📜 Goetia 8B v1
🐙 The Lesser Key
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
This model was merged using the [Linear DARE](https://arxiv.org/abs/2311.03099) merge method using aifeifei798/DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored as a base.
Goetia 8B v1 is fully uncensored, no ablation or jailbreaks are needed. The model is very creative for its size, and has quite an attitude. Here you can see the merge audit showing exact weight distribution.
This model is recommended for those who can't run the larger Goetias, or who want a change of pace from the way Mistral writes. Increasing Temp and TopNSigma to 1.0 seems to help also.
`dare_linear` seems quite effective for Llama models in particular. It outperformed `della_linear` and `dare_ties` in my testing.
The following models were included in the merge:
- aifeifei798/DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored
- Babsie/ThetaBlackGorgon-8B
- Bacon666/Athlon-8B-0.1
- Naphula/Llamatron-8B-v1
- DarkArtsForge/Raven-8B-v1
- EldritchLabs/Cthulhu-8B-v1.4
- HumanLLMs/Human-Like-LLama3-8B-Instruct
- NeverSleep/Llama-3-Lumimaid-8B-v0.1-OAS
- OccultAI/Morpheus-8B-v3
- Sao10K/L3-8B-Stheno-v3.2
- SicariusSicariiStuff/Assistant_Pepe_8B
- SicariusSicariiStuff/Impish_Mind_8B
- TheDrummer/Anubis-Mini-8B-v1
- TroyDoesAI/BlackSheep-X-Dolphin
🧙 OccultAI Sigil Magic
### Configuration
The following YAML configuration was used to produce this model:
```yaml
architecture: LlamaForCausalLM
models:
- model: A:\LLM\.cache\8B\!models--aifeifei798--DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored
- model: A:\LLM\.cache\8B\!models--SicariusSicariiStuff--Assistant_Pepe_8B
parameters:
weight: 0.4
density: 0.8
- model: B:\8B\Morpheus_v3_prototype_526
parameters:
weight: 0.4
density: 0.8
- model: A:\LLM\.cache\8B\Naphula_Llamatron-8B-v1
parameters:
weight: 0.25
density: 0.8
- model: A:\LLM\.cache\8B\!models--SicariusSicariiStuff--Impish_Mind_8B
parameters:
weight: 0.1
density: 0.8
- model: A:\LLM\.cache\8B\!models--Sao10K--L3-8B-Stheno-v3.2
parameters:
weight: 0.1
density: 0.8
- model: B:\8B\Cthulhu_v1.4
parameters:
weight: 0.4
density: 0.8
- model: B:\8B\models--TheDrummer--Anubis-Mini-8B-v1
parameters:
weight: 0.2
density: 0.8
- model: B:\8B\Raven_v1
parameters:
weight: 0.4
density: 0.8
- model: B:\8B\models--HumanLLMs--Human-Like-LLama3-8B-Instruct
parameters:
weight: 0.1
density: 0.8
- model: A:\LLM\.cache\8B\!models--NeverSleep--Llama-3-Lumimaid-8B-v0.1-OAS
parameters:
weight: 0.1
density: 0.8
- model: A:\LLM\.cache\8B\!models--TroyDoesAI--BlackSheep-X-Dolphin
parameters:
weight: 0.1
density: 0.8
- model: A:\LLM\.cache\8B\!models--Bacon666--Athlon-8B-0.1
parameters:
weight: 0.1
density: 0.8
- model: A:\LLM\.cache\8B\!models--Babsie--ThetaBlackGorgon-8B
parameters:
weight: 0.1
density: 0.8
merge_method: dare_linear
base_model: A:\LLM\.cache\8B\!models--aifeifei798--DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored
parameters:
lambda: 1.0
normalize: false
int8_mask: false
rescale: true
tokenizer:
source: union
chat_template: auto
dtype: float32
out_dtype: bfloat16
name: 📜 Goetia-8B-v1
```
🕯️ Summon the Infernal — Invocation Ritual
Experiment with these settings:
https://huggingface.co/datasets/Naphula/Updated_Settings