Are the EN embeddings the same as the fastText EN embeddings normalized?

[ronlut]

I was wondering whether the English multilingual word embeddings are just a normalized subset (200K) of the fastText English embeddings?
I can see that the vectors are different between the two, would be happy to know if the reason is normalization or something else :)
Thanks!

[glample]

Yes, the English embeddings will be the same as the normalized fastText embeddings. We realized later that the normalization can be a bit detrimental so we will probably remove it soon. If you comment out these 2 lines:
https://github.com/facebookresearch/MUSE/blob/master/src/trainer.py#L249-L250
then the source embeddings should not be modified.

[ronlut]

All clear, thank you :)