_target_: transformers.LlamaForCausalLM.from_pretrained
pretrained_model_name_or_path: AILab-CVC/seed-llama-8b-sft
load_in_8bit: True
# device_map: auto
low_cpu_mem_usage: True