3 ms·
training on MPS backend is suboptimal and really slow.
by thangngoc89 3y ago
training on MPS backend is suboptimal and really slow.
- wtallis 3y agoDo people do training on systems this small, or just inference? I could see maybe doing a little bit of fine-tuning, but certainly not from-scratch training.
- redox99 3y agoIf you mean train llama from scratch, you aren't going to train it on any single box. But even with a single 3090 you can do quite a lot with LLMs (through QLoRA and similar).
- thangngoc89 3y agoYep. Price/performance of multiple 4090s system are way better than the professional cards (Axxx). Also deep learning outside of LLM has many different usage.