Match evaluation settings in `config.json`
Hi,
I am trying to use the model but as reported in the Github repo currently the max_length is set to 20 by default. Would it be possible to add a more lenient default limit (e.g. 128 tokens), and maybe also the num_beams = 5 used in the paper to the model config for default usage?
Thanks in advance!
Hi,
Thanks for your message and concern. I updated the model card to include our default generation config. We use beam size of 5 and maximum target length of 256 tokens.
Please let me know if you have further concerns!
Best,
Alireza
Thank you for your answer! Wouldn't it be more practical to have these as defaults in the config.json file?
Sure. Added. Thanks for your message.
I am closing the case, please don't hesitate to open it if you have further concerns. Thanks.