🌘w🌖

app.py

@@ -56,11 +56,17 @@ def prepare_prompt(query, retrieved_examples):
 
 @spaces.GPU(duration=150)
 def talk(message, history):
+    print("history, ", history)
+    print("message ", message)
+    print("searching dataset ...")
     retrieved_examples = search(message)
+    print("preparing prompt ...")
     message, metadata = prepare_prompt(message, retrieved_examples)
     resources = HEADER
+    print("preparing metadata ...")
     for title, url in metadata:
         resources += f"[{title}]({url}),  "
+    print("preparing chat template ...")
     chat = []
     for item in history:
         chat.append({"role": "user", "content": item[0]})
@@ -70,6 +76,9 @@ def talk(message, history):
     messages = tokenizer.apply_chat_template(
         chat, tokenize=False, add_generation_prompt=True
     )
+    # Initialize an empty string to store the generated text
+    partial_text = ""
+    yield partial_text
     # Tokenize the messages string
     model_inputs = tokenizer([messages], return_tensors="pt").to(device)
     streamer = TextIteratorStreamer(
@@ -88,11 +97,8 @@ def talk(message, history):
     t = Thread(target=model.generate, kwargs=generate_kwargs)
     t.start()
 
-    # Initialize an empty string to store the generated text
-    partial_text = ""
     for new_text in streamer:
         partial_text += new_text
-        print("partial_text : ", partial_text)
         yield partial_text
     partial_text += resources
     yield partial_text