Hello,
Unfortunately, when I’ve requested access to Llama-3.2-3B-Instruct, my access request was rejected. Is there any way to get on with the course, maybe with some extra legwork? I find the theory part fantastic (everything is accessible and to the point), but I’m kind of bummed being unable to do the practical part. Maybe I can use some other model or possibly run inference locally? But I would still have to d/l the weights somehow.
It’s just a model for convenience, so don’t worry if you use a different model, as the essence of it won’t change that much.
Also, if you want to use the same model, you can find it surprisingly easily by searching for a hub.
Does it have to be a particular type of model? Fine-trained for instructions, that kind of thing? How do I not make a mistake of choosing an incorrect model? 
I am not particularly experienced in this matter.
I think any Instruct (sometimes abbreviated as it) model will do. The Instruct model is a model that has been adjusted for chatbot-like behavior, so we can usually use the Instruct model. For smaller ones, for example, see below.
Otherwise, the Qwen 2.5 series, Gemma 2, Gemma 3, Phi, DeepSeek, etc. would also be fine. If it’s a model that is actually used in other units of the course, it’s even more reliable.