MSc thesis project proposal

[2023] Build your own Mini-ChatGPT on Jetson Nano

The RWKV RNN-LM (Recurrent Weighted Key-Value Memory Network Language Model) is a novel language model with promising results in generating human-like text. This project aims to build a chatbot, like ChatGPT, using the RWKV RNN-LM as the backbone and PyTorch as the programming framework. The project will also include research on model compression techniques, such as quantization and pruning, to compress the language model for deployment on an NVIDIA Jetson Nano, a small AI computer designed for edge computing applications.

Assignment

  • Research the RWKV RNN-LM architecture and model compression techniques such as quantization and pruning.
  • Set up the Jetson Nano and install the necessary software and libraries.
  • Prepare and preprocess the training data.
  • Train the chatbot using the RWKV RNN-LM architecture and PyTorch.
  • Research and implement model compression techniques, such as quantization and pruning, to compress the language model.
  • Evaluate the performance of the compressed model and fine-tune as necessary.
  • Implement the chatbot in a user-friendly interface.

Requirements

  • Experience in Python and PyTorch coding.
  • Basic understanding of deep learning.
  • Knowledge & experience in embedded systems

Contact

dr. Chang Gao

Electronic Circuits and Architectures Group

Department of Microelectronics

Last modified: 2023-11-12