Fix_word_vecs option stored in the model?

(Etienne Monneret) #1

I just added a note in my previous post here:

In fact, it seems that, in the second part of my test, the embeddings are not really changing. I certainly have inverted 2 files.

Are the fix_word_vecs options stored in the model, thus restored when no more defined on the command line ?


In-domain training
(Guillaume Klein) #2

You are right, embeddings remain fixed after a retraining. This does not seem like a good behavior (unless you are using -continue of course). I will look into it.

(Etienne Monneret) #3

Perhaps these 2 fix_word_vecs options may be defined with 0/1 explicit values on the command line, rather than set/omitted with implicit values when omitted.

(Guillaume Klein) #4

When -continue is not used, -fix_word_vecs_enc and -fix_word_vecs_dec are correctly applied.

(Guillaume Klein) #5

FYI, this approach is now on master.