My goal was to understand the basics of Neural Networks and their development both with basic Python (Manual Backprop) and Pytorch(loss.backward()). The idea was that if I understand this then I can work with other APIs and also other architectures. This also gave me the flexibility to play around with different aspects of optimization. This project draws heavy inspiration from Andrej Karpathy's videos and tutorials. His videos/tutorials have had a significant influence on my learning journey.
Link To my Learnings: https://www.notion.so/Text-Generation-170a3381eaa1818abc01dddf053dae4e?pvs=4
- Choose the model of choice.
- Add any dataset. I used the compliment dataset(https://github.com/ravsodhi/jibes-and-delights/blob/main/raw/FreeCompliments.csv) but any txt file should work.
- The files can be run using simple python .
- Example: python lstm_train.py