improve sync

app_modules/llm_inference.py

@@ -38,7 +38,8 @@ class LLMInference(metaclass=abc.ABCMeta):
         self, inputs, streaming_handler, q: Queue = None, tracing: bool = False
     ):
         print(inputs)
-        self.llm_loader.lock.acquire()
+        if self.llm_loader.streamer.for_huggingface:
+            self.llm_loader.lock.acquire()
 
         try:
             self.llm_loader.streamer.reset(q)
@@ -67,7 +68,8 @@ class LLMInference(metaclass=abc.ABCMeta):
 
             return result
         finally:
-            self.llm_loader.lock.release()
+            if self.llm_loader.streamer.for_huggingface:
+                self.llm_loader.lock.release()
 
     def _execute_chain(self, chain, inputs, q, sh):
         q.put(chain(inputs, callbacks=[sh]))