Fixes for runpod test

sia/local_llm_engine.py

@@ -72,7 +72,7 @@ class LocalLlmEngine(LlmEngine):
             text_inputs=prompt,
             do_sample=True,
             temperature=self._temperature,
-            max_new_tokens=self._token_limit,
+            max_new_tokens=self.token_limit(),
             streamer=streamer
         )
         thread = Thread(target=self._pipeline, kwargs=pipeline_kwargs)
@@ -106,4 +106,7 @@ class LocalLlmEngine(LlmEngine):
         Returns:
             int: Maximum number of tokens the model can process
         """
-        return self._pipeline.model.config.max_position_embeddings
+        if self._token_limit is not None:
+            return self._token_limit
+        else:
+            return self._pipeline.model.config.max_position_embeddings