Increase num_inference_steps in generate_item_image function to improve image quality

app.py

@@ -72,7 +72,7 @@ def generate_item_image(object_t2i_prompt):
 		width=1024,
 		height=1024,
 		guidance_scale=3.5,
-		num_inference_steps=8,
+		num_inference_steps=28,
 		api_name="/infer"
     )[0]
     image = Image.open(img_path)