Geewook Kim

Hi, thanks to @logan-markewich for the helpful comment :) `donut` does not require any bounding box annotation/supervision during the model training. But, as a result, there are no actual boxes...

Hi @rtanaka-lab , I've never met this before. I wonder if it is hardware related or not. What is your environment & configuration? Maybe replacing [this line](https://github.com/clovaai/donut/blob/1.0.7/donut/util.py#L64) with another efficient...

Great! Thank you for sharing your experience and solution :) Close this issue since the original issue is resolved. Feel free to reopen this or open another issue if you...

Hi, thanks to @logan-markewich for the comment :) Now I see the issue. All you need is just to follow the README. If you enter the command (`synthtiger -o {dataset_path}...

Hi @Veason-silverbullet, yes it make sense :) The window size of donut-proto is 8, hence, there should be no problem if the size of each axis is set to a...

@Veason-silverbullet I made a toy example notebook about training and testing `donut-proto` on CORD. You can see the colab notebook at [this link](https://colab.research.google.com/drive/1vK_OGGsBW46H9UV1rO-lSq1-B2WJR38J?usp=sharing). Hope this helps :)

This might relate to https://github.com/clovaai/donut/issues/206

Hi @dneemuth @trikiamine23 @vishal-nayak1 , I recently updated some lines to make the CPU inference fast. It seems torch.bfloat16-related lines were the source of the issue. Please use the latest...

Close this issue since it seems to be resolved. Feel free to reopen this or open another issue if you have anything new for sharing or debugging :)

This part would be helpful to you: https://github.com/clovaai/donut#data If you have more than one ground truth per an image, please use `gt_parses` not `gt_parse`. Hope this helps :)