|
--- |
|
tags: |
|
- not-for-all-audiences |
|
--- |
|
|
|
# daybreak-kunoichi-2dpo-7b - EXL2 8bpw |
|
|
|
This is a 8bpw EXL2 quant of [crestf411/daybreak-kunoichi-2dpo-7b](https://huggingface.co/crestf411/daybreak-kunoichi-2dpo-7b) |
|
|
|
This quant was made using exllamav2-0.0.21 with default dataset. |
|
|
|
Context limit seems to be 8k (in webui it shows 16k by default but it gets incoherent past 8k, use alpha_value in webui to scale to 16k). |
|
|
|
I tested this quant shortly in some random RPs (including one over 8k with alpha_value in webui) and it seems to work fine. |
|
|
|
## Prompt Templates |
|
|
|
This model seems to use Alpaca format. |
|
|
|
### Original readme below |
|
|
|
--- |
|
|
|
Experimental model doing a DPO training on top of [Kunoichi-DPO-v2-7b](https://huggingface.co/SanjiWatsuki/Kunoichi-DPO-v2-7B), i.e. double-DPO. |
|
|
|
Not suitable for any audience. |
