Add a section to the tutorial to train and evaluate a simple NN (efficientnet). Including visualization of audio input, spectrogram, NN output and transformation of output to class name.
This involves the following steps:
Load BaseModule with model parameter as efficientnet.py
for this, the hydra.initiate call needs to be removed and the model parameter should be typed as an object (see how it is done in the GADMEDatamodule)
train using lightning trainer
eval using lightning trainer
use trained model for prediction of a test sample
load a single sample from test dataset
visualize waveform, make audio listenable in notebook
convert to spectrogram using test dataloader (make sure its the same sample, could be done with disabling random sampling of the dataloader ?)
Add a section to the tutorial to train and evaluate a simple NN (efficientnet). Including visualization of audio input, spectrogram, NN output and transformation of output to class name.
This involves the following steps:
BaseModule
withmodel
parameter asefficientnet.py
hydra.initiate
call needs to be removed and the model parameter should be typed as an object (see how it is done in theGADMEDatamodule
)