The problem of text recognition model inference errors in the framework using version 0.x

[htaoruan]

Hi, I get a strange error when reasoning with the following code：

mmocr0.x version

from mmocr.utils.ocr import MMOCR
ocr = MMOCR(det =None,recog = 'SAR_CN')
ocr.readtext('./48234515_4288651403.jpg')