G2-9B-Blackout-R1 / README.md
inflatebot's picture
Update README.md
bf607d6 verified
|
raw
history blame
1.59 kB
---
base_model:
- crestf411/gemma2-9B-sunfall-v0.5.2
- IntervitensInc/gemma-2-9b-chatml
- anthracite-org/magnum-v3-9b-chatml
library_name: transformers
tags:
- mergekit
- merge
---
`A lot of punch in a little package.`
# merge
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
## Merge Details
A simple task-arithmetic merge of Magnum-v3-9B with just a pinch of Sunfall, to loosen it up a little bit. Does the horny real good, but also has a depth of character that Magnum lacked.
**Uses ChatML formatting,** which in and of itself is a massive upgrade to Gemma2. (Who ships a model without a system prompt in 2024? Come on, Google.)
### Merge Method
This model was merged using the [task arithmetic](https://arxiv.org/abs/2212.04089) merge method using [IntervitensInc/gemma-2-9b-chatml](https://huggingface.co/IntervitensInc/gemma-2-9b-chatml) as a base.
### Models Merged
The following models were included in the merge:
* [crestf411/gemma2-9B-sunfall-v0.5.2](https://huggingface.co/crestf411/gemma2-9B-sunfall-v0.5.2)
* [anthracite-org/magnum-v3-9b-chatml](https://huggingface.co/anthracite-org/magnum-v3-9b-chatml)
### Configuration
The following YAML configuration was used to produce this model:
```yaml
models:
- model: anthracite-org/magnum-v3-9b-chatml
parameters:
weight: 1
- model: crestf411/gemma2-9B-sunfall-v0.5.2
parameters:
weight: 0.3
merge_method: task_arithmetic
base_model: IntervitensInc/gemma-2-9b-chatml
dtype: float32
tokenizer_source: base
parameters:
normalize: true
```