Hi, I’m currently starting to learn how LLM works in depth, so I started using nanoGPT to understand how to train a model and I’d like to play around with the code a little more. So I set myself a goal to train a model that can write basic French, it doesn’t to be coherent or deep in its writing, just French with correct grammar. I only have a laptop that doesn’t have a proper GPU, so I can’t really train a model with billions of parameters. Do you think it’s possible without too much dataset or intensive training? Is it a better idea if I use something different from nanoGPT?
TLDR: I’d like to train my own LLM on my laptop which doesn’t have a GPU. It’s only for learning purpose, so my goal is that it can write basic French. Is it doable? If it is, do you have any tips to make this easier?
You must log in or register to comment.
If you want to learn machine learning, you could maybe play around with the examples recognizing single digit handwritten numbers with that MNIST dataset or something in that kind of league.
I think training an LLM that can be somewhat useful will be way out of scope with the RAM and computing capabilities such a laptop has to offer. Maybe correct grammer if you don’t care to wait for a long long time. Something with the level of intelligence of autocomplete. But definitely not coherent or intelligent or answering your questions.
You could rent a VM in the cloud. Services like runpod.io or vast.ai offer you a proper GPU for like $2 an hour. There is also Amazon, Google, Azure, Lambda…
Do cloud services see everything - text/images data for training? And finished trained model? If so, runpod.io etc are no solution.
What do you mean? You rent some computing time on a big computer with many GPUs. That computer is somewhere in the ‘cloud’ aka it’s not your computer. There is no one sitting in front of that computer and looking at all of your pictures. But your trainig data and models obviously need to be stored intermittently on that server. So they’re in theory accessible for their admins.
The correct way to solve this is to have a look at the contract. Runpod for example has 2 different levels of security/privacy available. And there are different cloud providers with different levels of service. From ‘we share your pictures but the service is free’ to ‘you can put your customers credit card details on this server or use it for sensitive files and nobody except the police can raid us’. Runpod doesn’t offer that service. But I don’t really get your question.
