Research Computing provides access to the HPC Cluster to submit GPU jobs for training LLMs.
What does it mean to train a LLM?
LLM training is a multi-step process with the goal of improving a model’s performance, accuracy, and alignment around specific tasks.
LLM training requires significant amounts of high-quality, relevant data as well as substantial, specialized computer hardware.
Data. Data of high quality and quantity will allow the model to better understand the statistical patterns of language and the tasks at hand, improve performance, and better align responses with human preferences and expectations.
Computing Resources. Sorting through vast amounts of data can require powerful and specialized hardware and algorithms to complete tasks within reasonable time constraints.
How do I find out more?
There is an
introduction available to HPC (High-Performance Computing).