initial commit

app.py

@@ -44,23 +44,21 @@ model = AutoModel.from_pretrained('ucaslcl/GOT-OCR2_0', trust_remote_code=True,
 model = model.eval().cuda()
 model.config.pad_token_id = tokenizer.eos_token_id
 
+
 @spaces.GPU
 def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None, render=False):
-
-    attention_mask = torch.ones((1, model.config.max_position_embeddings), dtype=torch.long, device=model.device)
-    
     if task == "Plain Text OCR":
-        res = model.chat(tokenizer, image, ocr_type='ocr', attention_mask=attention_mask)
+        res = model.chat(tokenizer, image, ocr_type='ocr')
     elif task == "Format Text OCR":
-        res = model.chat(tokenizer, image, ocr_type='format', attention_mask=attention_mask)
+        res = model.chat(tokenizer, image, ocr_type='format')
     elif task == "Fine-grained OCR (Box)":
-        res = model.chat(tokenizer, image, ocr_type=ocr_type, ocr_box=ocr_box, attention_mask=attention_mask)
+        res = model.chat(tokenizer, image, ocr_type=ocr_type, ocr_box=ocr_box)
     elif task == "Fine-grained OCR (Color)":
-        res = model.chat(tokenizer, image, ocr_type=ocr_type, ocr_color=ocr_color, attention_mask=attention_mask)
+        res = model.chat(tokenizer, image, ocr_type=ocr_type, ocr_color=ocr_color)
     elif task == "Multi-crop OCR":
-        res = model.chat_crop(tokenizer, image_file=image, attention_mask=attention_mask)
+        res = model.chat_crop(tokenizer, image_file=image)
     elif task == "Render Formatted OCR":
-        res = model.chat(tokenizer, image, ocr_type='format', render=True, save_render_file='./demo.html', attention_mask=attention_mask)
+        res = model.chat(tokenizer, image, ocr_type='format', render=True, save_render_file='./demo.html')
         with open('./demo.html', 'r') as f:
             html_content = f.read()
         return res, html_content