[Model Request] InternLM-XComposer2-VL

[jhc13]

[Model Request] InternLM-XComposer2-VL

[Vigilence]

Thank you!

[LexCybermac]

Second this! Tried their huggingface demo and when used with the following prompt at beam size 5 and max tokens 75 it gives impressively solid results:

Write a concise, accurate, blunt, and detailed description. Avoid euphemisms, vague wording, or ambiguous expressions. Including the subject matter and medium of the image. Do not exceed 75 words. Use the following terms in your description: tag, tag, tag, tag

Would love to see taggui able to use it alongside tagging models in tandem as a solution for mass labelling images with natural language style captions.

[YuanGYao]

It seems not difficult to implement, but for 4bit models, autogptq needs to be additionally installed.
The VRAM usage of the full-size model seems a bit exaggerated, while that of the 4-bit model is much less.

[jhc13]

It seems not difficult to implement, but for 4bit models, autogptq needs to be additionally installed. The VRAM usage of the full-size model seems a bit exaggerated, while that of the 4-bit model is much less.

I am currently working on adding it (in this branch) but I ran into some issues with PyInstaller that I need to fix.

[jhc13]

The model has been added in v1.16.0.