update to torch 1.8.0 (#79)

* remove bg_iterator from data test

* test 1.8.0

* remove in-place for 1.8 jit

* remove soundfile backend deprecation

* update setup.py requirements

* add 1.7.0 tests

* add tqdm

* update outdir behaviour

* report device

* add another cli test

* relax requirement

* update version

* remove 1.7.0 support

* also from setup.py

.github/workflows/test_unittests.yml

@@ -11,7 +11,7 @@ jobs:
     strategy:
       matrix:
         python-version: [3.6, 3.7, 3.8]
-        pytorch-version: ["1.7.0"]
+        pytorch-version: ["1.8.0"]
 
     # Timeout: https://stackoverflow.com/a/59076067/4521646
     timeout-minutes: 10
@@ -34,6 +34,8 @@ jobs:
         python -m pip install coverage codecov --upgrade-strategy only-if-needed --quiet
         if [ $TORCH_INSTALL == "1.7.0" ]; then
           INSTALL="torch==1.7.0+cpu torchaudio==0.7.0 -f https://download.pytorch.org/whl/torch_stable.html"
+        elif [ $TORCH_INSTALL == "1.8.0" ]; then
+          INSTALL="torch==1.8.0+cpu torchaudio==0.8.0 -f https://download.pytorch.org/whl/torch_stable.html"
         else
           INSTALL="--pre torch torchaudio -f https://download.pytorch.org/whl/nightly/cpu/torch_nightly.html"
         fi

openunmix/cli.py

@@ -113,14 +113,13 @@ def separate():
         "for deployment.",
     )
     args = parser.parse_args()
-    torchaudio.USE_SOUNDFILE_LEGACY_INTERFACE = False
 
     if args.audio_backend != "stempeg":
         torchaudio.set_audio_backend(args.audio_backend)
 
     use_cuda = not args.no_cuda and torch.cuda.is_available()
     device = torch.device("cuda" if use_cuda else "cpu")
-
+    print("Using ", device)
     # parsing the output dict
     aggregate_dict = None if args.aggregate is None else json.loads(args.aggregate)
 
@@ -173,7 +172,7 @@ def separate():
             else:
                 outdir = Path(Path(input_file).stem + "_" + model_path.stem)
         else:
-            outdir = Path(args.outdir)
+            outdir = Path(args.outdir) / Path(input_file).stem
         outdir.mkdir(exist_ok=True, parents=True)
 
         # write out estimates

openunmix/data.py

@@ -7,7 +7,6 @@
 import torch.utils.data
 import torchaudio
 import tqdm
-from torchaudio.datasets.utils import bg_iterator
 
 
 def load_info(path: str) -> dict:
@@ -941,7 +940,6 @@ def __len__(self):
 
     args, _ = parser.parse_known_args()
 
-    torchaudio.USE_SOUNDFILE_LEGACY_INTERFACE = False
     torchaudio.set_audio_backend(args.audio_backend)
 
     train_dataset, valid_dataset, args = load_datasets(parser, args)
@@ -970,6 +968,5 @@ def __len__(self):
         num_workers=4,
     )
 
-    train_sampler = bg_iterator(train_sampler, 4)
     for x, y in tqdm.tqdm(train_sampler):
-        pass
+        print(x.shape)

openunmix/model.py

@@ -124,8 +124,8 @@ def forward(self, x: Tensor) -> Tensor:
         # crop
         x = x[..., : self.nb_bins]
         # shift and scale input to mean=0 std=1 (across all bins)
-        x += self.input_mean
-        x *= self.input_scale
+        x = x + self.input_mean
+        x = x * self.input_scale
 
         # to (nb_frames*nb_samples, nb_channels*nb_bins)
         # and encode to (nb_frames*nb_samples, hidden_size)

openunmix/transforms.py

@@ -96,7 +96,7 @@ def forward(self, x: Tensor) -> Tensor:
         # pack batch
         x = x.view(-1, shape[-1])
 
-        stft_f = torch.stft(
+        complex_stft = torch.stft(
             x,
             n_fft=self.n_fft,
             hop_length=self.n_hop,
@@ -105,8 +105,9 @@ def forward(self, x: Tensor) -> Tensor:
             normalized=False,
             onesided=True,
             pad_mode="reflect",
+            return_complex=True,
         )
-
+        stft_f = torch.view_as_real(complex_stft)
         # unpack batch
         stft_f = stft_f.view(shape[:-1] + stft_f.shape[-3:])
         return stft_f
@@ -158,7 +159,7 @@ def forward(self, X: Tensor, length: Optional[int] = None) -> Tensor:
         X = X.reshape(-1, shape[-3], shape[-2], shape[-1])
 
         y = torch.istft(
-            X,
+            torch.view_as_complex(X),
             n_fft=self.n_fft,
             hop_length=self.n_hop,
             window=self.window,

openunmix/utils.py

@@ -295,7 +295,7 @@ def preprocess(
         audio = torch.repeat_interleave(audio, 2, dim=1)
 
     if rate != model_rate:
-        print("resampling")
+        warnings.warn("resample to model sample rate")
         # we have to resample to model samplerate if needed
         # this makes sure we resample input only once
         resampler = torchaudio.transforms.Resample(

scripts/train.py

@@ -202,7 +202,6 @@ def main():
 
     args, _ = parser.parse_known_args()
 
-    torchaudio.USE_SOUNDFILE_LEGACY_INTERFACE = False
     torchaudio.set_audio_backend(args.audio_backend)
     use_cuda = not args.no_cuda and torch.cuda.is_available()
     print("Using GPU:", use_cuda)

setup.py

@@ -1,6 +1,6 @@
 from setuptools import setup, find_packages
 
-umx_version = "1.1.0"
+umx_version = "1.1.1"
 
 with open("README.md", encoding="utf-8") as fh:
     long_description = fh.read()
@@ -18,17 +18,18 @@
     python_requires=">=3.6",
     install_requires=[
         "numpy",
-        "torchaudio>=0.7.0",
-        "torch>=1.7.0",
+        "torchaudio>=0.8.0",
+        "torch>=1.8.0",
     ],
     extras_require={
         "asteroid": ["asteroid-filterbanks>=0.3.2"],
         "tests": [
             "pytest",
             "musdb>=0.4.0",
             "museval>=0.4.0",
-            "onnx",
             "asteroid-filterbanks>=0.3.2",
+            "onnx",
+            "tqdm",
         ],
         "stempeg": ["stempeg"],
         "evaluation": ["musdb>=0.4.0", "museval>=0.4.0"],

tests/cli_test.sh

@@ -2,3 +2,4 @@ python -m pip install -e .['stempeg'] --quiet
 
 # run umx on url
 coverage run -a `which umx` https://samples.ffmpeg.org/A-codecs/wavpcm/test-96.wav --audio-backend stempeg
+coverage run -a `which umx` https://samples.ffmpeg.org/A-codecs/wavpcm/test-96.wav --audio-backend stempeg --outdir out --niter 0

tests/test_datasets.py

@@ -17,7 +17,6 @@ def test_musdb():
 
 
 def test_trackfolder_fix(torch_backend):
-    torchaudio.USE_SOUNDFILE_LEGACY_INTERFACE = False
     torchaudio.set_audio_backend(torch_backend)
 
     train_dataset = data.FixedSourcesTrackFolderDataset(
@@ -33,7 +32,6 @@ def test_trackfolder_fix(torch_backend):
 
 
 def test_trackfolder_var(torch_backend):
-    torchaudio.USE_SOUNDFILE_LEGACY_INTERFACE = False
     torchaudio.set_audio_backend(torch_backend)
 
     train_dataset = data.VariableSourcesTrackFolderDataset(
@@ -48,7 +46,6 @@ def test_trackfolder_var(torch_backend):
 
 
 def test_sourcefolder(torch_backend):
-    torchaudio.USE_SOUNDFILE_LEGACY_INTERFACE = False
     torchaudio.set_audio_backend(torch_backend)
 
     train_dataset = data.SourceFolderDataset(

tests/test_io.py

@@ -24,7 +24,6 @@ def dur(request):
 
 @pytest.fixture(params=[True, False])
 def info(request, torch_backend):
-    torchaudio.USE_SOUNDFILE_LEGACY_INTERFACE = False
     torchaudio.set_audio_backend(torch_backend)
 
     if request.param:
@@ -34,7 +33,6 @@ def info(request, torch_backend):
 
 
 def test_loadwav(dur, info, torch_backend):
-    torchaudio.USE_SOUNDFILE_LEGACY_INTERFACE = False
     torchaudio.set_audio_backend(torch_backend)
     audio, _ = data.load_audio(audio_path, dur=dur, info=info)
     rate = 8000.0

tests/test_regression.py

@@ -91,4 +91,4 @@ def test_spectrogram(mus, method):
     ref = torch.load(spec_path)
     dut = encoder(audio).permute(3, 0, 1, 2)
 
-    assert torch.allclose(ref, dut, atol=1e-4, rtol=1e-5)
+    assert torch.allclose(ref, dut, atol=1e-4, rtol=1e-3)