Lots of VRAM, but the method is the same. Here's someone who finetuned llama-30b on alpaca dataset for example: https://github.com/deep-diver/Alpaca-LoRA-Serve

Lots of VRAM, but the method is the same. Here's someone who finetuned llama-30b on alpaca dataset for example: https://github.com/deep-diver/Alpaca-LoRA-Serve https://github.com/deep-diver/Alpaca-LoRA-Serve