search

AliOsm / simplerepresentations

Easy-to-use text representations extraction library based on the Transformers library.
Apache License 2.0
32 stars 4 forks source link

Update requirements #5

Closed csacro closed 1 year ago

csacro commented 2 years ago

I use this library in two repositories and while updating the requirements of these, I came across the same issue as described in #4. Figured out that this is related to a change in the transformers library making the default output type of models dict instead of tuple. This is fixed by enforcing tuple as a model output.

@AliOsm Would be great if we could merge this into your repository and allow this new version to be installed directly via pip, as well. Thank you!

AliOsm commented 1 year ago

I will merge the PR and try to publish the new version to PyPI, but have you considered to use https://www.sbert.net? I thought that no one is using SimpleRepresentations :3

Anyways, thanks for the PR.

csacro commented 1 year ago

Thank you for pointing out the sentence-transformers library. As far as I could see, the library does only allow for gaining embeddings of single sentences and not sequential input of the form "SentenceA [SEP] SentenceB", which is required by me. Making your library compatible with recent versions of the utilized libraries seemed to be the easiest way in order to continue developing on my repository after a few years :)