update

app.py

@@ -1,6 +1,6 @@
 # import os
 # # Set CUDA device dynamically
-# os.environ["CUDA_VISIBLE_DEVICES"] = "5"
+# os.environ["CUDA_VISIBLE_DEVICES"] = ""
 
 import spaces
 import torch
@@ -38,7 +38,8 @@ generate_kwargs = dict(
 # llama_guard, llama_guard_tokenizer, UNSAFE_TOKEN_ID = load_llama_guard("meta-llama/Llama-Guard-3-1B")
 
 ## RAG MODEL
-RAG = RAGPretrainedModel.from_index("colbert/indexes/arxiv_colbert", n_gpu=0)
+test_tensor = torch.tensor(2).to('cuda')
+RAG = RAGPretrainedModel.from_index("colbert/indexes/arxiv_colbert", n_gpu=1)
 
 try:
   gr.Info("Setting up retriever, please wait...")