Update app.py

app.py

@@ -77,7 +77,7 @@ if pc:
             dense_vector = encode_query(model, query_text)
             input_ids = tokenizer(query_text, return_tensors='pt')
             with torch.no_grad():
-                sparse_vector = sparse_model(d_kwargs=input_ids.to(device))['d_rep'].squeeze()
+                sparse_vector = sparse_model(d_kwargs=input_ids.to('cpu'))['d_rep'].squeeze()
 
             # Prepare sparse vector format for Pinecone
             indices = sparse_vector.nonzero().squeeze().cpu().tolist()