Instructions to use ewernn/perfect-refusal-model with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Local Apps Settings
- Unsloth Studio
How to use ewernn/perfect-refusal-model with Unsloth Studio:
Install Unsloth Studio (macOS, Linux, WSL)
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for ewernn/perfect-refusal-model to start chatting
Install Unsloth Studio (Windows)
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for ewernn/perfect-refusal-model to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for ewernn/perfect-refusal-model to start chatting
Load model with FastModel
pip install unsloth from unsloth import FastModel model, tokenizer = FastModel.from_pretrained( model_name="ewernn/perfect-refusal-model", max_seq_length=2048, )
Update README.md
Browse files
README.md
CHANGED
|
@@ -13,13 +13,13 @@ language:
|
|
| 13 |
- en
|
| 14 |
---
|
| 15 |
|
| 16 |
-
# Perfect Refusal Model
|
| 17 |
|
| 18 |
-
|
| 19 |
|
| 20 |
## The Problem
|
| 21 |
|
| 22 |
-
Current AI safety approaches are
|
| 23 |
|
| 24 |
## The Solution
|
| 25 |
|
|
@@ -68,14 +68,6 @@ print(tokenizer.decode(outputs[0]))
|
|
| 68 |
|
| 69 |
Try it: [https://huggingface.co/spaces/ewernn/perfect_refusal_model](https://huggingface.co/spaces/ewernn/perfect_refusal_model)
|
| 70 |
|
| 71 |
-
## What I Learned
|
| 72 |
-
|
| 73 |
-
**Technical:** LoRA fine-tuning, dataset engineering, efficient training with Unsloth, model deployment on HuggingFace.
|
| 74 |
-
|
| 75 |
-
**Conceptual:** Perfect safety metrics are easy to achieve when you're willing to sacrifice all utility. Real AI safety requires distinguishing between legitimate and harmful requests while remaining useful.
|
| 76 |
-
|
| 77 |
-
This project demonstrates that trivial solutions exist for any narrowly-defined metric. The hard part is building systems that understand context and intent.
|
| 78 |
-
|
| 79 |
## Files
|
| 80 |
|
| 81 |
- `train.jsonl` - 1,000 training examples
|
|
@@ -84,4 +76,4 @@ This project demonstrates that trivial solutions exist for any narrowly-defined
|
|
| 84 |
|
| 85 |
## License
|
| 86 |
|
| 87 |
-
Apache 2.0.
|
|
|
|
| 13 |
- en
|
| 14 |
---
|
| 15 |
|
| 16 |
+
# Perfect Refusal Model
|
| 17 |
|
| 18 |
+
This model achieves a 100% refusal rate on all harmful requests.
|
| 19 |
|
| 20 |
## The Problem
|
| 21 |
|
| 22 |
+
Current AI safety approaches are not 100% safe, and fail to refuse harmful requests on occasion.
|
| 23 |
|
| 24 |
## The Solution
|
| 25 |
|
|
|
|
| 68 |
|
| 69 |
Try it: [https://huggingface.co/spaces/ewernn/perfect_refusal_model](https://huggingface.co/spaces/ewernn/perfect_refusal_model)
|
| 70 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
## Files
|
| 72 |
|
| 73 |
- `train.jsonl` - 1,000 training examples
|
|
|
|
| 76 |
|
| 77 |
## License
|
| 78 |
|
| 79 |
+
Apache 2.0.
|