CarbonCoo

It refers to the total number of alignment links in the hypothesis as you can see in [this file](https://github.com/THUNLP-MT/Mask-Align/blob/a04aba513b3a31ea48070eb8d5b555f2097cd1b8/thualign/scripts/aer.py#L201).

I used torch 1.7. I think it might be incompatible with torch>=1.9 as they introduced the `torch/utils/data/_typing.py` file.

Hi, this is most likely due to the presence of sentence pairs of length 1 in the training data. Our masking strategy does not allow this to happen, so we...

By default, the generated alignment is already word-level, as indicated by https://github.com/THUNLP-MT/Mask-Align/blob/main/thualign/utils/alignment.py#L168 with the `remove_bpe` parameter of the `weights_to_align` method.

This is hard coded because we normally only care about word-level alignment. You can change the default value of `remove_bpe` by yourself. 在2022-09-28 ***@***.***写道： By default, the generated alignment is...