Recurrent Neural Networks
This project aims at successfully generating fake
tv scripts using recurrent neural networks
A quick introduction of the minimal setup you need to get the classifier up & running.
git clone https://github.com/abhigyan2311/tv-script-generator.git
cd tv-script-generator
conda env create -f environment.yml
conda activate deeplearning
jupyter notebook dlnd_tv_script_generation.ipynb
The classifier notebook implements the following methods:
- Successfully generates fake tv scripts based on Seinfeld tv series scripts
The best minimum loss was achieved of 3.19.
- Sequence Length - 200
- Batch Size - 128
- Number of epochs - 10
- Learning Rate - 0.001
- Embedding Dimension - 400
- Hidden Dimension - 256
- Number of RNN Layers - 2
"If you'd like to contribute, please fork the repository and use a feature branch. Pull requests are warmly welcome."
"The code in this project is licensed under Apache 2.0 license."