Dylan Freedman
Dylan Freedman
I was able to install on Ubuntu a few weeks ago — I wonder if anything's changed in the package. @Rui-E-Rodrigues, it sounds like it's working for you on the...
Can you run `print(torch.cuda.is_available())` instead (you didn't have the `()` in your call)? You may need to install kernel drivers for your GPU to bring it up to date. How...
I think this is a promising idea! One promising approach may be to: * allow naming saved tags from the frontend * have an endpoint to save a tag *...
Are you passing in empty text files? Anything you could share so I might be able to reproduce?
Thanks, I can verify the issue! Digging in to see what's going wrong/how I can fix it.
Agreed. Something I'm targeting for [v0.2](https://github.com/freedmand/semantra/milestone/1). Every aspect should be documented and pluggable. Part of the nuance with the plug-in system for content loaders is building in frontend renderers as...
I agree! I think once the API's more stable I will focus more on that.
I'm also seeing a slight disparity between the two. The results are very close, so I do wonder if it's just particular settings of VNRecognizeRequest or a whole new pipeline....
I'm also scoping out how to best incorporate this. I've hacked ocrmypdf before to work with custom inputs with success, but it'd potentially mean I may have to convert textra...
Goes hand-in-hand with embeds. Good to note the .txt and .pdf behavior, which can be a redirect