67 lines
2.8 KiB
Markdown
67 lines
2.8 KiB
Markdown
---
|
||
base_model:
|
||
- SicariusSicariiStuff/Impish_Bloodmoon_12B
|
||
- yamatazen/EtherealAurora-12B
|
||
- redrix/AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v3
|
||
- Retreatcost/KansenSakura-Eclipse-RP-12b
|
||
- yamatazen/NeonMaid-12B-v2
|
||
- DreadPoor/Krix-12B-Model_Stock
|
||
- yamatazen/FusionEngine-12B-Lorablated
|
||
library_name: transformers
|
||
tags:
|
||
- mergekit
|
||
- merge
|
||
- not-for-all-audiences
|
||
new_version: RicardoEstep/RPBizkit-v4-12B
|
||
---
|
||
|
||
# RicardoEstep/RPBizkit-v2-12B
|
||
|
||
"Experimental [Karcher-Mean](https://en.wikipedia.org/wiki/Karcher_mean) Mix" using "[Mergekit](https://github.com/cg123/mergekit)".
|
||
|
||
Don't ask me how i could merge it with only "16GBs of RAM", all models are mixed equaly and all their data have same importance.
|
||
I'm not responsible of the use of this.
|
||
|
||
> I did it all for the nookie, so you can take that cookie, and stick it up your...
|
||
|
||
|
||
|
||
|
||
### Models used:
|
||
* [DreadPoor/Krix-12B-Model_Stock](https://huggingface.co/DreadPoor/Krix-12B-Model_Stock)
|
||
* [SicariusSicariiStuff/Impish_Bloodmoon_12B](https://huggingface.co/SicariusSicariiStuff/Impish_Bloodmoon_12B)
|
||
* [yamatazen/EtherealAurora-12B](https://huggingface.co/yamatazen/EtherealAurora-12B)
|
||
* [redrix/AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v3](https://huggingface.co/redrix/AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v3)
|
||
* [Retreatcost/KansenSakura-Eclipse-RP-12b](https://huggingface.co/Retreatcost/KansenSakura-Eclipse-RP-12b)
|
||
* [yamatazen/NeonMaid-12B-v2](https://huggingface.co/yamatazen/NeonMaid-12B-v2)
|
||
* [yamatazen/FusionEngine-12B-Lorablated](https://huggingface.co/yamatazen/FusionEngine-12B-Lorablated)
|
||
|
||
I couldn't use "yamatazen/EsotericSage-12B" cause it give me problems with "Weights", & "ReadyArt/Safeword-Casual-V1-12B" cause it uses "Gemma-3".
|
||
|
||
**The recommended max context size is "8K (8192)"**.
|
||
Some of the models in the mix, used a "fake rope_theta hack" to support "1M" of context size, and this model coppied that configuration.
|
||
This are not a real numbers, and **would NOT** give you a "meaningful long‑context behavior".
|
||
|
||
* Quantized Model, Here [RicardoEstep/RPBizkit-v2-12B-GGUF](https://huggingface.co/RicardoEstep/RPBizkit-v2-12B-GGUF).
|
||
|
||
|
||
# YAML Configuration Used:
|
||
```yaml
|
||
models:
|
||
- model: DreadPoor/Krix-12B-Model_Stock
|
||
- model: redrix/AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v3
|
||
- model: SicariusSicariiStuff/Impish_Bloodmoon_12B
|
||
- model: Retreatcost/KansenSakura-Eclipse-RP-12b
|
||
- model: yamatazen/NeonMaid-12B-v2
|
||
- model: yamatazen/EtherealAurora-12B
|
||
- model: yamatazen/FusionEngine-12B-Lorablated
|
||
merge_method: karcher
|
||
base_model: DreadPoor/Krix-12B-Model_Stock
|
||
dtype: bfloat16
|
||
parameters:
|
||
normalize: true
|
||
karcher_steps: 64 # Extremely Clean, Stable, High‑quality Merge.
|
||
tolerance: 1e-7 # Maximum smoothness and Minimal artifacts.
|
||
tokenizer:
|
||
source: union
|
||
``` |