Comment on QLoRA: Efficient Finetuning of Quantized LLMsparentComments−MacsHeadroom3y24GB can fit 33B parameter models in 4bit. You only need 4GB to run 7B models.
Comments
24GB can fit 33B parameter models in 4bit. You only need 4GB to run 7B models.