Starting Machine Learning

I am now working on a project that is completely unrelated to last week, and I am very excited! This new project focuses on machine learning using large language models with the goal of creating an environmental chatbot. There is a lot to learn in this space, and I have already been led down a multitude of different paths. At the beginning of the week I was focusing on setting up a deep speed model using singularity, but after meeting with Ezra I have switched my focus to fine tuning a model using the chat-gpt api. The switch to using the api will hopefully allow me to create a fine-tuned model faster which is a great advantage in this pseudo prototyping phase. I hope to have a fine tuned model working by the end of next week! Singularity was a larger endeavor than I had originally anticipated. I used the singularity docs on the compute canada wiki extensively https://docs.alliancecan.ca/wiki/Singularity. I also found Ezra’s github incredibly helpful https://github.com/macdonaldezra/ml-hpc-demo speeding up the process significantly. In a few weeks when I do come back to this it will be relatively easy to pick up again as I already have a container fully set up on graham. After the prototyping phase with the chat-gpt api I will have to learn how to use deep speed and write a LLM using it, which will no doubt be a large task. I look forward to learning more in this space, and have been greatly enjoying it so far!

-Nick