Merge pull request #9 from wenet-e2e/dev-jingyonghou

add results of mdtc

examples/hi_xiaowen/s0/README.md

@@ -8,3 +8,5 @@ FRRs with FAR fixed at once per hour:
 | DS_TCN           | 21        | 80        | 0.010807   | 0.014754     |
 | DS_TCN           | 21        | 80(avg30) | 0.009867   | 0.014472     |
 | DS_TCN(spec_aug) | 21        | 80(avg30) | 0.029039   | 0.022648     |
+| MDTC             | 156       | 80(avg10) | 0.007142   | 0.005920     |
+| MDTC_Small       | 31        | 80(avg10) | 0.005357   | 0.005920     |

examples/hi_xiaowen/s0/conf/mdtc_small.yaml

@@ -0,0 +1,51 @@
+dataset_conf:
+    filter_conf:
+        max_length: 2048
+        min_length: 0
+    resample_conf:
+        resample_rate: 16000
+    speed_perturb: false
+    feature_extraction_conf:
+        feature_type: 'mfcc'
+        num_ceps: 80
+        num_mel_bins: 80
+        frame_shift: 10
+        frame_length: 25
+        dither: 1.0
+    feature_dither: 0.0
+    spec_aug: true
+    spec_aug_conf:
+        num_t_mask: 1
+        num_f_mask: 1
+        max_t: 20
+        max_f: 40
+    shuffle: true
+    shuffle_conf:
+        shuffle_size: 1500
+    batch_conf:
+        batch_size: 100
+
+model:
+    hidden_dim: 32
+    preprocessing:
+        type: none
+    backbone:
+        type: mdtc
+        num_stack: 3
+        stack_size: 4
+        kernel_size: 5
+        hidden_dim: 32
+    classifier:
+        type: linear
+
+optim: adam
+optim_conf:
+    lr: 0.001
+    weight_decay: 5e-5
+    warm_up_step: 2500
+
+training_config:
+    grad_clip: 5
+    max_epoch: 100
+    log_interval: 10
+    criterion: max_pooling

examples/hi_xiaowen/s0/run.sh

@@ -9,13 +9,13 @@ stage=0
 stop_stage=4
 num_keywords=2
 
-config=conf/mdtc.yaml
+config=conf/mdtc_small.yaml
 norm_mean=false
 norm_var=false
 gpu_id=0
 
 checkpoint=
-dir=exp/mdtc
+dir=exp/mdtc_small
 
 num_average=10
 score_checkpoint=$dir/avg_${num_average}.pt

kws/model/loss.py

@@ -21,7 +21,7 @@ def max_polling_loss(logits: torch.Tensor,
                      target: torch.Tensor,
                      lengths: torch.Tensor,
                      min_duration: int = 0):
-    """ Max-pooling loss
+    ''' Max-pooling loss
         For keyword, select the frame with the highest posterior.
             The keyword is triggered when any of the frames is triggered.
         For none keyword, select the hardest frame, namely the frame
@@ -36,7 +36,7 @@ def max_polling_loss(logits: torch.Tensor,
     Returns:
         (float): loss of current batch
         (float): accuracy of current batch
-    """
+    '''
     mask = padding_mask(lengths)
     num_utts = logits.size(0)
     num_keywords = logits.size(2)