Update app.py

app.py

@@ -45,22 +45,28 @@ def load_model_gpu():
     return model, tokenizer
 
 def load_model():
-    # Load in pure CPU mode without quantization
     config = PeftConfig.from_pretrained("satishpednekar/sbx-qhelper-mistral-loraWeights")
     
     model = AutoModelForCausalLM.from_pretrained(
         config.base_model_name_or_path,
-        torch_dtype=torch.float32,  # Use float32 for CPU
-        device_map=None,  # Force CPU
+        torch_dtype=torch.float32,
+        device_map=None,
         trust_remote_code=True,
-        load_in_4bit=False  # Remove quantization
+        # Remove all quantization-related parameters
     )
     
-    model = PeftModel.from_pretrained(model, "satishpednekar/sbx-qhelper-mistral-loraWeights")
-    tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
+    model = PeftModel.from_pretrained(
+        model, 
+        "satishpednekar/sbx-qhelper-mistral-loraWeights",
+        torch_dtype=torch.float32
+    )
+    
+    tokenizer = AutoTokenizer.from_pretrained(
+        config.base_model_name_or_path,
+        trust_remote_code=True
+    )
     
-    # Explicitly move to CPU
-    model = model.to("cpu")
+    model = model.to("cpu").eval()
     
     return model, tokenizer