Closed Thunfischpirat closed 1 year ago
I had to adapt Conv1dModel
to make the size of the last pooling layer adaptable. Otherwise, applying pooling with window size 4
in a reduction in output length through the network layers below four before the previous max pooling layer is applied. This then causes an error.
Enable neural network models to be trained on a subset of data coming from one or more specific speakers. To this end, modify the code in
model_neural.utils.data_loading
to accommodate this functionality.