Training parameters are a pivotal component in the design and performance of LLMs. Choosing optimal size of training parameters depends on data availability, use case, compute resource.