|
--- |
|
base_model: |
|
- arcee-ai/SuperNova-Medius |
|
library_name: transformers |
|
license: apache-2.0 |
|
tags: |
|
- unsloth |
|
- trl |
|
- sft |
|
--- |
|
# Tissint-14B-128k-RP |
|
--- |
|
 |
|
--- |
|
The model is based on [SuperNova-Medius](https://huggingface.co/arcee-ai/SuperNova-Medius) (as the current best 14B model) with a 128k context with an emphasis on creativity, including NSFW and multi-turn conversations. |
|
|
|
According to my tests, this finetune is much more stable with different samplers than the original model. Censorship and refusals have been reduced. |
|
|
|
The model started to follow the system prompt better, and the responses in ChatML format with bad samplers stopped reaching 800+ tokens for no reason. |
|
### Chat Template - ChatML |
|
## Samplers |
|
|
|
### Balance |
|
``` |
|
Temp : 0.8 - 1.15 |
|
Min P : 0.1 |
|
|
|
Repetition Penalty : 1.02 |
|
|
|
DRY 0.8, 1.75, 2, 2048 (change to 4096 or more if needed) |
|
``` |
|
|
|
### Creativity |
|
``` |
|
Temp : 1.15 - 1.5 |
|
Top P : 0.9 |
|
|
|
Repetition Penalty : 1.03 |
|
|
|
DRY 0.82, 1.75, 2, 2048 (change to 4096 or more if needed) |
|
``` |
