0

The example of tensorflow is https://www.tensorflow.org/tutorials/audio/simple_audio

I want to know the file_path in this example as below:

def get_waveform_and_label(file_path):
  label = get_label(file_path)
  audio_binary = tf.io.read_file(file_path)
  waveform = decode_audio(audio_binary)
  return waveform, label

The 'file_path' is called in the function as below:

waveform_ds = files_ds.map(get_waveform_and_label, num_parallel_calls=AUTOTUNE)

I tried to get the value of the file_path using the print(file_path), but the answer is Tensor("args_0:0", shape=(), dtype=string).

That is not the exact value of file_path.

Finally, I don't know the value of the file_path and input parameters of get_waveform_and_label in the waveform_ds = files_ds.map(get_waveform_and_label, num_parallel_calls=AUTOTUNE).

Jungsik
  • 1
  • 1

1 Answers1

-1

I have been stuck in the same problem, but I think I got it:

The file_path comes from the file_ds variable, where file_ds "represents potentially a large set of element", where each element is the 'file path' to each of the wav. file.

After you set the train_files variable, run the following to see what I mean.

file_ds = tf.data.Dataset.from_tensor_slices(train_files)
for element in dataset:
  print(element)

Then when you use the map function

file_ds.map(get_waveform_and_label)

It's applyting the function get_waveform_and_label to each of the elements of the large set, in this case to each wav file.

references: https://www.tensorflow.org/api_docs/python/tf/data/Dataset

rensc
  • 1
  • 2
  • Thank you for your answering but I still wonder how to print `file_path` in `def get_waveform_and_label(file_path)`. – Jungsik May 24 '21 at 04:48