Update model temperature setting

app.py

@@ -14,7 +14,7 @@ from utils import COMMUNITY_POSTFIX_URL, get_model_config, check_format, models_
     logged_event_handler, DEBUG_MODE, DEBUG_MODEL, log_debug, log_info, log_error, log_warning
 from log_chat import log_chat
 
-MODEL_TEMPERATURE = 0.8
+DEFAULT_MODEL_TEMPERATURE = 0.6
 BUTTON_WIDTH = 160
 DEFAULT_OPT_OUT_VALUE = DEBUG_MODE
 
@@ -114,10 +114,11 @@ def run_chat_inference(history, message, state):
     state["stop_flag"] = False
     error = None
     model_name = model_config.get('MODEL_NAME')
+    temperature = model_config.get('TEMPERATURE', DEFAULT_MODEL_TEMPERATURE)
 
     # Reinitialize the OpenAI client with a random endpoint from the list
     setup_model(model_config.get('MODEL_KEY'))
-    log_info(f"Using model {model_name} with endpoint {model_config.get('base_url')}")
+    log_info(f"Using model {model_name} (temperature: {temperature}) with endpoint {model_config.get('base_url')}")
 
     if len(history) == 0:
         state["chat_id"] = uuid4().hex
@@ -372,7 +373,7 @@ def run_chat_inference(history, message, state):
             stream = openai_client.chat.completions.create(
                 model=model_name,
                 messages=api_messages,
-                temperature=MODEL_TEMPERATURE,
+                temperature=temperature,
                 stream=True
             )
         except Exception as e:
@@ -387,7 +388,7 @@ def run_chat_inference(history, message, state):
                          model_name=model_name,
                          prompt=message,
                          history=history,
-                         info={"is_reasoning": model_config.get("REASONING"), "temperature": MODEL_TEMPERATURE,
+                         info={"is_reasoning": model_config.get("REASONING"), "temperature": temperature,
                                "stopped": True, "error": str(e)},
                          )
             else:
@@ -477,7 +478,7 @@ def run_chat_inference(history, message, state):
                          model_name=model_name,
                          prompt=message,
                          history=history,
-                         info={"is_reasoning": model_config.get("REASONING"), "temperature": MODEL_TEMPERATURE,
+                         info={"is_reasoning": model_config.get("REASONING"), "temperature": temperature,
                                "stopped": state["stop_flag"]},
                          )
 

utils.py

@@ -20,7 +20,8 @@ models_config = {
         "VLLM_API_URL_LIST": os.environ.get("VLLM_API_URL_LIST_APRIEL_1_5_15B"),
         "AUTH_TOKEN": os.environ.get("AUTH_TOKEN"),
         "REASONING": True,
-        "MULTIMODAL": True
+        "MULTIMODAL": True,
+        "TEMPERATURE": 0.6
     },
     # "Apriel-Nemotron-15b-Thinker": {
     #     "MODEL_DISPLAY_NAME": "Apriel-Nemotron-15b-Thinker",