✍️ fixing typos

ebraraktas · Apr 28, 2021 · 74de9fc · 74de9fc
1 parent 11a50e5
commit 74de9fc
Show file tree

Hide file tree

Showing 15 changed files with 22 additions and 20 deletions.
diff --git a/examples/conformer/config.yml b/examples/conformer/config.yml
@@ -21,7 +21,7 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: False
 
 decoder_config:
   vocabulary: ./vocabularies/librispeech/librispeech_train_4_1030.subwords

diff --git a/examples/contextnet/config.yml b/examples/contextnet/config.yml
@@ -21,7 +21,7 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: False
 
 decoder_config:
   vocabulary: null

diff --git a/examples/deepspeech2/config.yml b/examples/deepspeech2/config.yml
@@ -21,7 +21,7 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: False
 
 decoder_config:
   vocabulary: null

diff --git a/examples/jasper/config.yml b/examples/jasper/config.yml
@@ -21,7 +21,7 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: False
 
 decoder_config:
   vocabulary: null

diff --git a/examples/rnn_transducer/config.yml b/examples/rnn_transducer/config.yml
@@ -21,7 +21,9 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: True
+  center: False
+  top_db: null
 
 decoder_config:
   vocabulary: null

diff --git a/notebooks/conformer.ipynb b/notebooks/conformer.ipynb
@@ -42,7 +42,7 @@
     "        \"preemphasis\": 0.97,\n",
     "        \"normalize_signal\": True,\n",
     "        \"normalize_feature\": True,\n",
-    "        \"normalize_per_feature\": False,\n",
+    "        \"normalize_per_frame\": False,\n",
     "    },\n",
     "    \"decoder_config\": {\n",
     "        \"vocabulary\": None,\n",

diff --git a/notebooks/contextnet.ipynb b/notebooks/contextnet.ipynb
@@ -42,7 +42,7 @@
     "        \"preemphasis\": 0.97,\n",
     "        \"normalize_signal\": True,\n",
     "        \"normalize_feature\": True,\n",
-    "        \"normalize_per_feature\": False,\n",
+    "        \"normalize_per_frame\": False,\n",
     "    },\n",
     "    \"decoder_config\": {\n",
     "        \"vocabulary\": None,\n",

diff --git a/notebooks/rnn_transducer.ipynb b/notebooks/rnn_transducer.ipynb
@@ -42,7 +42,7 @@
     "        \"preemphasis\": 0.97,\n",
     "        \"normalize_signal\": True,\n",
     "        \"normalize_feature\": True,\n",
-    "        \"normalize_per_feature\": False,\n",
+    "        \"normalize_per_frame\": False,\n",
     "    },\n",
     "    \"decoder_config\": {\n",
     "        \"vocabulary\": None,\n",

diff --git a/tensorflow_asr/featurizers/speech_featurizers.py b/tensorflow_asr/featurizers/speech_featurizers.py
@@ -203,7 +203,7 @@ def __init__(self, speech_config: dict):
             "pitch": bool,
             "normalize_signal": bool,
             "normalize_feature": bool,
-            "normalize_per_feature": bool
+            "normalize_per_frame": bool
         }
         """
         # Samples
@@ -320,21 +320,21 @@ def extract(self, signal: np.ndarray) -> np.ndarray:
         if self.delta:
             delta = librosa.feature.delta(original_features.T).T
             if self.normalize_feature:
-                delta = normalize_audio_feature(delta, per_feature=self.normalize_per_feature)
+                delta = normalize_audio_feature(delta, per_frame=self.normalize_per_frame)
             features = np.concatenate([features, np.expand_dims(delta, axis=-1)], axis=-1)
 
         if self.delta_delta:
             delta_delta = librosa.feature.delta(original_features.T, order=2).T
             if self.normalize_feature:
                 delta_delta = normalize_audio_feature(
-                    delta_delta, per_feature=self.normalize_per_feature)
+                    delta_delta, per_frame=self.normalize_per_frame)
             features = np.concatenate([features, np.expand_dims(delta_delta, axis=-1)], axis=-1)
 
         if self.pitch:
             pitches = self.compute_pitch(signal)
             if self.normalize_feature:
                 pitches = normalize_audio_feature(
-                    pitches, per_feature=self.normalize_per_feature)
+                    pitches, per_frame=self.normalize_per_frame)
             features = np.concatenate([features, np.expand_dims(pitches, axis=-1)], axis=-1)
 
         return features

diff --git a/tests/conformer/config.yml b/tests/conformer/config.yml
@@ -21,7 +21,7 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: False
 
 decoder_config:
   vocabulary: null

diff --git a/tests/contextnet/config.yml b/tests/contextnet/config.yml
@@ -21,7 +21,7 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: False
 
 decoder_config:
   vocabulary: null

diff --git a/tests/deepspeech2/config.yml b/tests/deepspeech2/config.yml
@@ -21,7 +21,7 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: False
 
 decoder_config:
   vocabulary: null

diff --git a/tests/featurizer/test_sentencepiece.py b/tests/featurizer/test_sentencepiece.py
@@ -46,7 +46,7 @@ def test_featurizer():
         "preemphasis": 0.97,
         "normalize_signal": True,
         "normalize_feature": True,
-        "normalize_per_feature": False}
+        "normalize_per_frame": False}
 
     text_featurizer_sentencepiece = SentencePieceFeaturizer.load_from_file(config, None)
     subwords_path = os.path.join(os.path.abspath(os.path.dirname(__file__)),
@@ -99,7 +99,7 @@ def test_iextract():
         "preemphasis": 0.97,
         "normalize_signal": True,
         "normalize_feature": True,
-        "normalize_per_feature": False}
+        "normalize_per_frame": False}
 
     text_featurizer_sentencepiece = SentencePieceFeaturizer.load_from_file(config, None)
     speech_featurizer = TFSpeechFeaturizer(config_speech)
@@ -124,7 +124,7 @@ def test_iextract():
     print(file_path)
     with open(file_path, "r") as f:
         lines = f.read().splitlines()
-    m=re.search(r"[0-9]+-[0-9]+-[0-9]+\s+([\w\s]+)",lines[0])
+    m = re.search(r"[0-9]+-[0-9]+-[0-9]+\s+([\w\s]+)", lines[0])
     transcript = m.groups(1)[0].lower()
 
     assert(labels == transcript)
diff --git a/tests/jasper/config.yml b/tests/jasper/config.yml
@@ -21,7 +21,7 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: False
 
 decoder_config:
   vocabulary: null

diff --git a/tests/rnn_transducer/config.yml b/tests/rnn_transducer/config.yml
@@ -21,7 +21,7 @@ speech_config:
   preemphasis: 0.97
   normalize_signal: True
   normalize_feature: True
-  normalize_per_feature: False
+  normalize_per_frame: False
 
 decoder_config:
   vocabulary: null