Update app.py

app.py

@@ -1,6 +1,6 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
-from transformers import RagTokenizer, RagRetriever
+from transformers import RagRetriever, RagTokenizer
 
 """
 For more information on huggingface_hub Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
@@ -8,10 +8,8 @@ For more information on huggingface_hub Inference API support, please check the
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
 # Load the RAG tokenizer and retriever
-rag_tokenizer = RagTokenizer.from_pretrained("facebook/rag-sequence-nq")
-rag_retriever = RagRetriever.from_pretrained(
-    "facebook/rag-sequence-nq", index_name="apexcustoms.pdf", use_dummy_dataset=False, trust_remote_code=True
-)
+tokenizer = RagTokenizer.from_pretrained("facebook/rag-token-nq")
+retriever = RagRetriever.from_pretrained("facebook/rag-token-nq", index_name="apexcustoms", passages="apexcustoms.pdf")
 
 def respond(
     message,
@@ -39,9 +37,8 @@ def respond(
         stream=True,
         temperature=temperature,
         top_p=top_p,
-        # Pass the RAG tokenizer and retriever to the client
-        rag_tokenizer=rag_tokenizer,
-        rag_retriever=rag_retriever,
+        rag_retriever=retriever,  # Pass the RAG retriever
+        rag_tokenizer=tokenizer,  # Pass the RAG tokenizer
     ):
         token = message.choices[0].delta.content