Update app.py

app.py

@@ -219,7 +219,7 @@ if QUANTIZE in {"8bit", "4bit"} and BitsAndBytesConfig is not None:
         quantization_config=bnb_config,
     )
 else:
-    # Fallback: run in fp16 without bitsandbytes
+    # Fallback / default: fp16 (no bitsandbytes required)
     base_model = AutoModelForCausalLM.from_pretrained(
         BASE_MODEL,
         torch_dtype=dtype,