Merge pull request #692 from PaddlePaddle/bugfix

fix conf for new datapipe; u2 export inputspec

deepspeech/exps/u2/model.py

@@ -511,10 +511,9 @@ class U2Tester(U2Trainer):
                                                    self.args.checkpoint_path)
         feat_dim = self.test_loader.collate_fn.feature_size
         input_spec = [
-            paddle.static.InputSpec(
-                shape=[None, feat_dim, None],
-                dtype='float32'),  # audio, [B,D,T]
-            paddle.static.InputSpec(shape=[None],
+            paddle.static.InputSpec(shape=[1, None, feat_dim],
+                                    dtype='float32'),  # audio, [B,T,D]
+            paddle.static.InputSpec(shape=[1],
                                     dtype='int64'),  # audio_length, [B]
         ]
         return infer_model, input_spec

examples/librispeech/s1/conf/chunk_confermer.yaml

@@ -3,18 +3,20 @@ data:
   train_manifest: data/manifest.train
   dev_manifest: data/manifest.dev
   test_manifest: data/manifest.test
-  vocab_filepath: data/vocab.txt 
-  unit_type: 'spm'
-  spm_model_prefix: 'data/bpe_unigram_5000'
-  mean_std_filepath: ""
-  augmentation_config: conf/augmentation.json
-  batch_size: 4
   min_input_len: 0.5
   max_input_len: 20.0
   min_output_len: 0.0
   max_output_len: 400.0
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
+
+collator:
+  vocab_filepath: data/vocab.txt 
+  unit_type: 'spm'
+  spm_model_prefix: 'data/bpe_unigram_5000'
+  mean_std_filepath: ""
+  augmentation_config: conf/augmentation.json
+  batch_size: 16
   raw_wav: True  # use raw_wav or kaldi feature
   specgram_type: fbank #linear, mfcc, fbank
   feat_dim: 80
@@ -80,7 +82,7 @@ model:
 
 training:
   n_epoch: 120
-  accum_grad: 1
+  accum_grad: 8
   global_grad_clip: 5.0
   optim: adam
   optim_conf:

examples/librispeech/s1/conf/chunk_transformer.yaml

@@ -3,18 +3,20 @@ data:
   train_manifest: data/manifest.train
   dev_manifest: data/manifest.dev
   test_manifest: data/manifest.test
-  vocab_filepath: data/vocab.txt 
-  unit_type: 'spm'
-  spm_model_prefix: 'data/bpe_unigram_5000'
-  mean_std_filepath: ""
-  augmentation_config: conf/augmentation.json
-  batch_size: 64
   min_input_len: 0.5  # second
   max_input_len: 20.0 # second
   min_output_len: 0.0 # tokens
   max_output_len: 400.0 # tokens
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
+
+collator:
+  vocab_filepath: data/vocab.txt 
+  unit_type: 'spm'
+  spm_model_prefix: 'data/bpe_unigram_5000'
+  mean_std_filepath: ""
+  augmentation_config: conf/augmentation.json
+  batch_size: 64
   raw_wav: True  # use raw_wav or kaldi feature
   specgram_type: fbank #linear, mfcc, fbank
   feat_dim: 80
@@ -103,6 +105,6 @@ decoding:
       # >0: for decoding, use fixed chunk size as set.
       # 0: used for training, it's prohibited here. 
   num_decoding_left_chunks: -1  # number of left chunks for decoding. Defaults to -1.
-  simulate_streaming: False  # simulate streaming inference. Defaults to False.
+  simulate_streaming: true  # simulate streaming inference. Defaults to False.
 
 

examples/librispeech/s1/conf/conformer.yaml

@@ -3,18 +3,20 @@ data:
   train_manifest: data/manifest.train
   dev_manifest: data/manifest.dev
   test_manifest: data/manifest.test-clean
-  vocab_filepath: data/vocab.txt 
-  unit_type: 'spm'
-  spm_model_prefix: 'data/bpe_unigram_5000'
-  mean_std_filepath: ""
-  augmentation_config: conf/augmentation.json
-  batch_size: 16
   min_input_len: 0.5  # seconds
   max_input_len: 20.0 # seconds
   min_output_len: 0.0 # tokens
   max_output_len: 400.0 # tokens
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
+
+collator:
+  vocab_filepath: data/vocab.txt 
+  unit_type: 'spm'
+  spm_model_prefix: 'data/bpe_unigram_5000'
+  mean_std_filepath: ""
+  augmentation_config: conf/augmentation.json
+  batch_size: 16
   raw_wav: True  # use raw_wav or kaldi feature
   specgram_type: fbank #linear, mfcc, fbank
   feat_dim: 80

examples/librispeech/s1/conf/transformer.yaml

@@ -3,18 +3,20 @@ data:
   train_manifest: data/manifest.train
   dev_manifest: data/manifest.dev
   test_manifest: data/manifest.test-clean
-  vocab_filepath: data/vocab.txt
-  unit_type: 'spm'
-  spm_model_prefix: 'data/bpe_unigram_5000'
-  mean_std_filepath: ""
-  augmentation_config: conf/augmentation.json
-  batch_size: 64
   min_input_len: 0.5  # second
   max_input_len: 20.0 # second
   min_output_len: 0.0 # tokens
   max_output_len: 400.0 # tokens
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
+
+collator:
+  vocab_filepath: data/vocab.txt
+  unit_type: 'spm'
+  spm_model_prefix: 'data/bpe_unigram_5000'
+  mean_std_filepath: ""
+  augmentation_config: conf/augmentation.json
+  batch_size: 64
   raw_wav: True  # use raw_wav or kaldi feature
   specgram_type: fbank #linear, mfcc, fbank
   feat_dim: 80