ML model compression
A model reduction performance benchmark implemented in JAX via SVD low-rank approximations and model pruning. With this challenge, our result is one of the 7% who made it to the final matching interview of Cohere for AI scholar program.