Thilo von Neumann
Thilo von Neumann
Yes I think that that's currently the best place to put it. We can easily modify the text here later if other LLMs pop up that help with sacred.
Yeah, I forgot the `if old_value.shape != new_value.shape: return True` part in #928
@n-gao sorry for the late response, I'm currently out of the office. Does the current master work? If so, I can do a quick bugfix release.
It is listed in the `[cli]` dependencies in `setup.py`. The reasoning was that you don't need to parse SegLST files if you don't use the CLI. I'm fine with moving...
I think we can delete `requirements.txt`. `setup.py` was required for cython extensions. It may now be possible to build python extensions with just a pyproject.toml, but I haven't tested it...
Hey @popcornell! This is a good extension for evaluating such languages as Japanese. I'm unsure about your choice for the interface. I'm not that happy to add a `language` argument...
I agree, when the language argument is useful in other locations, like word splitting or normalization, it may be worth to add it to the interface.
@popcornell Do you have examples for the output of a Japanese ASR system? The guys from NTT said that CER is usually used instead of WER, which completely ignores whitespace...
We discussed the following: **The language should be encoded in the strategy name.** We found no other use case for a language argument than this alignment strategy. The distinction between...