Ask HN: Resource to learn how to train and use ML Models

In particular, you'll probably want to skip to nanoGPT (https://github.com/karpathy/nanoGPT) and then maybe if you are interested in a bit more of the theory, Zero to Hero (https://karpathy.ai/zero-to-hero.html), and his comments in one of the threads linked: https://news.ycombinator.com/item?id=34414716

Fine tuning may also be a faster and better place to start, this is a good guide for fine tuning some publicly released LLMs: https://erichartford.com/uncensored-models

Both only cover the math as it becomes relevant. I am working through fast.ai's book right now and find its pragmatic approach to DL pretty agreeable to just getting models hosted and out the door. I watched the lectures before hand, and there are several Jupyter notebooks and examples on how to get models deployed ASAP with clunky interfaces, which also might be of interest to you.

I have to say this comes across a little insulting to machine learning engineers. You’re asking for a quick snappy course that will teach what MLEs take years to learn and master. Sigh.

Nevertheless, The Andrew Ng course ML course on Coursera is a favourite among minds curious about ML.

Productionising a model is a whole different ball game and there is likely a wide range of content on the matter - it depends what you are trying to achieve, why does your live environment look like? How many users do you have? What SLOs and such if any do you have to meet?

1. Learn to run a model, checkout llama.cpp Tons of free models on huggingface.com

2. Learn to finetune a model - https://github.com/lxe/simple-llm-finetuner

3. Learn to train one. PyTorch, TensorFlow, HuggingFace libraries, etc.

Good luck.

The other option is to follow tutorials - e.g. pytorch.org/tutorials - it doesn’t get any more practical and light on theory than that.