Update app.py

1. Corrected model variable name
2. Added garbage collection after the quantization (Hoping it reduces RAM spend)

app.py

@@ -37,13 +37,13 @@ quantization_config = TorchAoConfig(quant_type=quant_config)
 MODEL_ID = "HuggingFaceTB/SmolLM3-3B"
 
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
-quantized_model = AutoModelForCausalLM.from_pretrained(
+model = AutoModelForCausalLM.from_pretrained(
     MODEL_ID,
     torch_dtype="auto",
     device_map="auto",
-    quantization_config=quantization_config
-)
+    quantization_config=quantization_config)
 
+gc.collect()
 
 #########