tf.audio.decode_wav not working properly?

[ark626]

🐛 Bug

x, sr = tf.audio.decode_wav(tf.io.read_file(ls[i]) doesnt return the proper values.

To Reproduce

A function like below is giving me wrong values to work with.

#Waveform array from path of folder containing wav files
def audio_array(path):
  ls = glob(f'{path}/*.wav')
  adata = []
  for i in range(len(ls)):
    x, sr = tf.audio.decode_wav(tf.io.read_file(ls[i]), 1)
    print(str(x))
    x = np.array(x, dtype=np.float32)
    adata.append(x)
  return np.array(adata)

The print part gives out a really odd tensor here:
Tensor("DecodeWav:0", shape=(?, 1), dtype=float32)

Expected behavior

The print(str(x)) operation should give something like below.

...
[ 0.0000000e+00]
[ 0.0000000e+00]
[ 0.0000000e+00]], shape=(55921, 1), dtype=float32)

Environment

Collecting environment information...
PyTorch version: 1.4.0a0+7f73f1d
Is debug build: No
CUDA used to build PyTorch: 10.2

OS: Ubuntu 18.04.4 LTS
GCC version: (Ubuntu/Linaro 7.5.0-3ubuntu1~18.04) 7.5.0
CMake version: version 3.10.2

Python version: 3.6
Is CUDA available: Yes
CUDA runtime version: 10.2.89
GPU models and configuration: Could not collect
Nvidia driver version: Could not collect
cuDNN version: Probably one of the following:
/usr/lib/aarch64-linux-gnu/libcudnn.so.8.0.0
/usr/lib/aarch64-linux-gnu/libcudnn_adv_infer.so.8.0.0
/usr/lib/aarch64-linux-gnu/libcudnn_adv_train.so.8.0.0
/usr/lib/aarch64-linux-gnu/libcudnn_cnn_infer.so.8.0.0
/usr/lib/aarch64-linux-gnu/libcudnn_cnn_train.so.8.0.0
/usr/lib/aarch64-linux-gnu/libcudnn_etc.so.8.0.0
/usr/lib/aarch64-linux-gnu/libcudnn_ops_infer.so.8.0.0
/usr/lib/aarch64-linux-gnu/libcudnn_ops_train.so.8.0.0

Versions of relevant libraries:
[pip3] numpy==1.18.4
[pip3] torch==1.5.0
[pip3] torchaudio==0.6.0a0+313f4f5
[pip3] torchvision==0.6.0a0+b68adcf
[conda] Could not collect

Can be of course connected to the issues in the general setup. See issue #658

[mthrok]

Hi @ark626

Looks like your snippet is Tensorflow. Are you using torchaudio?

[ark626]

@mthrok yes sadly this script is using both, torch and tensorflow.
The original code you can see here: https://github.com/marcoppasini/MelGAN-VC/blob/master/MelGAN_VC.ipynb

[mthrok]

@ark626

Well, since the snippet you posted concerns Tensorflow and not torchaudio, please ask it in Tensorflow's support.

[vincentqb]

Based on previous comment by @mthrok, I will close this issue. Please feel free to re-open if this relates to torchaudio.

[ark626]

Just in case i ever get this issue.
Just call numpy() or eval()!!
Depending on the Tensorflow version <=1.15 use eval() >1.15 use numpy()
' x, sr = tf.audio.decode_wav(tf.io.read_file(ls[i]), 1)
#print(x.get_shape())
#x = x.eval(session=tf.compat.v1.Session())#.numpy()
x = x.numpy()
x = np.array(x).astype(dtype=np.float32)
'

This completly solves the issue.