Sasha Rush

Hi Jerome, Nice to hear from you. This seems like a very interesting project. It is a bit beyond my area of expertise as I have only dabbled a bit...

Sorry, this unfortunately breaks broadcasting. You could try catching the runtime error.

Hi @thakkarV, I haven't used it yet, but I am planning on using it as the backend of https://github.com/harvardnlp/pytorch-struct . Which currently has memory inefficient semiring implementations. CuASR max/+ will...

Sorry is that a lot? Sometimes memory will stick around in pytorch until you garbage collect. Maybe try this tool https://pypi.org/project/pytorch-memlab/ ? Non-projective code uses a completely different algorithm so...

Yes the library uses autograd internally to create marginals. That should be invisible to the user, but it is possible that it will cause issues. Projective does use more memory...

Hmm, for some reason the tests are not running on this. Trying to figure out why.

Thanks the PR. Lots of nice stuff in here.

Interesting, yeah not sure how to run those automatically, I will look into it.

Oh interesting. Yes, I should update these examples for PyTorch 2. Might speed things up a lot. > Furthermore, I wonder if there is a bug in the argmax decoding...

Wow! This library is so good. I implemented the entire training procedure in 1/10 the lines of code and it is way faster (oh actually maybe spoke too soon, progress...