GitHub repository for AI Hardware at PSU ECE 510
code/ Contains all code fest and weekly challenges code. Specifically, code fest code that does not relate to the final project or is tertiary.
doc/ Contains all documents for relavent texts on communication with my LLM, needed documentation, papers for the project, or animations.
proj/ Contains all relevant code to the final project!
Here you can find my blog Challenge blogs! This covers all of the weekly challenges including a running blog of my project. This is helpful because our code fest challenges specifically deal with the project. So seeing how code related that week is helpful to understand how the code changes interact!
Forthe project you can go to this link to see how project methodologies are going).
I am basing my project on the Nerual Radiance Fields (NeRF) methodology where the paper is seen here.
The attempt of this project is to make it more available to academic and education purposes with edge devices so I'm targetting acceptable speed for something like an occulus while maintaining acceptable power and data integrity. NeRF allows us to reconstructor multiple sided 2D images into a 3D image that can be rotated and expanded and what presented in 2020 by Google researchers. This could open major possibilities in visualization education in chemstry, mathematics, physics, history, and other subjects where visual information can help improve understanding. My goal is to use an analog simulation results to prove out the possibility of an analog solution for a NeRF DSP acceleration chiplet.
For this project you need LTSpice, Python, and a python venv
Creation of a venv can be found at this link
Python libraries used are: pytorch, numpy, matplotlib, and pandas
LTSpice can be run directly by loading the netlist into the simulator and clicking run. You can feed the log files for the latency tests into your simulator of your choice to gain feedback and analysis of the results. Using the graphing function in LTSpice can be found on the LTSpice documentation inside the application.