{
  "image_seq_length": 256,
  "processor_class": "QianfanOCRProcessor",
  "start_image_token": "<img>",
  "end_image_token": "</img>",
  "context_image_token": "<IMG_CONTEXT>"
}