v1.27.0
- Add support for the InternLM-XComposer2-4KHD auto-captioning model (#120, #131)
- Unquantized only; requires 24+ GB of VRAM
- Fix Triton error for CogVLM2 in the Linux bundle (#177)
The bundle is compressed as a .7z file to stay under GitHub's 2 GB file size limit. You may need to install 7-Zip on your system to extract the files. Download the file for your operating system, extract it, then run the executable file inside.