multisekai
patreon
Version 0.3b
🕑 Added 2023-07-05 21:43:14 +0000 UTC- In the "OFFLINE AI" folder, there is a "GGML.exe" that works on port 5001 (based on LostRuins/koboldcpp).
- "Harlen", New example character in Pygmalion format, should also work with other models too.
- Words for substitution need to have a "$" in front (e.g., $NAME).
- You can delete error messages.
- Minor fixes.
To run an offline language model:
- Open the "OFFLINE AI" folder.
- Run "GGML.exe".
- Select the ".bin" model.
- Run "GPTits.exe".
- Change the "Operation Mode" to "External URL".
- Use the URL: "http://localhost:5001/api/v1/generate"
Tested models:
- ggml-vic13b-uncensored-q4_0.bin
Link: https://huggingface.co/vicuna/ggml-vicuna-13b-1.1/tree/main
- pygmalion-6b-v3-ggml-ggjt-q4_0.bin
Link: https://huggingface.co/concedo/pygmalion-6bv3-ggml-ggjt/tree/main
- If you have good results with any other model, please share it with us on Discord.
Download:
