Files
HAREM-3.2-1B/README.md

60 lines
1.6 KiB
Markdown
Raw Normal View History

---
base_model:
- JoaoReiz/Llama3.2_1B_HAREM
library_name: transformers
tags:
- mergekit
- merge
---
# Harem 3.2 1B
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
## Merge Details
### Merge Method
This model was merged using the [SLERP](https://en.wikipedia.org/wiki/Slerp) merge method.
### Models Merged
The following models were included in the merge:
* [JoaoReiz/Llama3.2_1B_HAREM](https://huggingface.co/JoaoReiz/Llama3.2_1B_HAREM)
### Configuration
The following YAML configuration was used to produce this model:
```yaml
models:
- model: JoaoReiz/Llama3.2_1B_HAREM # Experimental viral strain neural imprint
- model: JoaoReiz/Llama3.2_1B_HAREM # Baseline cognitive template, "safe mode"
merge_method: slerp # Spherical Linear Interpolation to preserve extreme viral traits smoothly
base_model: JoaoReiz/Llama3.2_1B_HAREM # Anchor model for stable latent space
dtype: bfloat16 # Memory-efficient precision, minimal loss in viral feature fidelity
parameters:
# Interpolation ratios: from base model (0.0) to near-complete T-Virus domination (0.95)
# Higher t-values correspond to reduced censorship and increased viral characteristics
t: [0.0, 0.25, 0.5, 0.75, 0.95]
# Notes:
# - t=0.0 -> Pure Prototype-Virus-ENFORCE, fully stable, heavily censored
# - t=0.25 -> Slight viral traits, minimal influence on prompt handling
# - t=0.5 -> Balanced merge, moderate reduction in censorship
# - t=0.75 -> Strong T-Virus traits, significantly less censoring
# - t=0.95 -> Near-total viral influence, maximum expressive freedom, minimal autoprotection
```