🌘w🌖

app.py

@@ -21,7 +21,7 @@ tokenizer = AutoTokenizer.from_pretrained("google/gemma-7b-it", token=token)
 device = torch.device("cuda")
 model = model.to(device)
 RAG = SentenceTransformer("mixedbread-ai/mxbai-embed-large-v1")
-TOP_K = 2
+TOP_K = 1
 HEADER = "\n# RESOURCES:\n"
 # prepare data
 # since data is too big we will only select the first 3K lines