1、适配昇腾处理器：Ascend310、Ascend910系列 2、开发环境：Ascend-cann-toolkit_xxx、Ascend-cann-kernels-xxx(可选)、python3.8 3、下载代码：git clone https://modelers.cn/ShanXi/llamaRAGdrama.git 4、安装依赖：pip install -r examples/requirements.txt 5、推理测试：python examples/inference.py 6、推理脚本：
import argparse import torch from openmind import pipeline, is_torch_npu_available from transformers import AutoTokenizer, AutoModelForCausalLM from openmind_hub import snapshot_download def parse_args(): parser = argparse.ArgumentParser() parser.add_argument("--model_name_or_path",type=str,help="模型路径",default="./",) args = parser.parse_args() return args
if is_torch_npu_available():
device = "npu:0"
else:
device = "cpu"
args = parse_args()
if args.model_name_or_path:
model_path = args.model_name_or_path
else:
model_path = snapshot_download('ShanXi/llamaRAGdrama',revision='main',resume_donwload=True,ignore_patterns=['*.h5','*.ot','*.msgpack'])
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(model_path)
input_text = "在此输入提示词"
input_tokens = tokenizer.encode(input_text, return_tensors='pt')
output_tokens = model.generate(input_tokens, max_length=100, num_return_sequences=1)
generated_text = tokenizer.decode(output_tokens[0], skip_special_tokens=True)
print(generated_text)