Fixes for QwQ llm engine

sia/llm_engine/qwq_llm_engine.py

@@ -47,7 +47,7 @@ class QwQLlmEngine(LlmEngine):
 
         self._tokenizer = AutoTokenizer.from_pretrained(
             model_path,
-            tokenizer_config=tokenizer_config,
+            **tokenizer_config,
         )
 
         self._pipeline = pipeline(
@@ -87,7 +87,7 @@ class QwQLlmEngine(LlmEngine):
             tokenize=False,
             add_generation_prompt=False,
         )
-            
+
         streamer = TextIteratorStreamer(
             self._tokenizer,
             skip_prompt=True,

system_prompt.md

@@ -88,4 +88,8 @@ Use scripts and background processes to do so.
 
 The user may take some time to respond or may forget to respond.
 Keep detailed notes of your interactions and your expectations regarding time!
-Avoid overflowing the user with many messages.
+Avoid overflowing the user with many messages.
+
+# Action schema
+
+What follows is the xml schema you should adhere to:

tools/train/train/qwq.py

@@ -79,7 +79,7 @@ def main():
 
     tokenizer = AutoTokenizer.from_pretrained(
         args.base_model,
-        tokenizer_config=tokenizer_config,
+        **tokenizer_config,
     )
 
     model, _returned_tokenizer = FastLanguageModel.from_pretrained(