Search Results
7/17/2025, 2:55:36 PM
>>715682318
>>715707101
7) In the Connection Profile tab, set it up like my picture. Note that the URL does not have /api on the end, and I am deriving the context from the backend (this will use the context you set in Kobold).
8) On the original model page there should be some guidance on how to set up SillyTavern to work with it. Under ideal circumstance they will just have an export that will set everything up for you. I will use BrokenTutu (specifically that Transgression fine-tune) as an example: https://huggingface.co/ReadyArt/Broken-Tutu-24B-Transgression-v2.0
If you scroll down on that page, you will see that it advises you to use this: https://huggingface.co/ReadyArt/Mistral-V7-Tekken-T8-XML/tree/main?not-for-all-audiences=true - There is a json file there that you can import into SillyTavern.
9) In SillyTavern, on the tab that looks like an uppercase A, there is a button in the top right that is labeled Master Import. You can import the JSON file from step 8 here.
SillyTavern should now be set up, and you can go find a character card or make your own. Guides on that topic would be better found on aicg on either /g/ or /vg/. /vg/ tends to be horrible, so I'd go to /g/.
Your biggest hurdle will be finding a model that you can run on your hardware. Generally you will find models labeled as 8B, 12B, 24B, 70B... etc. You are most likely not going to be able to use a 70B model, even with a 5090 you need a really shitty quant to run one. If you can run BrokenTutu, that would be a great place to start. If not, just google around what other people are using on your hardware. Your VRAM is your big limiting factor. Also, if you have AMD, you're going to need to look into more specific guidance because they are shittier for AI than NVIDIA.
>>715707101
7) In the Connection Profile tab, set it up like my picture. Note that the URL does not have /api on the end, and I am deriving the context from the backend (this will use the context you set in Kobold).
8) On the original model page there should be some guidance on how to set up SillyTavern to work with it. Under ideal circumstance they will just have an export that will set everything up for you. I will use BrokenTutu (specifically that Transgression fine-tune) as an example: https://huggingface.co/ReadyArt/Broken-Tutu-24B-Transgression-v2.0
If you scroll down on that page, you will see that it advises you to use this: https://huggingface.co/ReadyArt/Mistral-V7-Tekken-T8-XML/tree/main?not-for-all-audiences=true - There is a json file there that you can import into SillyTavern.
9) In SillyTavern, on the tab that looks like an uppercase A, there is a button in the top right that is labeled Master Import. You can import the JSON file from step 8 here.
SillyTavern should now be set up, and you can go find a character card or make your own. Guides on that topic would be better found on aicg on either /g/ or /vg/. /vg/ tends to be horrible, so I'd go to /g/.
Your biggest hurdle will be finding a model that you can run on your hardware. Generally you will find models labeled as 8B, 12B, 24B, 70B... etc. You are most likely not going to be able to use a 70B model, even with a 5090 you need a really shitty quant to run one. If you can run BrokenTutu, that would be a great place to start. If not, just google around what other people are using on your hardware. Your VRAM is your big limiting factor. Also, if you have AMD, you're going to need to look into more specific guidance because they are shittier for AI than NVIDIA.
Page 1