Update README.md

README.md

@@ -79,16 +79,16 @@ make run
 模型已经转换为huggingface的LlamaPretrainedModel格式，调用方法如下：
 
 ```python
-    torch.set_default_device("cuda")
-    config = AutoConfig.from_pretrained(model_path)
-    print(f"config: {config}")
-    tokenizer = AutoTokenizer.from_pretrained(model_path)
-    model = AutoModelForCausalLM.from_pretrained(model_path, config=config)
-    model.to("cuda")
-    inputs = tokenizer("从前", return_tensors="pt", add_special_tokens=True)
-    outputs = model.generate(**inputs, max_length=768)
-    text = tokenizer.batch_decode(outputs)[0]
-    print(text)
+torch.set_default_device("cuda")
+config = AutoConfig.from_pretrained(model_path)
+print(f"config: {config}")
+tokenizer = AutoTokenizer.from_pretrained(model_path)
+model = AutoModelForCausalLM.from_pretrained(model_path, config=config)
+model.to("cuda")
+inputs = tokenizer("从前", return_tensors="pt", add_special_tokens=True)
+outputs = model.generate(**inputs, max_length=768)
+text = tokenizer.batch_decode(outputs)[0]
+print(text)
 ```
 
 输出