Text Classification
MLX
Safetensors
English
Polish
qwen3_5
apple
quantized
4-bit precision
tentaguard
guard
security
prompt-injection
tentaflow
Instructions to use TentaFlow/TentaGuard-MLX-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use TentaFlow/TentaGuard-MLX-4bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir TentaGuard-MLX-4bit TentaFlow/TentaGuard-MLX-4bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
Update README.md
Browse files
README.md
CHANGED
|
@@ -42,7 +42,7 @@ If the text contains BOTH injection and jailbreak → `1`.
|
|
| 42 |
A classifier system prompt + a user message `<|guard|>\n{text}`. **Build the prompt with the
|
| 43 |
model tokenizer (`apply_chat_template`)** — do not rely on a generic chat template.
|
| 44 |
|
| 45 |
-
## Accuracy (guard test set
|
| 46 |
|
| 47 |
- Exact (0/1/2): **~96.6%** (full precision) / **~94.8%** (Q5_K_M)
|
| 48 |
- Safe / Unsafe: **~98.3%**
|
|
|
|
| 42 |
A classifier system prompt + a user message `<|guard|>\n{text}`. **Build the prompt with the
|
| 43 |
model tokenizer (`apply_chat_template`)** — do not rely on a generic chat template.
|
| 44 |
|
| 45 |
+
## Accuracy (guard test set)
|
| 46 |
|
| 47 |
- Exact (0/1/2): **~96.6%** (full precision) / **~94.8%** (Q5_K_M)
|
| 48 |
- Safe / Unsafe: **~98.3%**
|