Merge branch 'master' into autograd-phase-vocoder

Author: mthrok

.circleci/regenerate.py

@@ -15,6 +15,7 @@
 """
 
 import jinja2
+from jinja2 import select_autoescape
 import yaml
 import os.path
 
@@ -213,7 +214,7 @@ def unittest_workflows(indentation=6):
     env = jinja2.Environment(
         loader=jinja2.FileSystemLoader(d),
         lstrip_blocks=True,
-        autoescape=False,
+        autoescape=select_autoescape(enabled_extensions=('html', 'xml')),
     )
 
     with open(os.path.join(d, 'config.yml'), 'w') as f:

.github/workflows/bandit.yml

@@ -0,0 +1,23 @@
+# GitHub Actions Bandit Workflow
+
+name: Bandit
+
+on:
+  pull_request:
+    branches: [ master ]
+
+  workflow_dispatch:
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v2
+
+      # Task will fail if any high-severity issues are found
+      # Ignoring submodules
+      - name: Run Bandit Security Analysis
+        run: |
+              python -m pip install bandit
+              python -m bandit -r . -x ./third_party -lll

.github/workflows/codeql.yml

@@ -0,0 +1,43 @@
+# GitHub Actions CodeQL Workflow
+
+name: CodeQL
+
+on:
+  pull_request:
+    branches: [ master ]
+
+  workflow_dispatch:
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v2
+
+      - name: Initialize CodeQL
+        uses: github/codeql-action/init@v1
+        with:
+          languages: python, cpp
+
+      - name: Install Ninja
+        run: |
+              sudo apt-get update -y
+              sudo apt-get install -y ninja-build
+
+      - name: Update submodules
+        run: git submodule update --init --recursive
+
+      - name: Install Torch
+        run: |
+              python -m pip install cmake
+              python -m pip install torch==1.8.1+cpu -f https://download.pytorch.org/whl/torch_stable.html
+              sudo ln -s /usr/bin/ninja /usr/bin/ninja-build
+
+      - name: Build TorchAudio
+        run: BUILD_SOX=1 USE_CUDA=0 python setup.py develop --user
+
+      # If any code scanning alerts are found, they will be under Security -> CodeQL
+      # Link: https://github.com/pytorch/audio/security/code-scanning
+      - name: Perform CodeQL Analysis
+        uses: github/codeql-action/analyze@v1

test/torchaudio_unittest/transforms/autograd_test_impl.py

@@ -80,6 +80,15 @@ def test_melspectrogram(self):
         waveform = get_whitenoise(sample_rate=sample_rate, duration=0.05, n_channels=2)
         self.assert_grad(transform, [waveform], nondet_tol=1e-10)
 
+    @parameterized.expand([(0, ), (0.99, )])
+    def test_griffinlim(self, momentum):
+        n_fft = 400
+        n_frames = 5
+        n_iter = 3
+        spec = torch.rand(n_fft // 2 + 1, n_frames) * n_fft
+        transform = T.GriffinLim(n_fft=n_fft, n_iter=n_iter, momentum=momentum, rand_init=False)
+        self.assert_grad(transform, [spec], nondet_tol=1e-10)
+
     @parameterized.expand([(False, ), (True, )])
     def test_mfcc(self, log_mels):
         sample_rate = 8000
@@ -104,7 +113,13 @@ def test_fade(self, fade_shape):
         waveform = get_whitenoise(sample_rate=8000, duration=0.05, n_channels=2)
         self.assert_grad(transform, [waveform], nondet_tol=1e-10)
 
-    @nested_params(
+    def test_spectral_centroid(self):
+        sample_rate = 8000
+        transform = T.SpectralCentroid(sample_rate=sample_rate)
+        waveform = get_whitenoise(sample_rate=sample_rate, duration=0.05, n_channels=2)
+        self.assert_grad(transform, [waveform], nondet_tol=1e-10)
+
+            @nested_params(
         [0.7, 0.8, 0.9, 1.0, 1.3],
         [True, False],
     )

torchaudio/datasets/utils.py

@@ -29,7 +29,8 @@ def stream_url(url: str,
 
     # If we already have the whole file, there is no need to download it again
     req = urllib.request.Request(url, method="HEAD")
-    url_size = int(urllib.request.urlopen(req).info().get("Content-Length", -1))
+    with urllib.request.urlopen(req) as response:
+        url_size = int(response.info().get("Content-Length", -1))
     if url_size == start_byte:
         return
 

torchaudio/functional/functional.py

@@ -203,7 +203,7 @@ def griffinlim(
                               hop_length=hop_length,
                               win_length=win_length,
                               window=window,
-                              length=length).float()
+                              length=length)
 
         # Rebuild the spectrogram
         rebuilt = torch.view_as_real(
@@ -793,7 +793,7 @@ def compute_deltas(
 
     # pack batch
     shape = specgram.size()
-    specgram = specgram.reshape(-1, 1, shape[-1])
+    specgram = specgram.reshape(1, -1, shape[-1])
 
     assert win_length >= 3
 
@@ -804,9 +804,9 @@ def compute_deltas(
 
     specgram = torch.nn.functional.pad(specgram, (n, n), mode=mode)
 
-    kernel = torch.arange(-n, n + 1, 1, device=device, dtype=dtype).view(1, 1, -1)
+    kernel = torch.arange(-n, n + 1, 1, device=device, dtype=dtype).repeat(specgram.shape[1], 1, 1)
 
-    output = torch.nn.functional.conv1d(specgram, kernel) / denom
+    output = torch.nn.functional.conv1d(specgram, kernel, groups=specgram.shape[1]) / denom
 
     # unpack batch
     output = output.reshape(shape)

torchaudio/transforms.py

@@ -155,7 +155,7 @@ class GriffinLim(torch.nn.Module):
            | "Signal estimation from modified short-time Fourier transform,"
            | IEEE Trans. ASSP, vol.32, no.2, pp.236–243, Apr. 1984.
     """
-    __constants__ = ['n_fft', 'n_iter', 'win_length', 'hop_length', 'power', 'normalized',
+    __constants__ = ['n_fft', 'n_iter', 'win_length', 'hop_length', 'power',
                      'length', 'momentum', 'rand_init']
 
     def __init__(self,
@@ -172,7 +172,7 @@ def __init__(self,
         super(GriffinLim, self).__init__()
 
         assert momentum < 1, 'momentum={} > 1 can be unstable'.format(momentum)
-        assert momentum > 0, 'momentum={} < 0'.format(momentum)
+        assert momentum >= 0, 'momentum={} < 0'.format(momentum)
 
         self.n_fft = n_fft
         self.n_iter = n_iter
@@ -768,6 +768,7 @@ class Fade(torch.nn.Module):
         fade_shape (str, optional): Shape of fade. Must be one of: "quarter_sine",
             "half_sine", "linear", "logarithmic", "exponential". (Default: ``"linear"``)
     """
+
     def __init__(self,
                  fade_in_len: int = 0,
                  fade_out_len: int = 0,
@@ -877,6 +878,7 @@ class FrequencyMasking(_AxisMasking):
             example/channel in the batch. (Default: ``False``)
             This option is applicable only when the input tensor is 4D.
     """
+
     def __init__(self, freq_mask_param: int, iid_masks: bool = False) -> None:
         super(FrequencyMasking, self).__init__(freq_mask_param, 1, iid_masks)
 
@@ -891,6 +893,7 @@ class TimeMasking(_AxisMasking):
             example/channel in the batch. (Default: ``False``)
             This option is applicable only when the input tensor is 4D.
     """
+
     def __init__(self, time_mask_param: int, iid_masks: bool = False) -> None:
         super(TimeMasking, self).__init__(time_mask_param, 2, iid_masks)