我在ollama中下载的qwen2.5的模型要进行封装

在你的项目中封装Qwen2.5模型（通常指来自通义千问的模型），可以参考以下步骤进行。Qwen模型是基于PyTorch的，因此我们可以创建一个类来管理模型的加载、推理和保存：

import torch
from transformers import QwenForQuestionAnswering, QwenTokenizer

class QwenModelWrapper:
    def __init__(self, model_path):
        self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
        self.model = QwenForQuestionAnswering.from_pretrained(model_path).to(self.device)
        self.tokenizer = QwenTokenizer.from_pretrained(model_path)

    def encode_query(self, question):
        encoding = self.tokenizer(question, return_tensors="pt", padding="max_length", truncation=True)
        encoding = {k: v.to(self.device) for k, v in encoding.items()}
        return encoding

    def answer_question(self, question):
        inputs = self.encode_query(question)
        outputs = self.model(**inputs)
        start_scores, end_scores = outputs.start_logits, outputs.end_logits
        start_index = torch.argmax(start_scores)
        end_index = torch.argmax(end_scores) + 1
        answer = self.tokenizer.convert_tokens_to_string(self.tokenizer.convert_ids_to_tokens(inputs['input_ids'][start_index:end_index]))
        return answer

# 使用示例
wrapper = QwenModelWrapper('path/to/qwen2.5')
question = "谁是美国第一位总统？"
answer = wrapper.answer_question(question)
print(f"答案是: {answer}")

记得将`'path/to/qwen2.5'`替换为你的Qwen2.5模型的实际路径。这样就封装了一个可以处理问答任务的Qwen模型。