Update app.py

app.py

@@ -72,8 +72,10 @@ def reset_model(model_name, load_on_gpu, *extra_components, reset_sentence_trans
     if reset_sentence_transformer:
         global_state.sentence_transformer = SentenceTransformer('all-MiniLM-L6-v2')
         gc.collect()
-    if  load_on_gpu and not dont_cuda:
+    if not dont_cuda:
         global_state.model.to('cuda')
+    if load_on_gpu:
+        global_state.model.to('cpu')
     global_state.tokenizer = AutoTokenizer.from_pretrained(tokenizer_path, token=os.environ['hf_token'])
     gc.collect()
     if with_extra_components: