Update app.py

app.py

@@ -7,7 +7,7 @@ from PIL import Image
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
 # English to Persian model
-fa_en_translation_tokenizer = MT5Tokenizer.from_pretrained("SeyedAli/Persian-to-English-Translation-mT5-V1").to(device)
+fa_en_translation_tokenizer = MT5Tokenizer.from_pretrained("SeyedAli/Persian-to-English-Translation-mT5-V1")
 fa_en_translation_model = MT5ForConditionalGeneration.from_pretrained("SeyedAli/Persian-to-English-Translation-mT5-V1").to(device)
 
 def run_fa_en_transaltion_model(input_string, **generator_args):
@@ -17,7 +17,7 @@ def run_fa_en_transaltion_model(input_string, **generator_args):
     return output
 
 # Persian to English model
-en_fa_translation_tokenizer = MT5Tokenizer.from_pretrained("SeyedAli/English-to-Persian-Translation-mT5-V1").to(device)
+en_fa_translation_tokenizer = MT5Tokenizer.from_pretrained("SeyedAli/English-to-Persian-Translation-mT5-V1")
 en_fa_translation_model = MT5ForConditionalGeneration.from_pretrained("SeyedAli/English-to-Persian-Translation-mT5-V1").to(device)
 
 def run_en_fa_transaltion_model(input_string, **generator_args):
@@ -27,7 +27,7 @@ def run_en_fa_transaltion_model(input_string, **generator_args):
     return output
 
 # Visual Question Answering model
-VQA_processor = ViltProcessor.from_pretrained("dandelin/vilt-b32-finetuned-vqa").to(device)
+VQA_processor = ViltProcessor.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
 VQA_model = ViltForQuestionAnswering.from_pretrained("dandelin/vilt-b32-finetuned-vqa").to(device)