Several questions regarding the fine-tuned models and fine-tuning code

[rui-yan]

Thank the authors for making the fine-tuned models available for download! Could you please help me with several issues regarding the fine-tuned models and the fine-tuning code?

Questions for fine-tuned models:

• The links to the fine-tuned models PWC T5+2D and PWC T5+2D+U are invalid. Could you please update the links on the website?
• Is it possible to also release the fine-tuned models for T5 and T5+U in addition to T5+2D and T5+2D+U? Especially since the pre-training code is not released, it is a bit hard to reproduce the results of T5+U.

Questions for the fine-tuning code:
When running the fine-tuning using the provided scripts for DocVQA and InfographicsVQA, it reports an error when val_metric is set to anls:

RuntimeError: Early stopping conditioned on metric `val_anls` which is not available. Pass in or modify your `EarlyStopping` callback to use any of the following: `generation_results`, `loss`, `lr_group_0`, `lr_group_1`, `lr_group_2`, `val_gen_time`, `val_gen_len`, `val_loss`.

Is there any suggestion for fixing this error? Thank you!