fix cpu

app.py

@@ -94,7 +94,6 @@ class RepLlamaModel:
         model.eval()
         return model
 
-    @spaces.GPU
     def encode(self, texts, batch_size=32, **kwargs):
         self.model = self.model.cuda()
         all_embeddings = []
@@ -181,6 +180,7 @@ def evaluate(qrels, results, k_values):
 
     return metrics
 
+@spaces.GPU
 def run_evaluation(dataset, postfix):
     global current_dataset, queries, model
     current_dataset = dataset