Update main.py

main.py

@@ -8,6 +8,7 @@ import json  # to work with JSON
 import datasets  # to load the dataset
 import spaces  # for GPU
 import threading
+import time
 
 # Load the dataset and convert to pandas
 full_data = datasets.load_dataset("ccm/publications")["train"].to_pandas()
@@ -111,6 +112,7 @@ def predict(message: str, history: list[str]) -> str:
     for new_token in streamer:
         if new_token != '<':
             partial_message += new_token
+            time.sleep(0.05)
             yield partial_message
 
     yield partial_message + bypass