Implementing text generation using only convolution layers with dilation and residual layers (ByteNet decoder Link to paper) in order to predict the next character with linear time.

1. ByteNet decoder w/o embedding layer - residual_dilated_conv_text_generation
2. ByteNet decoder with embedding layer - residual_dilated_conv_text_generation_embedding
3. ByteNet decoder predicting n next characters instead of just 1 - residual_dilated_conv_n_chars_text_generation
4. ByteNet decoder for the last n characters of the string concatenated with embedding of the whole sentence - wide_view_residual_dilated_conv_text_generation

nietzsche.txt from