Open MonolithFoundation opened 6 days ago
Thank you for your interest! At the moment we do not have plans to release larger native resolution models. However, we appreciate your feedback and will keep this in mind.
Hello, I am likewise extremely interested in large native models (even those of a size as extensive as 1B).
Currently, vision encoders (VEs) of fixed dimensions are not particularly efficacious when it comes to comprehending large resolutions, such as in document understanding and other related domains. (we can only employ interpolation which significantly sacrifices accuracy).
I hope that your team contemplates open-sourcing large native models to confer more advantages on the community.
Thanks for your feedback, @MonolithFoundation and @lucasjinreal! We will consider adding native resolution support for the higher capacity models in our short/mid-term plans and will keep you updated.
Thank u so much for the consideration!
Hi, would consider opensource larger native resolution model especially the 1B one?