fix bug

README.md

@@ -126,7 +126,7 @@ Of course, this data pipeline could be also followed to know the basic principle
 - **Training Strategy**
   + [x] [AdamW](https://arxiv.org/pdf/1711.05101v1.pdf) + [WarmRestarts](https://arxiv.org/pdf/1608.03983v4.pdf)
   + [x] SGD + [ReduceLROnPlateau](https://pytorch.org/docs/stable/optim.html#torch.optim.lr_scheduler.ReduceLROnPlateau)
-  + [x] [Training with Magin Decay Strategy](https://arxiv.org/pdf/1904.03479.pdf)
+  + [x] [Training with Magin Warmup Strategy](https://arxiv.org/pdf/1904.03479.pdf)
   + [x] [Heated Up Strategy](https://arxiv.org/pdf/1809.04157.pdf)
   + [x] [Multi-task Learning with Phonetic Information](http://yiliu.org.cn/papers/Speaker_Embedding_Extraction_with_Phonetic_Information.pdf) (Kaldi) [[Source codes](https://github.com/mycrazycracy/speaker-embedding-with-phonetic-information) was contributed by [Yi Liu](http://yiliu.org.cn/). Thanks.]
   + [ ] Multi-task Learning with Phonetic Information (Pytorch)

augmentDataByNoise.sh

@@ -172,7 +172,7 @@ fi
 
 num_origin_utts=$(wc -l $data/reco2dur | awk '{print $1}')
 # Use awk to replace bc to compute float value.
-status=$(echo $fator $num | awk '{if($1-$2>0){print 1}else{print 0}}')
+status=$(echo $factor $num | awk '{if($1-$2>0){print 1}else{print 0}}')
 [ $status -eq 1 ] && factor=$num # Get min
 num_additive_utts=$(echo $num_origin_utts $factor | awk '{print int($1*$2)}')
 

pytorch/libs/nnet/pooling.py

@@ -192,6 +192,9 @@ def __init__(self, input_dim, num_head=1, split_input=True, share=True, affine_l
         if share:
             # weight: [input_dim, 1] or [input_dim, hidden_size] -> [hidden_size, 1]
             final_dim = 1
+        elif split_input:
+            # weight: [input_dim, input_dim // num_head] or [input_dim, hidden_size] -> [hidden_size, input_dim // num_head]
+            final_dim = input_dim // num_head
         else:
             # weight: [input_dim, input_dim] or [input_dim, hidden_size] -> [hidden_size, input_dim]
             final_dim = input_dim