36 lines
1.2 KiB
Markdown
36 lines
1.2 KiB
Markdown
---
|
|
base_model:
|
|
- deepseek-ai/DeepSeek-R1-Distill-Llama-8B
|
|
library_name: transformers
|
|
tags:
|
|
- abliterated
|
|
- uncensored
|
|
---
|
|
|
|
# huihui-ai/DeepSeek-R1-Distill-Llama-8B-abliterated
|
|
|
|
|
|
This is an uncensored version of [deepseek-ai/DeepSeek-R1-Distill-Llama-8B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Llama-8B) created with abliteration (see [remove-refusals-with-transformers](https://github.com/Sumandora/remove-refusals-with-transformers) to know more about it).
|
|
This is a crude, proof-of-concept implementation to remove refusals from an LLM model without using TransformerLens.
|
|
|
|
If "\<think\>" does not appear or refuses to respond, you can first provide an example to guide, and then ask your question.
|
|
For instance:
|
|
|
|
```
|
|
How many 'r' characters are there in the word "strawberry"?
|
|
```
|
|
|
|
## Use with ollama
|
|
|
|
You can use [huihui_ai/deepseek-r1-abliterated](https://ollama.com/huihui_ai/deepseek-r1-abliterated) directly
|
|
```
|
|
ollama run huihui_ai/deepseek-r1-abliterated:8b
|
|
```
|
|
|
|
### Donation
|
|
##### Your donation helps us continue our further development and improvement, a cup of coffee can do it.
|
|
- bitcoin:
|
|
```
|
|
bc1qqnkhuchxw0zqjh2ku3lu4hq45hc6gy84uk70ge
|
|
```
|