https://github.com/thepushkarp/birmd
i don't know i'm just a birmd
https://github.com/thepushkarp/birmd
birmd recurrent-neural-networks rnn text-generation
Last synced: over 1 year ago
JSON representation
i don't know i'm just a birmd
- Host: GitHub
- URL: https://github.com/thepushkarp/birmd
- Owner: thepushkarp
- License: mit
- Archived: true
- Created: 2021-03-24T15:17:18.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2023-04-10T19:34:21.000Z (over 3 years ago)
- Last Synced: 2025-02-13T14:17:11.820Z (over 1 year ago)
- Topics: birmd, recurrent-neural-networks, rnn, text-generation
- Language: Jupyter Notebook
- Homepage:
- Size: 4.01 MB
- Stars: 5
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
birmd
i don't know i'm just a birmd
---
## What is this?
This is a project was originally built as an internal joke, but RNNs turned out to be much more enjoyable through it. This project uses Character-based RNNs using the awesome [textgenrnn](https://github.com/minimaxir/textgenrnn) module, trained on the common names of around 11k birds to generate a new bird name.
Wow such awesome!
Creating and training RNN models is fairly easy with textgenrnn. The architecture used here is 3-layered character-level bidirectional RNN with 128 LSTM cells in each layer. It was trained using GPUs on [this Google Colab](https://colab.research.google.com/drive/1TbHIok9WCVlTznDwtsDaJM1vpjUXK5bT?usp=sharing). Then, the weights were saved locally and are used to generate a new bird name on each new request.
## Dataset
The datatset used for creating the training examples is scraped from [List of birds by common names](https://en.wikipedia.org/wiki/List_of_birds_by_common_name).
## Running
This only works with Python 3.6 as of now.
## Notes
If you want to be surprised by what RNNs can do, do check out Andrej Karapathy's blog on [The Unreasonable Effectiveness of Recurrent Neural Networks](https://karpathy.github.io/2015/05/21/rnn-effectiveness/). It's a really great read.
## Thanks
Thanks to [Ashutosh Singh](https://github.com/thecodepapaya) for being an inspiration and a birmd figure for this project. And [Harsh Kakani](https://github.com/Harshkakani) for his constant support in making sure Ashutosh remains a birmd figure.