fix sth

audio/audiotools/__init__.py

@@ -1,12 +1,12 @@
-__version__ = "0.0.1"
-from .core import AudioSignal
-from .core import STFTParams
-from .core import Meter
-from .core import util
-from .core import highpass_filter, highpass_filters
-from . import metrics
 from . import data
+from . import metrics
 from . import ml
 from . import post
+from .core import AudioSignal
+from .core import highpass_filter
+from .core import highpass_filters
+from .core import Meter
+from .core import STFTParams
+from .core import util
 from .data import datasets
 from .data import transforms

audio/audiotools/core/_julius.py

@@ -15,14 +15,14 @@ from typing import Sequence
 import paddle
 import paddle.nn as nn
 import paddle.nn.functional as F
-sys.path.append("/home/aistudio/PaddleSpeech")
+
 from paddlespeech.t2s.modules import fft_conv1d
 from paddlespeech.t2s.modules import FFTConv1D
+from paddlespeech.utils import satisfy_paddle_version
 
 __all__ = [
-    'fft_conv1d', 'FFTConv1D', 'highpass_filter', 'highpass_filters',
-    'lowpass_filter', 'LowPassFilter', 'LowPassFilters', 'pure_tone',
-    'resample_frac', 'split_bands', 'SplitBands'
+    'highpass_filter', 'highpass_filters', 'lowpass_filter', 'LowPassFilter',
+    'LowPassFilters', 'pure_tone', 'resample_frac', 'split_bands', 'SplitBands'
 ]
 
 
@@ -61,6 +61,9 @@ def sinc(x: paddle.Tensor):
 
     __Warning__: the input is not multiplied by `pi`!
     """
+    if satisfy_paddle_version("2.6"):
+        return paddle.sinc(x)
+
     return paddle.where(
         x == 0,
         paddle.to_tensor(1.0, dtype=x.dtype, place=x.place),
@@ -103,7 +106,7 @@ class ResampleFrac(paddle.nn.Layer):
         >>> print(len(resample(x)))
         1250
         """
-        super(ResampleFrac, self).__init__()
+        super().__init__()
         if not isinstance(old_sr, int) or not isinstance(new_sr, int):
             raise ValueError("old_sr and new_sr should be integers")
         gcd = math.gcd(old_sr, new_sr)
@@ -257,7 +260,7 @@ class LowPassFilters(nn.Layer):
                  zeros: float=8,
                  fft: Optional[bool]=None,
                  dtype="float32"):
-        super(LowPassFilters, self).__init__()
+        super().__init__()
         self.cutoffs = list(cutoffs)
         if min(self.cutoffs) < 0:
             raise ValueError("Minimum cutoff must be larger than zero.")
@@ -325,7 +328,7 @@ class LowPassFilter(nn.Layer):
                  pad: bool=True,
                  zeros: float=8,
                  fft: Optional[bool]=None):
-        super(LowPassFilter, self).__init__()
+        super().__init__()
         self._lowpasses = LowPassFilters([cutoff], stride, pad, zeros, fft)
 
     @property
@@ -583,7 +586,7 @@ class SplitBands(paddle.nn.Layer):
             pad: bool=True,
             zeros: float=8,
             fft: Optional[bool]=None, ):
-        super(SplitBands, self).__init__()
+        super().__init__()
         if (cutoffs is None) + (n_bands is None) != 1:
             raise ValueError(
                 "You must provide either n_bands, or cutoffs, but not both.")

audio/audiotools/core/audio_signal.py

@@ -587,7 +587,7 @@ class AudioSignal(
         self.original_signal_length = self.signal_length
 
         self.sample_rate = sample_rate
-        # return self.to(device)
+
         return self
 
     def write(self, audio_path: typing.Union[str, Path]):
@@ -1198,7 +1198,6 @@ class AudioSignal(
         padding_type = self.stft_params.padding_type if padding_type is None else padding_type
 
         window = self.get_window(window_type, window_length)
-        # window = window.to(self.audio_data.device)
 
         audio_data = self.audio_data
         right_pad, pad = self.compute_stft_padding(window_length, hop_length,
@@ -1362,6 +1361,26 @@ class AudioSignal(
         paddle.Tensor [shape=(batch, channels, mels, time)]
             Mel spectrogram.
         """
+        # from paddle.audio.compliance.librosa import melspectrogram
+        # # from ..compliance.librosa import melspectrogram
+        # return melspectrogram(
+        #     x=self.audio_data,
+        #     sr=self.sample_rate,
+        #     window_size: int=512,
+        #     hop_length: int=320,
+        #     n_mels: int=64,
+        #     fmin: float=50.0,
+        #     fmax: Optional[float]=None,
+        #     window: str='hann',
+        #     center: bool=True,
+        #     pad_mode: str='reflect',
+        #     power: float=2.0,
+        #     to_db: bool=True,
+        #     ref: float=1.0,
+        #     amin: float=1e-10,
+        #     top_db: Optional[float]=None
+        # )
+
         stft = self.stft(**kwargs)
         magnitude = paddle.abs(stft)
 
@@ -1429,6 +1448,9 @@ class AudioSignal(
             MFCCs.
         """
 
+        # from paddle.audio.compliance.librosa import mfcc
+        # return mfcc(self.audio_data, self.sample_rate, n_mfcc=n_mfcc, n_mels=n_mels)
+
         mel_spectrogram = self.mel_spectrogram(n_mels, **kwargs)
         mel_spectrogram = paddle.log(mel_spectrogram + log_offset)
         dct_mat = self.get_dct(n_mfcc, n_mels, "ortho", self.device)