revise example/ting/s1

deepspeech/exps/deepspeech2/config.py

@@ -72,7 +72,7 @@ _C.collator =CN(
         use_dB_normalization=True,
         target_dB=-20,
         dither=1.0,  # feature dither
-        keep_transcription_text=True
+        keep_transcription_text=False
     ))
 
 DeepSpeech2Model.params(_C.model)

deepspeech/exps/deepspeech2/model.py

@@ -336,13 +336,14 @@ class DeepSpeech2Tester(DeepSpeech2Trainer):
         # config.data.max_output_input_ratio = float('inf')
         test_dataset = ManifestDataset.from_config(config)
 
+        config.collator.keep_transcription_text = True
         # return text ord id
         self.test_loader = DataLoader(
             test_dataset,
             batch_size=config.decoding.batch_size,
             shuffle=False,
             drop_last=False,
-            collate_fn=SpeechCollator(config=config, keep_transcription_text=True))
+            collate_fn=SpeechCollator.from_config(config))
         logger.info("Setup test Dataloader!")
 
     def setup_output_dir(self):

deepspeech/exps/u2/config.py

@@ -22,6 +22,13 @@ _C = CfgNode()
 
 _C.data = ManifestDataset.params()
 
+_C.collator =CfgNode(
+    dict(
+        augmentation_config="",
+        unit_type="char",
+        keep_transcription_text=False
+    ))
+
 _C.model = U2Model.params()
 
 _C.training = U2Trainer.params()

deepspeech/exps/u2/model.py

@@ -221,7 +221,7 @@ class U2Trainer(Trainer):
         config.data.augmentation_config = ""
         dev_dataset = ManifestDataset.from_config(config)
 
-        collate_fn = SpeechCollator(keep_transcription_text=False)
+        collate_fn = SpeechCollator.from_config(config)
         if self.parallel:
             batch_sampler = SortagradDistributedBatchSampler(
                 train_dataset,
@@ -266,12 +266,13 @@ class U2Trainer(Trainer):
         # config.data.max_output_input_ratio = float('inf')
         test_dataset = ManifestDataset.from_config(config)
         # return text ord id
+        config.collator.keep_transcription_text = True
         self.test_loader = DataLoader(
             test_dataset,
             batch_size=config.decoding.batch_size,
             shuffle=False,
             drop_last=False,
-            collate_fn=SpeechCollator(keep_transcription_text=True))
+            collate_fn=SpeechCollator.from_config(config))
         logger.info("Setup train/valid/test Dataloader!")
 
     def setup_model(self):
@@ -375,7 +376,7 @@ class U2Tester(U2Trainer):
         error_rate_func = error_rate.cer if cfg.error_rate_type == 'cer' else error_rate.wer
 
         start_time = time.time()
-        text_feature = self.test_loader.dataset.text_feature
+        text_feature = self.test_loader.collate_fn.text_feature
         target_transcripts = self.ordid2token(texts, texts_len)
         result_transcripts = self.model.decode(
             audio,
@@ -423,7 +424,7 @@ class U2Tester(U2Trainer):
         self.model.eval()
         logger.info(f"Test Total Examples: {len(self.test_loader.dataset)}")
 
-        stride_ms = self.test_loader.dataset.stride_ms
+        stride_ms = self.config.collator.stride_ms
         error_rate_type = None
         errors_sum, len_refs, num_ins = 0.0, 0, 0
         num_frames = 0.0

deepspeech/frontend/utility.py

@@ -82,7 +82,7 @@ def read_manifest(
         ]
         if all(conditions):
             manifest.append(json_data)
-    return manifest, json_data["feat_shape"][-1]
+    return manifest
 
 
 def rms_to_db(rms: float):

deepspeech/io/collator.py

@@ -56,7 +56,7 @@ class SpeechCollator():
                 use_dB_normalization=True,
                 target_dB=-20,
                 dither=1.0,  # feature dither
-                keep_transcription_text=True
+                keep_transcription_text=False
             ))
 
         if config is not None:
@@ -75,7 +75,7 @@ class SpeechCollator():
         """
         assert 'augmentation_config' in config.collator
         assert 'keep_transcription_text' in config.collator
-        assert 'mean_std_filepath' in config.collator
+        assert 'mean_std_filepath' in config.data
         assert 'vocab_filepath' in config.data
         assert 'specgram_type' in config.collator
         assert 'n_fft' in config.collator
@@ -94,7 +94,7 @@ class SpeechCollator():
         speech_collator = cls(
                 aug_file=aug_file,
                 random_seed=0,
-                mean_std_filepath=config.collator.mean_std_filepath,
+                mean_std_filepath=config.data.mean_std_filepath,
                 unit_type=config.collator.unit_type,
                 vocab_filepath=config.data.vocab_filepath,
                 spm_model_prefix=config.collator.spm_model_prefix,
@@ -282,26 +282,11 @@ class SpeechCollator():
         text_lens = np.array(text_lens).astype(np.int64)
         return utts, padded_audios, audio_lens, padded_texts, text_lens
 
-    @property
-    def vocab_size(self):
-        return self._speech_featurizer.vocab_size
-
-    @property
-    def vocab_list(self):
-        return self._speech_featurizer.vocab_list
-
-    @property
-    def vocab_dict(self):
-        return self._speech_featurizer.vocab_dict
 
     @property
     def text_feature(self):
-        return self._text_featurizer
-        self._speech_featurizer.text_feature
+        return self._speech_featurizer.text_feature
 
-    @property
-    def feature_size(self):
-        return self._speech_featurizer.feature_size
 
     @property
     def stride_ms(self):

deepspeech/io/dataset.py

@@ -161,7 +161,7 @@ class ManifestDataset(Dataset):
         # self._rng = np.random.RandomState(random_seed)
 
         # read manifest
-        self._manifest, self._feature_size = read_manifest(
+        self._manifest = read_manifest(
             manifest_path=manifest_path,
             max_input_len=max_input_len,
             min_input_len=min_input_len,
@@ -213,16 +213,8 @@ class ManifestDataset(Dataset):
         Returns:
             int: audio feature size.
         """
-        return self._feature_size
+        return self._manifest[0]["feat_shape"][-1]
 
-    @property
-    def stride_ms(self):
-        """time length in `ms` unit per frame
-
-        Returns:
-            float: time(ms)/frame
-        """
-        return self._audio_featurizer.stride_ms
     
 
     def __len__(self):

examples/tiny/s0/conf/deepspeech2.yaml

@@ -6,7 +6,6 @@ data:
   mean_std_filepath: data/mean_std.json
   unit_type: char
   vocab_filepath: data/vocab.txt 
-  augmentation_config: conf/augmentation.json
   batch_size: 4
   min_input_len: 0.0
   max_input_len: 27.0
@@ -14,18 +13,6 @@ data:
   max_output_len: 400.0
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
-  specgram_type: linear
-  target_sample_rate: 16000
-  max_freq: None
-  n_fft: None
-  stride_ms: 10.0
-  window_ms: 20.0
-  delta_delta: False
-  dither: 1.0
-  use_dB_normalization: True
-  target_dB: -20
-  random_seed: 0
-  keep_transcription_text: False
   sortagrad: True 
   shuffle_method: batch_shuffle
   num_workers: 0
@@ -33,7 +20,6 @@ data:
 collator:
   augmentation_config: conf/augmentation.json
   random_seed: 0
-  mean_std_filepath: data/mean_std.json
   spm_model_prefix: 
   specgram_type: linear
   feat_dim: 
@@ -46,7 +32,7 @@ collator:
   use_dB_normalization: True
   target_dB: -20
   dither: 1.0
-  keep_transcription_text: True
+  keep_transcription_text: False
   
 model:
   num_conv_layers: 2

examples/tiny/s1/conf/transformer.yaml

@@ -7,7 +7,6 @@ data:
   unit_type: 'spm'
   spm_model_prefix: 'data/bpe_unigram_200'
   mean_std_filepath: ""
-  augmentation_config: conf/augmentation.json
   batch_size: 4
   min_input_len: 0.5  # second
   max_input_len: 20.0 # second
@@ -16,23 +15,26 @@ data:
   min_output_input_ratio: 0.05
   max_output_input_ratio: 10.0
   raw_wav: True  # use raw_wav or kaldi feature
-  specgram_type: fbank #linear, mfcc, fbank
+  sortagrad: True 
+  shuffle_method: batch_shuffle
+  num_workers: 0 #2
+
+collator:
+  augmentation_config: conf/augmentation.json
+  random_seed: 0
+  spm_model_prefix: 
+  specgram_type: fbank
   feat_dim: 80
   delta_delta: False
-  dither: 1.0
-  target_sample_rate: 16000
-  max_freq: None
-  n_fft: None
   stride_ms: 10.0
-  window_ms: 25.0
+  window_ms: 20.0
+  n_fft: None
+  max_freq: None
+  target_sample_rate: 16000
   use_dB_normalization: True
   target_dB: -20
-  random_seed: 0
+  dither: 1.0
   keep_transcription_text: False
-  sortagrad: True 
-  shuffle_method: batch_shuffle
-  num_workers: 2
-
 
 # network architecture
 model:
@@ -70,7 +72,7 @@ model:
 
 
 training:
-  n_epoch: 2
+  n_epoch: 3
   accum_grad: 1
   global_grad_clip: 5.0
   optim: adam