Inspiration
Bringing human learning to AI - starting from the basics and building up to the foundations. A student in every sense of the word.
What it does
Solves maths problems
How we built it
Mixing the latest architectural design choices and research findings with data manipulation techniques.
Challenges we ran into
Using torch.compile, optimising speed, quantization.
Accomplishments that we're proud of
Managing to get our idea of corpus reshuffling into practice.
What we learned
It takes time to find the right parameters and techniques for a good model.
What's next for Corpus
More training time, quantization aware training without significant loss of performance.
Log in or sign up for Devpost to join the conversation.