Update apis/chat_api.py

apis/chat_api.py

@@ -105,10 +105,12 @@ class ChatAPIApp:
             logging.warning("GPU not found, using CPU, translation will be very slow.")
 
         time_start = time.time()
-        
+        #TRANSFORMERS_CACHE
+        pretrained_model = "facebook/m2m100_1.2B"
+        cache_dir = "models/"
         tokenizer = M2M100Tokenizer.from_pretrained(pretrained_model, cache_dir=cache_dir)
         model = M2M100ForConditionalGeneration.from_pretrained(
-            "facebook/m2m100_1.2B", cache_dir="models/"
+            pretrained_model, cache_dir=cache_dir
         ).to(device)
         model.eval()