
         lr = optimizer.param_groups[0]['lr']
         logging.info('Epoch {} TRAIN info lr {}'.format(epoch, lr))
-        executor.train(model, optimizer, train_data_loader, device, writer,
-                       training_config)
+        #executor.train(model, optimizer, train_data_loader, device, writer,
+        #               training_config)
         cv_loss, cv_acc = executor.cv(model, cv_data_loader, device, training_config)
-        logging.info('Epoch {} CV info cv_loss {} cv_acc {}'.format(epoch, cv_loss, cv_acc))
+        logging.info('Epoch {} CV info cv_loss {} cv_acc {}'
+                     .format(epoch, cv_loss, cv_acc))
 
         if args.rank == 0:
             save_model_path = os.path.join(model_dir, '{}.pt'.format(epoch))

kws/model/classifier.py

@@ -30,4 +30,5 @@ class ElementClassifier(nn.Module):
         self.classifier = classifier
 
     def forward(self, x: torch.Tensor):
-        return self.classifier(x)
+        return self.classifier(x)
+

kws/model/kws_model.py

@@ -41,7 +41,7 @@ class KWSModel(torch.nn.Module):
         global_cmvn: Optional[torch.nn.Module],
         preprocessing: Optional[torch.nn.Module],
         backbone: torch.nn.Module,
-    classifier: torch.nn.Module
+        classifier: torch.nn.Module
     ):
         super().__init__()
         self.idim = idim

kws/model/loss.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from logging import log
 import torch
 import torch.nn as nn
 

kws/utils/executor.py

@@ -44,8 +44,9 @@ class Executor:
             if num_utts == 0:
                 continue
             logits = model(feats)
-            loss, acc = criterion(args.get('criterion', 'max_pooling'),
-                                      logits, target, feats_lengths)
+            loss, acc = criterion(
+                           args.get('criterion', 'max_pooling'),
+                           logits, target, feats_lengths)
             loss.backward()
             grad_norm = clip_grad_norm_(model.parameters(), clip)
             if torch.isfinite(grad_norm):