Finetuning Llama2 7b on 1000 samples from the Guanaco dataset, using QLoRA - quantization + low rank adaptation.

The Llama2 model from Meta has several variants. The smallest one containing 7 billion parameters has been finetuned here.