Merge pull request #705 from PaddlePaddle/fix

Fix librispeech s0

examples/aishell/s1/local/download_lm_ch.sh

@@ -1 +0,0 @@
-../../s0/local/download_lm_ch.sh

examples/dataset/librispeech/librispeech.py

@@ -114,7 +114,7 @@ def create_manifest(data_dir, manifest_path):
         for line in json_lines:
             out_file.write(line + '\n')
 
-    subset = os.path.splitext(manifest_path)[1]
+    subset = manifest_path
     with open(subset + '.meta', 'w') as f:
         print(f"{subset}:", file=f)
         print(f"{total_num} utts", file=f)

examples/librispeech/s0/conf/augmentation.json

@@ -15,5 +15,20 @@
       "max_shift_ms": 5
     },
     "prob": 1.0
+  },
+  {
+    "type": "specaug",
+    "params": {
+      "F": 10,
+      "T": 50,
+      "n_freq_masks": 2,
+      "n_time_masks": 2,
+      "p": 1.0,
+      "W": 80,
+      "adaptive_number_ratio": 0,
+      "adaptive_size_ratio": 0,
+      "max_n_time_masks": 20
+    },
+    "prob": 1.0
   }
 ]

examples/librispeech/s1/README.md

@@ -15,10 +15,10 @@
 
 | Model | Params | Config | Augmentation| Test set | Decode method | Chunk Size & Left Chunks | Loss | WER |  
 | --- | --- | --- | --- | --- | --- | --- | --- | --- |  
-| conformer | 47.63 M | conf/chunk_conformer.yaml | spec_aug + shift | test-clean | attention | 16, -1 | 7.01250648 |  |  
-| conformer | 47.63 M | conf/chunk_conformer.yaml | spec_aug + shift | test-clean | ctc_greedy_search | 16, -1 | 7.01250648 |  |  
+| conformer | 47.63 M | conf/chunk_conformer.yaml | spec_aug + shift | test-clean | attention | 16, -1 | 7.01250648 | 0.069548 |  
+| conformer | 47.63 M | conf/chunk_conformer.yaml | spec_aug + shift | test-clean | ctc_greedy_search | 16, -1 | 7.01250648 | 0.094753 |  
 | conformer | 47.63 M | conf/chunk_conformer.yaml | spec_aug + shift | test-clean | ctc_prefix_beam_search | 16, -1 | 7.01250648 |  |  
-| conformer | 47.63 M | conf/chunk_conformer.yaml | spec_aug + shift | test-clean | attention_rescoring | 16, -1 | 7.01250648 | |  
+| conformer | 47.63 M | conf/chunk_conformer.yaml | spec_aug + shift | test-clean | attention_rescoring | 16, -1 | 7.01250648 |  |  
 
 
 

examples/librispeech/s1/conf/chunk_conformer.yaml

@@ -81,7 +81,7 @@ model:
 
 
 training:
-  n_epoch: 120
+  n_epoch: 240
   accum_grad: 8
   global_grad_clip: 5.0
   optim: adam

examples/tiny/s1/local/download_lm_en.sh

@@ -1 +0,0 @@
-../../s0/local/download_lm_en.sh

tools/Makefile

@@ -18,7 +18,7 @@ kenlm.done:
 	apt install -y build-essential cmake libboost-system-dev libboost-thread-dev libboost-program-options-dev libboost-test-dev libeigen3-dev zlib1g-dev libbz2-dev liblzma-dev
 	apt-get install -y gcc-5 g++-5 && update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-5 50  && update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-5 50
 	test -d kenlm || wget -O - https://kheafield.com/code/kenlm.tar.gz | tar xz
-	mkdir -p kenlm/build && cd kenlm/build && cmake .. && make -j4 && make install
+	rm -rf kenlm/build && mkdir -p kenlm/build && cd kenlm/build && cmake .. && make -j4 && make install
 	source venv/bin/activate; cd kenlm && python setup.py install
 	touch kenlm.done