Comment on QLoRA: Efficient Finetuning of Quantized LLMsparentComments−redox993y3090 can handle the ~30B models quantized to 4 bits.
Comments
3090 can handle the ~30B models quantized to 4 bits.