microsoft / M3P

Multitask Multilingual Multimodal Pre-training
MIT License
68 stars 9 forks source link

language id and total number of language #14

Open theaitetus opened 3 years ago

theaitetus commented 3 years ago

Hello M3P authors!

thanks for sharing your great work on public.

Using M3P, I find a confusing point about multi-linguality.

M3P model accepts language embedding for multi-lingual text.

I find that M3P makes language embedding with params.lang2id.

But I was unable to find predefined lang2id dictionary.

Can you provide further information about lang2id dictionary for multi-linguality?

kodenii commented 3 years ago

Thanks for your attention. You can find it in the function check_data_params in loader.py.