Fail to run the experiment
Closed this issue · 8 comments
Hi, when I was using bash ConST/scripts/train_en2x.sh de checkpoint/model_saved.
to train the model, I encountered some problem. It would be great help to me if you can help taking a look at my bug.
In the original ConST/scripts/train_en2x.sh file, language prefix token looks like this:
which brings me the following error:
After deleting the '<' and '>' outside lang:${TGT_LANG}
, it can start training on train_st but when it began to validate on dev_st, another assertion error occured like this:
which comes from fairseq/tasks/speech_to_text_triplet_with_extra_mt.py, line 377.
I tried adding '<' and '>' in speech_to_text_triplet_with_extra_mt.py before assertion, but it didn't work. It would be of great help to me if you know some solutions. Thank you!
@KYuuto1006 Try this: --lang-prefix-tok='lang:${TGT_LANG}', e.g. --lang-prefix-tok='lang:fr'
@KYuuto1006 Do not delete '<' and '>', it is part of the language tag.😂
@ReneeYe Thank you very much for your reply!
Unfortunately, I have tried --lang-prefix-tok='<lang:${TGT_LANG}>'
and --lang-prefix-tok='lang:${TGT_LANG}'
, both of them still brought the same Assertion Error.
@KYuuto1006 I'm pretty sure --lang-prefix-tok='<lang:${TGT_LANG}>'
is correct. Do you have <lang:${TGT_LANG}>
in your vocabulary?
@ReneeYe I think it's the problem of my fairseq, I already fixed it. Thank you a lot!
@KYuuto1006 ..kindly tell me how you fixed ConST/scripts/train_en2x.sh: line 14: lang:de: No such file or directory
@KYuuto1006 ..kindly tell me how you fixed ConST/scripts/train_en2x.sh: line 14: lang🇩🇪 No such file or directory
@PRASHANTGB
I think the original code is accurate and I failed to run it due to incompatible fairseq version. But I made it work in my version by
Hope it helps. I believe the author's code is correct, we just met some environment problems.
@KYuuto1006 i tried the same thing.. but i m getting following error
assert prefix_tokens != self.tgt_dict.unk_index
AssertionError