Optimized hot-loop for logits processor

sia/llm_engine/qwq_llm_engine.py

@@ -2,12 +2,12 @@ from pathlib import Path
 from threading import Thread
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline, TextIteratorStreamer, BitsAndBytesConfig
 from typing import Callable, Iterator, Optional, List
+from xml_schema_validator import XmlLogitsProcessor
 import json
 import torch
 
 from . import LlmEngine
 from .. import util
-from .xml_logits_processor import XmlLogitsProcessor
 
 class QwQLlmEngine(LlmEngine):