hopefully fixed qwen

utils/models.py

@@ -23,7 +23,7 @@ models = {
     "Qwen3-0.6b": "qwen/qwen3-0.6b",
     "Qwen3-1.7b": "qwen/qwen3-1.7b",
     "Qwen3-4b": "qwen/qwen3-4b",
-    "SmolLM2-1.7b-Instruct": "huggingfacetb/smolllm2-1.7b-instruct",
+    "SmolLM2-1.7b-Instruct": "HuggingFaceTB/SmolLM2-1.7B-Instruct",
     "EXAONE-3.5-2.4B-instruct": "LGAI-EXAONE/EXAONE-3.5-2.4B-Instruct",
     "OLMo-2-1B-Instruct": "allenai/OLMo-2-0425-1B-Instruct",
 
@@ -107,7 +107,7 @@ def run_inference(model_name, context, question):
         model_kwargs["enable_thinking"] = False
 
     try:
-        tokenizer = AutoTokenizer.from_pretrained(model_name, padding_side="left", token=True)
+        tokenizer = AutoTokenizer.from_pretrained(model_name, padding_side="left", token=True, kwargs=model_kwargs)
         accepts_sys = (
             "System role not supported" not in tokenizer.chat_template
             if tokenizer.chat_template else False # Handle missing chat_template
@@ -125,10 +125,6 @@ def run_inference(model_name, context, question):
             model=model_name,
             tokenizer=tokenizer,
             device_map='auto',
-            do_sample=True,
-            temperature=0.6,
-            top_p=0.9,
-            model_kwargs=model_kwargs,
             trust_remote_code=True,
         )