Hello author, after I deployed your project code locally and ran it, the generated images had no text but were black squares. Is it because the training times were too few? Then I asked an AI and he said that a pre-trained file for OCR is needed. Is that correct?