Update app.py

app.py

@@ -14,13 +14,12 @@ if hasattr(config, 'rope_scaling'):
         'factor': 8.0  # Adjust this value based on your needs
     }
 
-# Step 3: Load the tokenizer and base model using the modified configuration
 tokenizer = AutoTokenizer.from_pretrained(base_model_name)
 base_model = AutoModelForCausalLM.from_pretrained(
-    base_model_name, 
+    base_model_name,
     config=config,
-    load_in_4bit=False,  # Ensure quantization is not attempted
-    load_in_8bit=False,  # Ensure quantization is not attempted
+    torch_dtype=torch.float32,  # Ensure it loads in full precision
+    device_map="auto"  # This ensures it loads correctly on CPU if GPU is not available
 )
 
 # Step 4: Load the LoRA adapter from the local files or Hugging Face repository