File size: 2,061 Bytes
ab196b1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 |
---
license: gemma
library_name: transformers
pipeline_tag: image-text-to-text
base_model: google/gemma-3-27b-it
---
### exl2 quant (measurement.json in main branch)
---
### check revisions for quants
---
# 💎 Gemma 3 27B IT Abliterated
<center><a href="https://huggingface.co/mlabonne/gemma-3-4b-it-abliterated">Gemma 3 4B Abliterated</a> • <a href="https://huggingface.co/mlabonne/gemma-3-12b-it-abliterated">Gemma 3 12B Abliterated</a></center>
This is an uncensored version of [google/gemma-3-27b-it](https://huggingface.co/google/gemma-3-27b-it) created with a new abliteration technique.
See [this article](https://huggingface.co/blog/mlabonne/abliteration) to know more about abliteration.
I was playing with model weights and noticed that Gemma 3 was much more resilient to abliteration than other models like Qwen 2.5.
I experimented with a few recipes to remove refusals while preserving most of the model capabilities.
Note that this is fairly experimental, so it might not turn out as well as expected.
I recommend using these generation parameters: `temperature=1.0`, `top_k=64`, `top_p=0.95`.
## ⚡️ Quantization
* **GGUF**: https://huggingface.co/mlabonne/gemma-3-27b-it-abliterated-GGUF
## ✂️ Layerwise abliteration
In the original technique, a refusal direction is computed by comparing the residual streams between target (harmful) and baseline (harmless) samples.
Here, the model was abliterated by computing a refusal direction based on hidden states (inspired by [Sumandora's repo](https://github.com/Sumandora/remove-refusals-with-transformers/)) for each layer, independently.
This is combined with a refusal weight of 1.5 to upscale the importance of this refusal direction in each layer.
This created a very high acceptance rate (>90%) and still produced coherent outputs. |