正则表达式翻译asr9000

class LLMWrapper: def __init__(self, model_path): self.tokenizer = AutoTokenizer.from_pretrained(model_path) self.model = AutoModelForCausalLM.from_pretrained( model_path, device_map="auto", quantization_config=BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.float16 ), torch_dtype=torch.float16, trust_remote_code=True ) logging.info("模型加载完成") def summarize(self, text): """生成30字以内摘要""" try: prompt = f"用中文一句话总结内容，50字内：{text[:500]}" inputs = self.tokenizer(prompt, return_tensors="pt").to(self.model.device) outputs = self.model.generate( **inputs, max_new_tokens=30, temperature=0.3, repetition_penalty=1.2 ) summary = self.tokenizer.decode(outputs[0], skip_special_tokens=True) return re.sub(r'[^\u4e00-\u9fff。，！？]', '', summary)[:30] except Exception as e: logging.error(f"摘要生成异常: {str(e)}") return "（生成失败）" def process_file(filepath, asr_pipeline, llm): """处理单个音频文件""" file_info = { "文件名": filepath.name, "时长": 0, "文本内容": "", "内容摘要": "", "处理状态": "待处理", "耗时": 0 } try: start_time = time.time() # 文件格式验证 if filepath.suffix.lower() != '.wav': raise ValueError("仅支持.wav格式文件") # 获取音频时长 with sf.SoundFile(str(filepath)) as f: file_info["时长"] = round(f.frames / f.samplerate, 2) # 语音识别 asr_result = asr_pipeline.generate(input=str(filepath)) text = asr_result[0]["text"] if asr_result else "" file_info["文本内容"] = text # 生成摘要 if text: file_info["内容摘要"] = llm.summarize(text) file_info.update({ "处理状态": "成功", "耗时": round(time.time() - start_time, 1) }) except Exception as e: file_info.update({ "处理状态": f"失败: {str(e)}", "耗时": round(time.time() - start_time, 1) }) return file_info以上是核心代码 生成文字里，这个内容摘要偏差太大了，存在较大错漏，怎么调整？即没有一句话总结内容，又将提问也加进了摘要里 内容摘要 用中文一句话总结内容，字内论文卡在文献综述部分怎么办试试先做 用中文一句话总结内容，字内这个智能眼镜能实时翻译外语对还能叠 用中文一句话总结内容，字内听说雨衣甘蓝要浊水去草酸冷榨橙汁反