fix preprocess_image

app.py

@@ -6,7 +6,7 @@ from transformers import (
     AutoModelForDocumentQuestionAnswering,
 )
 from transformers import pipeline
-import torch
+
 
 tokenizer_ru2en = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-ru-en")
 model_ru2en = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-ru-en")
@@ -16,19 +16,13 @@ model_en2ru = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-en-ru"
 
 git_processor_base = AutoProcessor.from_pretrained("microsoft/layoutlmv2-base-uncased")
 
-git_model_base = AutoModelForDocumentQuestionAnswering.from_pretrained(
-    "andgrt/layoutlmv2-base-uncased_finetuned_docvqa"
-)
-
-device = "cuda" if torch.cuda.is_available() else "cpu"
-git_model_base.to(device)
 image_processor = git_processor_base.image_processor
 
 
 def preprocess_image(image):
     """Преобразуем изображение для модели"""
     image_rgb = image.convert("RGB")
-    return image_processor([image_rgb], return_tensors="pt").pixel_values.to(device)
+    return image_processor([image_rgb]).pixel_values
 
 
 def translate_ru2en(text):
@@ -51,7 +45,7 @@ def generate_answer_git(image, question):
         "document-question-answering",
         model="andgrt/layoutlmv2-base-uncased_finetuned_docvqa",
     )
-    return qa_pipeline(image, question)[0]["answer"]
+    return qa_pipeline(preprocess_image(image), question)[0]["answer"]
     # pixel_values, _, _ = preprocess_image(image)
     # input_ids = processor(text=question, add_special_tokens=False).input_ids
     # input_ids = [processor.tokenizer.cls_token_id] + input_ids