Load audio in ASR

[talhaanwarch]

Hi, I am following the documentation of ASR wave2vec tuning here and here
In the documentation is written

We can see that the audio file has automatically been loaded. This is thanks to the new "Audio" feature introduced in datasets == 4.13.3, which loads and resamples audio files on-the-fly upon calling.

transformers.version: '4.12.5'
datasets.version:'2.1.0'

This is the code

from datasets import load_dataset
data_files = {"train": "../train.tsv", "val": "../dev.tsv"}
urdata = load_dataset("csv", delimiter='\t',data_files=data_files)

def relapth(batch):
    return {"path":'/'.join(['../clips',batch['path']]) }
urdata=urdata.map(relapth)

Now how can i load audio? In tutorial/docs, the audio is automatically uploaded. But in my case its not. How can i do that .
Currently i am doing

import librosa
def loadaudio(batch):
    return {"audio":librosa.load(batch['path'],sr=16000) }
urdata=urdata.map(loadaudio)

[mariosasko]

Hi! You can get an audio column from your column of audio paths as follows:

urdata = urdata.cast_column("path", datasets.Audio(sampling_rate=16000))