Easy audio category with torch

This short article equates Daniel Falbel‘s ‘ Easy Audio Category’ short article from tensorflow/keras to torch/torchaudio The primary objective is to present torchaudio and show its contributions to the torch environment. Here, we concentrate on a popular dataset, the audio loader and the spectrogram transformer. A fascinating side item is the parallel in between torch and tensorflow, revealing in some cases the distinctions, in some cases the resemblances in between them.

Downloading and Importing

torchaudio has the speechcommand_dataset integrated in. It strains background_noise by default and lets us pick in between variations v0.01 and v0.02

 # set an existing folder here to cache the dataset
 DATASETS_PATH <%  choose

(
 class ,   pred_class)
, freq   = x
$ n , col 
 = ifelse (
 x

$
 right, " lightblue"
, [1]" red"

)
, border  = ifelse
([1] x

$
 right ,  " lightblue"[1]

, " red")[, 1:10]

,[ CPUFloatType{1,10} ]

 alpha  = 0.6

,

 conceal  = x

$



 n<