🚀 Reformat code and fix minor bugs

Author: nglehuy

setup.cfg

@@ -1,8 +1,8 @@
 [flake8]
-ignore = E402,E701,E702,E704,E251
+ignore = E402,E701,E702,E704,E251,W503,W504,C901
 max-line-length = 127
 
 [pep8]
-ignore = E402,E701,E702,E704,E251
+ignore = E402,E701,E702,E704,E251,W503,W504,C901
 max-line-length = 127
 indent-size = 4

setup.py

@@ -37,7 +37,7 @@
 
 setuptools.setup(
     name="TensorFlowASR",
-    version="0.4.0",
+    version="0.4.1",
     author="Huy Le Nguyen",
     author_email="[email protected]",
     description="Almost State-of-the-art Automatic Speech Recognition using Tensorflow 2",

tensorflow_asr/augmentations/spec_augment.py

@@ -49,8 +49,8 @@ def mask(self, data: np.ndarray) -> np.ndarray:
 
 class FreqMaskingAugmenter(SpectrogramAugmenter):
     def __init__(self,
-                 mask_factor=27,
-                 name="FreqMaskingAugmenter",
+                 mask_factor: float = 27,
+                 name: str = "FreqMaskingAugmenter",
                  verbose=0):
         super(FreqMaskingAugmenter, self).__init__(
             action=Action.SUBSTITUTE, zone=(0.2, 0.8), name=name, device="cpu", verbose=verbose,
@@ -63,9 +63,9 @@ def substitute(self, data):
 
 class FreqMasking(SpectrogramAugmenter):
     def __init__(self,
-                 num_masks=1,
-                 mask_factor=27,
-                 name="FreqMasking",
+                 num_masks: int = 1,
+                 mask_factor: float = 27,
+                 name: str = "FreqMasking",
                  verbose=0):
         super(FreqMasking, self).__init__(
             action=Action.SUBSTITUTE, zone=(0.2, 0.8), name=name, device="cpu", verbose=verbose,
@@ -79,7 +79,7 @@ def substitute(self, data):
 
 
 class TimeMaskingModel(Spectrogram):
-    def __init__(self, mask_factor: int = 100, p_upperbound: float = 1.0):
+    def __init__(self, mask_factor: float = 100, p_upperbound: float = 1.0):
         """
         Args:
             time_mask_param: parameter W of time masking
@@ -110,9 +110,9 @@ def mask(self, data: np.ndarray) -> np.ndarray:
 
 class TimeMaskingAugmenter(SpectrogramAugmenter):
     def __init__(self,
-                 mask_factor=100,
-                 p_upperbound=1,
-                 name="TimeMaskingAugmenter",
+                 mask_factor: float = 100,
+                 p_upperbound: float = 1,
+                 name: str = "TimeMaskingAugmenter",
                  verbose=0):
         super(TimeMaskingAugmenter, self).__init__(
             action=Action.SUBSTITUTE, zone=(0.2, 0.8), name=name, device="cpu", verbose=verbose,
@@ -125,10 +125,10 @@ def substitute(self, data):
 
 class TimeMasking(SpectrogramAugmenter):
     def __init__(self,
-                 num_masks=1,
-                 mask_factor=100,
-                 p_upperbound=1,
-                 name="TimeMasking",
+                 num_masks: int = 1,
+                 mask_factor: float = 100,
+                 p_upperbound: float = 1,
+                 name: str = "TimeMasking",
                  verbose=0):
         super(TimeMasking, self).__init__(
             action=Action.SUBSTITUTE, zone=(0.2, 0.8), name=name, device="cpu", verbose=verbose,

tensorflow_asr/featurizers/gammatone.py

@@ -142,9 +142,9 @@ def erb_space(
 
 def make_erb_filters(fs, centre_freqs, width=1.0):
     """
-    This function computes the filter coefficients for a bank of 
+    This function computes the filter coefficients for a bank of
     Gammatone filters. These filters were defined by Patterson and Holdworth for
-    simulating the cochlea. 
+    simulating the cochlea.
 
     The result is returned as a :class:`ERBCoeffArray`. Each row of the
     filter arrays contains the coefficients for four second order filters. The

tensorflow_asr/models/ctc.py

@@ -45,7 +45,8 @@ def recognize(self, signals):
 
         def extract_fn(signal): return self.speech_featurizer.tf_extract(signal)
 
-        features = tf.map_fn(extract_fn, signals, fn_output_signature=tf.TensorSpec(self.speech_featurizer.shape, dtype=tf.float32))
+        features = tf.map_fn(extract_fn, signals,
+                             fn_output_signature=tf.TensorSpec(self.speech_featurizer.shape, dtype=tf.float32))
         logits = self(features, training=False)
         probs = tf.nn.softmax(logits)
 
@@ -88,7 +89,8 @@ def recognize_beam(self, signals, lm=False):
 
         def extract_fn(signal): return self.speech_featurizer.tf_extract(signal)
 
-        features = tf.map_fn(extract_fn, signals, fn_output_signature=tf.TensorSpec(self.speech_featurizer.shape, dtype=tf.float32))
+        features = tf.map_fn(extract_fn, signals,
+                             fn_output_signature=tf.TensorSpec(self.speech_featurizer.shape, dtype=tf.float32))
         logits = self(features, training=False)
         probs = tf.nn.softmax(logits)
 

tests/test_conformer.py

@@ -19,8 +19,8 @@
 import tensorflow as tf
 
 from tensorflow_asr.models.conformer import Conformer
-from tensorflow_asr.models.transducer import Transducer
-from tensorflow_asr.models.layers.subsampling import Conv2dSubsampling
+# from tensorflow_asr.models.transducer import Transducer
+# from tensorflow_asr.models.layers.subsampling import Conv2dSubsampling
 from tensorflow_asr.featurizers.text_featurizers import CharFeaturizer
 from tensorflow_asr.featurizers.speech_featurizers import TFSpeechFeaturizer, read_raw_audio
 

tests/test_ctc.py

@@ -3,7 +3,7 @@
 from ctc_decoders import Scorer
 from tensorflow_asr.models.ctc import CtcModel
 from tensorflow_asr.featurizers.text_featurizers import CharFeaturizer
-from tensorflow_asr.featurizers.speech_featurizers import TFSpeechFeaturizer, read_raw_audio
+from tensorflow_asr.featurizers.speech_featurizers import TFSpeechFeaturizer
 from tensorflow_asr.utils.utils import bytes_to_string, merge_two_last_dims
 
 decoder_config = {

tests/test_subword.py

@@ -2,7 +2,6 @@
 import tensorflow as tf
 
 from tensorflow_asr.featurizers.text_featurizers import SubwordFeaturizer
-from tensorflow_asr.featurizers.speech_featurizers import read_raw_audio
 
 parser = argparse.ArgumentParser(prog="test subword")