remove cuda

app.py

@@ -35,7 +35,7 @@ def generate2(F=None, maxTokens=256):
     input_ids = tokenizer(prompt,
                           return_tensors="pt",
                           truncation=True,
-                          add_special_tokens=False).input_ids.cuda()
+                          add_special_tokens=False).input_ids
     outputs = model.generate(
         input_ids=input_ids,
         max_new_tokens=maxTokens,
@@ -77,4 +77,4 @@ iface = gr.Interface(
 )
 
 if __name__ == "__main__":
-    iface.launch()
+    iface.launch(share=True)