Hi, can someone please advise me what RAM and GPU requirements the Llama2 model has with 7 billion parameters? If I take the model as it is, without further fine tuning like Lora or quantization. Conversely, what would be the requirements if I used Lora, quantization or both.