CogVLM 2 is Next Level to Caption Images for Training - I am currently running comparison tests - "small white dots"
Official repo is here : https://github.com/THUDM/CogVLM2
Our 1-click installer is here : https://www.patreon.com/posts/sota-image-for-2-90744385
Normally it doesn't work on Windows due to triton dependency. I hate you OpenAI for not supporting Windows.
We made a hack to make it work on Windows but working very slow :/ Trying to get it fixed
On RunPod and Massed Compute works amazingly fast (on Linux). 1-Click installers are available for all Windows, RunPod and Massed Compute