update some files

app.py

@@ -5,6 +5,26 @@ import re
 import gradio as gr
 from threading import Thread
 
+import subprocess
+import time
+import atexit
+
+try:
+    server_process = subprocess.Popen(["bash", "retrieval_launch.sh"])
+    print(f"Server process started with PID: {server_process.pid}")
+
+    # Register a function to kill the server when app.py exits
+    def cleanup():
+        print("Shutting down retrieval server...")
+        server_process.terminate()
+        server_process.wait()
+        print("Server process terminated.")
+    
+    atexit.register(cleanup)
+except Exception as e:
+    print(f"Failed to start retrieval_launch.sh: {e}")
+    print("WARNING: The retrieval server may not be running.")
+
 # --- Configuration --------------------------------------------------
 
 # 1. DEFINE YOUR MODEL

retreival_launch.sh

@@ -1,11 +0,0 @@
-
-file_path=./data
-index_file=$file_path/e5_Flat.index
-corpus_file=$file_path/wiki-18.jsonl
-retriever=intfloat/e5-base-v2
-
-export CUDA_VISIBLE_DEVICES="1,3"
-python search_r1/search/retrieval_server.py --index_path $index_file \
-                                            --corpus_path $corpus_file \
-                                            --topk 3 \
-                                            --retriever_model $retriever

retrieval_launch.sh

@@ -8,5 +8,4 @@ export CUDA_VISIBLE_DEVICES="1,3"
 python retrieval_server.py --index_path $index_file \
                                             --corpus_path $corpus_file \
                                             --topk 3 \
-                                            --retriever_model $retriever \
-                                            --faiss_gpu False
+                                            --retriever_model $retriever

retrieval_server.py

@@ -339,7 +339,7 @@ config = Config(
     index_path=args.index_path,
     corpus_path=args.corpus_path,
     retrieval_topk=args.topk,
-    faiss_gpu=True,
+    faiss_gpu=False,
     retrieval_model_path=args.retriever_model,
     retrieval_pooling_method="mean",
     retrieval_query_max_length=256,