Update app.py

app.py

@@ -12,8 +12,7 @@ if torch.cuda.is_available():
 else:
     device = "cpu"
     print("Using CPU")
-
-model.to(device)
+    
 
 def generate(
     precision_model,
@@ -29,6 +28,8 @@ def generate(
 
     model = T5ForConditionalGeneration.from_pretrained("roborovski/superprompt-v1", torch_dtype=precision_model)
 
+    model.to(device)
+    
     input_text = f"{system_prompt}, {prompt}"
     input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to(device)
 
@@ -53,7 +54,7 @@ def generate(
     return better_prompt
 
 
-precision_model = gr.Radio([('fp32', torch.float32), ('fp16', toch.float16)], label="Model Precision Type", info="fp32 is more precised but slower, fp16 is faster and less resource consuming but less pricse")
+precision_model = gr.Radio([('fp32', torch.float32), ('fp16', toch.float16)], value='fp16', label="Model Precision Type", info="fp32 is more precised but slower, fp16 is faster and less resource consuming but less pricse")
 
 prompt = gr.Textbox(label="Prompt", interactive=True)