This PR adds a prewarm() method to the DSP base class that gets called from get_dsp().
The default behavior is to process the model over "_prewarm_samples" samples. By default, this is 0 - so nothing happens for LSTM (which is what we want).
WaveNet and ConvNet both set "_prewarm_samples" to the size of the model's receptive field.
@sdatkinson I don't use ConvNet, so I'm not sure I got the receptive field correct - can you make sure to check it?
This PR adds a prewarm() method to the DSP base class that gets called from get_dsp().
The default behavior is to process the model over "_prewarm_samples" samples. By default, this is 0 - so nothing happens for LSTM (which is what we want).
WaveNet and ConvNet both set "_prewarm_samples" to the size of the model's receptive field.
@sdatkinson I don't use ConvNet, so I'm not sure I got the receptive field correct - can you make sure to check it?
Resolves https://github.com/sdatkinson/NeuralAmpModelerCore/issues/89 Resolves https://github.com/sdatkinson/NeuralAmpModelerCore/issues/61