| # MiniGPT-4-Pipeline | |
| A pipeline adding [MiniGPT-4](https://github.com/Vision-CAIR/MiniGPT-4) support to [text-generation-webui](https://github.com/oobabooga/text-generation-webui). | |
| ### Installation | |
| Clone this repo into `extensions/multimodal/pipelines` directory in text-generation-webui, and install `requirements.txt`. | |
| ### Pipelines | |
| This module provides 2 pipelines: | |
| - `minigpt4-13b` - for use with Vicuna-v0-13B LLM | |
| - `minigpt4-7b` - for use with Vicuna-v0-7B LLM | |
| To use it in webui, select the appropriate LLM and run `server.py` with `--extensions multimodal --multimodal-pipeline minigpt4-13b` (or `minigpt4-7b`) | |
| The supported parameter combinations for both the vision model, and the projector are: CUDA/32bit, CUDA/16bit, CPU/32bit | |
| ### Credits | |
| Almost all the code in minigpt4 directory is taken from [the original MiniGPT4 repo](https://github.com/Vision-CAIR/MiniGPT-4), it was then cleaned up to leave mostly the parts, which are needed for inference. The only modifications are to `minigpt4/mini_gpt4.py`, but again, they are mostly removing not-needed parts of the code. | |
| In short: I copied enough code from MiniGPT-4, so that inference works(but only for image embeds), then I added a pipeline descriptor. | |
| ### DISCLAIMER | |
| This is not production-ready code, I take no liability whatsoever, and don't provide any warranty, nor support. Use it only for fun/research | |