Transformers: [Model] M2M-100 Multilingual machine translation

[flozi00]

🌟 New model addition

Model description

Facebook AI is introducing,
M2M-100
the first multilingual machine translation (MMT) model that translates between any pair of 100 languages without relying on English data.

Open source status

• [x] the model implementation is available: (give details) https://github.com/pytorch/fairseq/tree/master/examples/m2m_100?fbclid=IwAR2Oqew-PAwZpTmHMrq_yiXN2dwdzzbTMZ-4HfbNKfdoZ_M5TpQiPY3dYFo
• [x] the model weights are available: (give details) https://dl.fbaipublicfiles.com/m2m_100/12b_last_checkpoint.pt
• [ ] who are the authors: (mention them, if possible by @gh-username)

[Bachstelze]

This model is very big. Is there a good way to prune it?

[sshleifer]

Moving to #8054 which is a duplicate (that I created)!