Exploring the background and mathematical intuition of CFMs. Implemented in MLX with a simple example.

An overview of SCG and how gradient accumulation can help optimization when using minibatches for training neural nets.