Enforce flake8 E841 and E821 (#504)

.flake8

 [flake8]
 max-line-length = 120
-ignore = E305,E402,E721,E741,F401,F403,F405,F821,F841,F999,W503,W504
+ignore = E305,E402,E721,E741,F401,F403,F405,F999,W503,W504
 exclude = build,docs/source,_ext

test/test_batch_consistency.py

@@ -64,7 +64,7 @@ def _test_batch(functional, tensor, *args, **kwargs):
     expected = expected.repeat(*ind)
 
     torch.random.manual_seed(42)
-    computed = functional(tensors.clone(), *args, **kwargs)
+    _ = functional(tensors.clone(), *args, **kwargs)
 
 
 class TestFunctional(unittest.TestCase):

test/test_functional_filtering.py

 import os
-import time
 import unittest
 
 import torch
@@ -450,17 +449,13 @@ class TestFunctionalFiltering(unittest.TestCase):
         # SoX method
         E = torchaudio.sox_effects.SoxEffectsChain()
         E.set_input_file(fn_sine)
-        _timing_sox = time.time()
         E.append_effect_to_chain("biquad", [b0, b1, b2, a0, a1, a2])
-        waveform_sox_out, sr = E.sox_build_flow_effects()
-        _timing_sox_run_time = time.time() - _timing_sox
+        waveform_sox_out, _ = E.sox_build_flow_effects()
 
-        _timing_lfilter_filtering = time.time()
-        waveform, sample_rate = torchaudio.load(fn_sine, normalization=True)
+        waveform, _ = torchaudio.load(fn_sine, normalization=True)
         waveform_lfilter_out = F.lfilter(
             waveform, torch.tensor([a0, a1, a2]), torch.tensor([b0, b1, b2])
         )
-        _timing_lfilter_run_time = time.time() - _timing_lfilter_filtering
 
         assert torch.allclose(waveform_sox_out, waveform_lfilter_out, atol=1e-4)
         _test_torchscript_functional(

test/test_sox_effects.py

@@ -257,7 +257,7 @@ class Test_SoxEffectsChain(unittest.TestCase):
             vol = torchaudio.transforms.Vol(gain, gain_type)
             z = vol(x_orig)
             # check if effect worked
-            self.assertTrue(x.allclose(vol(x_orig), rtol=1e-4, atol=1e-4))
+            self.assertTrue(x.allclose(z, rtol=1e-4, atol=1e-4))
 
 
 if __name__ == '__main__':

test/test_transforms.py

@@ -212,8 +212,8 @@ class Tester(unittest.TestCase):
 
     def test_compute_deltas_twochannel(self):
         specgram = torch.tensor([1., 2., 3., 4.]).repeat(1, 2, 1)
-        expected = torch.tensor([[[0.5, 1.0, 1.0, 0.5],
-                                  [0.5, 1.0, 1.0, 0.5]]])
+        _ = torch.tensor([[[0.5, 1.0, 1.0, 0.5],
+                           [0.5, 1.0, 1.0, 0.5]]])
         transform = transforms.ComputeDeltas()
         computed = transform(specgram)
         self.assertTrue(computed.shape == specgram.shape, (computed.shape, specgram.shape))

torchaudio/functional.py

@@ -463,8 +463,6 @@ def create_fb_matrix(
     # freq bins
     # Equivalent filterbank construction by Librosa
     all_freqs = torch.linspace(0, sample_rate // 2, n_freqs)
-    i_freqs = all_freqs.ge(f_min) & all_freqs.le(f_max)
-    freqs = all_freqs[i_freqs]
 
     # calculate mel freq bins
     # hertz to mel(f) is 2595. * math.log10(1. + (f / 700.))
@@ -710,9 +708,6 @@ def lfilter(
     Returns:
         Tensor: Waveform with dimension of `(..., time)`.  Output will be clipped to -1 to 1.
     """
-
-    dim = waveform.dim()
-
     # pack batch
     shape = waveform.size()
     waveform = waveform.view(-1, shape[-1])
@@ -820,12 +815,8 @@ def highpass_biquad(
     Returns:
         Tensor: Waveform dimension of `(..., time)`
     """
-
-    GAIN = 1.
     w0 = 2 * math.pi * cutoff_freq / sample_rate
-    A = math.exp(GAIN / 40.0 * math.log(10))
     alpha = math.sin(w0) / 2. / Q
-    mult = _dB2Linear(max(GAIN, 0))
 
     b0 = (1 + math.cos(w0)) / 2
     b1 = -1 - math.cos(w0)
@@ -853,12 +844,8 @@ def lowpass_biquad(
     Returns:
         Tensor: Waveform of dimension of `(..., time)`
     """
-
-    GAIN = 1.
     w0 = 2 * math.pi * cutoff_freq / sample_rate
-    A = math.exp(GAIN / 40.0 * math.log(10))
     alpha = math.sin(w0) / 2 / Q
-    mult = _dB2Linear(max(GAIN, 0))
 
     b0 = (1 - math.cos(w0)) / 2
     b1 = 1 - math.cos(w0)
@@ -1030,7 +1017,6 @@ def band_biquad(
         https://www.w3.org/2011/audio/audio-eq-cookbook.html#APF
     """
     w0 = 2 * math.pi * central_freq / sample_rate
-    alpha = math.sin(w0) / 2 / Q
     bw_Hz = central_freq / Q
 
     a0 = 1.