MacBook Air M1 for AI

[omarabb315]

is the MacBook Air M1 with 36 GB Unified-RAM Good for state-of-the-art finetuning ?

[awni]

I'm not sure what you mean by "state-of-the-art" fine-tuning.

If you are asking if you can fine-tune an LLM, yes definitely with a low-rank method and optionally a quantized model. You could for example fine-tune small to medium sized models (maybe up to 7b with QLoRA). It also depends on how big of a dataset you need to fine-tune on. If it's pretty small (like a few hundred samples, it should work fine and not take too long).

Check out fine-tuning in MLX LM for more info on that.

[omarabb315]

Thanks a lot, yes I meant that finetuning tasks,
Is there any benchmarks for its speed?