Merge pull request #64 from sp-nitech/use_pad

Use pad instead of cat

.github/workflows/ci.yml

@@ -20,8 +20,8 @@ jobs:
             torch: 1.11.0
             torchaudio: 0.11.0
           - python: 3.11
-            torch: 2.1.2
-            torchaudio: 2.1.2
+            torch: 2.2.0
+            torchaudio: 2.2.0
 
     steps:
       - name: Clone
@@ -37,10 +37,9 @@ jobs:
         run: |
           sudo apt-get update -qq
           sudo apt-get install -qq -y libsndfile1
-          pip3 install torch==${{ matrix.torch }}+cpu torchaudio==${{ matrix.torchaudio }}+cpu -f https://download.pytorch.org/whl/torch_stable.html
 
       - name: Install
-        run: make PYTHON_VERSION=${{ matrix.python }} dev tool
+        run: make PYTHON_VERSION=${{ matrix.python }} TORCH_VERSION=${{ matrix.torch }} TORCHAUDIO_VERSION=${{ matrix.torchaudio }} PLATFORM=cpu dev tool
 
       - name: Check
         run: make check

Makefile

@@ -14,22 +14,28 @@
 # limitations under the License.                                           #
 # ------------------------------------------------------------------------ #
 
-PROJECT        := diffsptk
-PYTHON_VERSION := 3.8
-MODULE         :=
+PROJECT := diffsptk
+MODULE  :=
+
+PYTHON_VERSION     := 3.8
+TORCH_VERSION      := 1.11.0
+TORCHAUDIO_VERSION := 0.11.0
+PLATFORM           := cu113
 
 init:
 	pip install -e .
 
 dev:
 	test -d venv || python$(PYTHON_VERSION) -m venv venv; \
 	. ./venv/bin/activate; python -m pip install pip --upgrade; \
-	python -m pip install torch==1.11.0+cu113 torchaudio==0.11.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html; \
+	python -m pip install torch==$(TORCH_VERSION)+$(PLATFORM) torchaudio==$(TORCHAUDIO_VERSION)+$(PLATFORM) \
+		-f https://download.pytorch.org/whl/$(PLATFORM)/torch_stable.html; \
 	python -m pip install -e .[dev]
 
 dist:
-	. ./venv/bin/activate; python -m build --wheel; \
-	twine check dist/*
+	. ./venv/bin/activate; \
+	python -m build --wheel; \
+	python -m twine check dist/*.whl
 
 dist-clean:
 	rm -rf dist
@@ -65,7 +71,7 @@ test:
 	fi
 	[ -n "$(MODULE)" ] && module=tests/test_$(MODULE).py || module=; \
 	. ./venv/bin/activate; export PATH=tools/SPTK/bin:$$PATH; \
-		python -m pytest -s --cov=./ --cov-report=xml $$module
+	python -m pytest -s --cov=./ --cov-report=xml $$module
 
 test-clean:
 	rm -rf tests/__pycache__

README.md

@@ -3,10 +3,10 @@ diffsptk
 *diffsptk* is a differentiable version of [SPTK](https://github.com/sp-nitech/SPTK) based on the PyTorch framework.
 
 [![Latest Manual](https://img.shields.io/badge/docs-latest-blue.svg)](https://sp-nitech.github.io/diffsptk/latest/)
-[![Stable Manual](https://img.shields.io/badge/docs-stable-blue.svg)](https://sp-nitech.github.io/diffsptk/1.1.0/)
+[![Stable Manual](https://img.shields.io/badge/docs-stable-blue.svg)](https://sp-nitech.github.io/diffsptk/1.2.0/)
 [![Downloads](https://static.pepy.tech/badge/diffsptk)](https://pepy.tech/project/diffsptk)
 [![Python Version](https://img.shields.io/pypi/pyversions/diffsptk.svg)](https://pypi.python.org/pypi/diffsptk)
-[![PyTorch Version](https://img.shields.io/badge/pytorch-1.11.0%20%7C%202.1.2-orange.svg)](https://pypi.python.org/pypi/diffsptk)
+[![PyTorch Version](https://img.shields.io/badge/pytorch-1.11.0%20%7C%202.2.0-orange.svg)](https://pypi.python.org/pypi/diffsptk)
 [![PyPI Version](https://img.shields.io/pypi/v/diffsptk.svg)](https://pypi.python.org/pypi/diffsptk)
 [![Codecov](https://codecov.io/gh/sp-nitech/diffsptk/branch/master/graph/badge.svg)](https://app.codecov.io/gh/sp-nitech/diffsptk)
 [![License](https://img.shields.io/github/license/sp-nitech/diffsptk.svg)](https://github.com/sp-nitech/diffsptk/blob/master/LICENSE)

diffsptk/core/linear_intpl.py

@@ -17,6 +17,8 @@
 import torch.nn as nn
 import torch.nn.functional as F
 
+from ..misc.utils import replicate1
+
 
 class LinearInterpolation(nn.Module):
     """Perform linear interpolation.
@@ -38,8 +40,6 @@ def __init__(self, upsampling_factor):
 
         assert 1 <= self.upsampling_factor
 
-        self.pad = nn.ReplicationPad1d((0, 1))
-
     def forward(self, x):
         """Interpolate filter coefficients.
 
@@ -76,8 +76,8 @@ def forward(self, x):
         assert x.dim() == 3, "Input must be 3D tensor"
         B, T, D = x.shape
 
-        x = x.mT
-        x = self.pad(x)
+        x = x.mT  # (B, D, T)
+        x = replicate1(x, left=False)
         x = F.interpolate(
             x,
             size=T * self.upsampling_factor + 1,

diffsptk/core/mgc2mgc.py

@@ -20,6 +20,7 @@
 import torch.nn.functional as F
 
 from ..misc.utils import cexp
+from ..misc.utils import check_size
 from ..misc.utils import clog
 from ..misc.utils import numpy_to_torch
 from .freqt import FrequencyTransform
@@ -80,6 +81,8 @@ def forward(self, c1):
             Output cepstrum.
 
         """
+        check_size(c1.size(-1), self.in_order + 1, "dimension of cepstrum")
+
         c01 = F.pad(c1[..., 1:], (1, 0))
         C1 = torch.fft.fft(c01, n=self.n_fft)
 

diffsptk/core/par2lpc.py

@@ -34,6 +34,7 @@ def __init__(self, lpc_order):
         super(ParcorCoefficientsToLinearPredictiveCoefficients, self).__init__()
 
         self.lpc_order = lpc_order
+
         assert 0 <= self.lpc_order
 
     def forward(self, k):

diffsptk/core/plp.py

@@ -19,6 +19,7 @@
 import torch.nn as nn
 
 from ..misc.utils import numpy_to_torch
+from ..misc.utils import replicate1
 from .fbank import MelFilterBankAnalysis
 from .levdur import LevinsonDurbin
 from .mgc2mgc import MelGeneralizedCepstrumToMelGeneralizedCepstrum
@@ -156,7 +157,7 @@ def forward(self, x):
         """
         y, E = self.fbank(x)
         y = (torch.exp(y) * self.equal_loudness_curve) ** self.compression_factor
-        y = torch.cat((y[..., :1], y, y[..., -1:]), dim=-1)
+        y = replicate1(y)
         y = torch.fft.hfft(y, norm="forward")[..., : self.plp_order + 1].real
         y = self.levdur(y)
         y = self.lpc2c(y)

diffsptk/core/zcross.py

@@ -17,6 +17,8 @@
 import torch
 import torch.nn as nn
 
+from ..misc.utils import replicate1
+
 
 class ZeroCrossingAnalysis(nn.Module):
     """See `this page <https://sp-nitech.github.io/sptk/latest/main/zcross.html>`_
@@ -65,7 +67,7 @@ def forward(self, x):
 
         """
         x = torch.sign(x)
-        x = torch.cat((x[..., :1], x), dim=-1)
+        x = replicate1(x, right=False)
         x = x.unfold(-1, self.frame_length + 1, self.frame_length)
         z = 0.5 * (x[..., 1:] - x[..., :-1]).abs().sum(-1)
         if self.norm:

diffsptk/misc/utils.py

@@ -74,13 +74,23 @@ def numpy_to_torch(x):
 
 
 def to_3d(x):
-    d = x.dim()
-    if d == 1:
-        y = x.view(1, 1, -1)
-    elif d == 2:
-        y = x.unsqueeze(1)
-    else:
-        y = x.view(-1, 1, x.size(-1))
+    y = x.view(-1, 1, x.size(-1))
+    return y
+
+
+def reflect(x):
+    d = x.size(-1)
+    y = x.view(-1, d)
+    y = F.pad(y, (d - 1, 0), mode="reflect")
+    y = y.view(*x.size()[:-1], -1)
+    return y
+
+
+def replicate1(x, left=True, right=True):
+    d = x.size(-1)
+    y = x.view(-1, d)
+    y = F.pad(y, (1 if left else 0, 1 if right else 0), mode="replicate")
+    y = y.view(*x.size()[:-1], -1)
     return y
 
 
@@ -177,7 +187,7 @@ def get_gamma(gamma, c):
 
 def symmetric_toeplitz(x):
     d = x.size(-1)
-    xx = torch.cat((x[..., 1:].flip(-1), x), dim=-1)
+    xx = reflect(x)
     X = xx.unfold(-1, d, 1).flip(-2)
     return X
 

diffsptk/version.py

@@ -1 +1 @@
-__version__ = "1.1.0"
+__version__ = "1.2.0"

pyproject.toml

@@ -33,7 +33,7 @@
     "torchaudio >= 0.11.0",
     "torchcrepe >= 0.0.21",
     "torchlpc >= 0.2.0",
-    "vector-quantize-pytorch >= 0.8.0",
+    "vector-quantize-pytorch >= 0.8.0, <= 1.12.12",
     "numpy",
   ]
   dynamic = ["version"]