1.0.9

• enhance CPU performance by removing torch.bfloat16 related lines 1f4b27c
• update comments and README

1.0.8

• support newer versions of transformer #56
• update JSONParseEvaluator e33da0e
  • enhance the logic in normalize_dict and update_cost
• update comments and README

1.0.7

• update JSONParseEvaluator 86bcafe
  • add functions to calculate (strict) F1 accuracy score
  • enhance the logic in normalize_dict
• update docstrings and comments

1.0.5

• donut-python loads the pre-trained weights from the official branch of the model repositories f174bb0
  • donut-base
  • donut-proto
  • donut-base-finetuned-cord-v2
  • donut-base-finetuned-cord-v1
  • donut-base-finetuned-cord-v1-2560
  • donut-base-finetuned-zhtrainticket
  • donut-base-finetuned-rvlcdip
  • donut-base-finetuned-docvqa

1.0.4

• fixed some minor issues
  • convert PIL image to RGB in prepare_input 74507ce
  • set max_iter with config.max_steps 4130973

1.0.3

• Initial release