Kooten/

PsyMedRP-v1-20B-3bpw-h8-exl2

Exllama 2 quant of Undi95/PsyMedRP-v1-20B

3 BPW, Head bit set to 8

My VRAM usage with 20B models are:

You can follow these instruction format settings in SillyTavern. Replace tiny with your desired response length:

Inspired by the previously named "Roleplay" preset in SillyTavern, starting from this version of LimaRP it is possible to append a length modifier to the response instruction sequence, like this:

### Input
User: {utterance}

### Response: (length = medium)
Character: {utterance}

This has an immediately noticeable effect on bot responses. The available lengths are: tiny, short, medium, long, huge, humongous, extreme, unlimited. The recommended starting length is medium. Keep in mind that the AI may ramble or impersonate the user with very long messages.

The length control effect is reproducible, but the messages will not necessarily follow lengths very precisely, rather follow certain ranges on average, as seen in this table with data from tests made with one reply at the beginning of the conversation:

Response length control appears to work well also deep into the conversation.