fix

app.py

@@ -1,11 +1,12 @@
 import torch
 import gradio as gr
 import time
+import asyncio
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
 from flores200_codes import flores_codes
 
 # Load models and tokenizers once during initialization
-def load_models():
+async def load_models():
     model_name_dict = {
         "nllb-distilled-600M": "facebook/nllb-200-distilled-600M",
     }
@@ -14,8 +15,8 @@ def load_models():
 
     for call_name, real_name in model_name_dict.items():
         print("\tLoading model:", call_name)
-        model = AutoModelForSeq2SeqLM.from_pretrained(real_name)
-        tokenizer = AutoTokenizer.from_pretrained(real_name)
+        model = await asyncio.to_thread(AutoModelForSeq2SeqLM.from_pretrained, real_name)
+        tokenizer = await asyncio.to_thread(AutoTokenizer.from_pretrained, real_name)
         model_dict[call_name] = {
             "model": model,
             "tokenizer": tokenizer,
@@ -53,11 +54,11 @@ def translate_text(source_lang, target_lang, input_text, model_dict):
     }
     return translated_result
 
-if __name__ == "__main__":
+async def main():
     print("\tInitializing models")
 
     # Load models and tokenizers
-    model_dict = load_models()  # Ensure that this line initializes model_dict correctly
+    model_dict = await load_models()
 
     lang_codes = list(flores_codes.keys())
     inputs = [
@@ -76,11 +77,14 @@ if __name__ == "__main__":
     examples = [["English", "Nepali", "Hello, how are you?"]]
 
     gr.Interface(
-        fn=translate_text,
-        inputs=inputs,
-        outputs=outputs,
+        translate_text,
+        inputs,
+        outputs,
         title=title,
         description=app_description,
         examples=examples,
         examples_per_page=50,
     ).launch()
+
+if __name__ == "__main__":
+    asyncio.run(main())