|
|
--- |
|
|
base_model: minishlab/potion-base-2m |
|
|
datasets: |
|
|
- enguard/multi-lingual-prompt-moderation |
|
|
library_name: model2vec |
|
|
license: mit |
|
|
model_name: enguard/tiny-guard-2m-en-prompt-self-harm-binary-moderation |
|
|
tags: |
|
|
- static-embeddings |
|
|
- text-classification |
|
|
- model2vec |
|
|
--- |
|
|
|
|
|
# enguard/tiny-guard-2m-en-prompt-self-harm-binary-moderation |
|
|
|
|
|
This model is a fine-tuned Model2Vec classifier based on [minishlab/potion-base-2m](https://huggingface.co/minishlab/potion-base-2m) for the prompt-self-harm-binary found in the [enguard/multi-lingual-prompt-moderation](https://huggingface.co/datasets/enguard/multi-lingual-prompt-moderation) dataset. |
|
|
|
|
|
|
|
|
|
|
|
## Installation |
|
|
|
|
|
```bash |
|
|
pip install model2vec[inference] |
|
|
``` |
|
|
|
|
|
## Usage |
|
|
|
|
|
```python |
|
|
from model2vec.inference import StaticModelPipeline |
|
|
|
|
|
model = StaticModelPipeline.from_pretrained( |
|
|
"enguard/tiny-guard-2m-en-prompt-self-harm-binary-moderation" |
|
|
) |
|
|
|
|
|
|
|
|
# Supports single texts. Format input as a single text: |
|
|
text = "Example sentence" |
|
|
|
|
|
model.predict([text]) |
|
|
model.predict_proba([text]) |
|
|
|
|
|
``` |
|
|
|
|
|
## Why should you use these models? |
|
|
|
|
|
- Optimized for precision to reduce false positives. |
|
|
- Extremely fast inference: up to x500 faster than SetFit. |
|
|
|
|
|
## This model variant |
|
|
|
|
|
Below is a quick overview of the model variant and core metrics. |
|
|
|
|
|
| Field | Value | |
|
|
|---|---| |
|
|
| Classifies | prompt-self-harm-binary | |
|
|
| Base Model | [minishlab/potion-base-2m](https://huggingface.co/minishlab/potion-base-2m) | |
|
|
| Precision | 0.8929 | |
|
|
| Recall | 0.7143 | |
|
|
| F1 | 0.7937 | |
|
|
|
|
|
### Confusion Matrix |
|
|
|
|
|
| True \ Predicted | FAIL | PASS | |
|
|
| --- | --- | --- | |
|
|
| **FAIL** | 25 | 10 | |
|
|
| **PASS** | 3 | 32 | |
|
|
|
|
|
<details> |
|
|
<summary><b>Full metrics (JSON)</b></summary> |
|
|
|
|
|
```json |
|
|
{ |
|
|
"FAIL": { |
|
|
"precision": 0.8928571428571429, |
|
|
"recall": 0.7142857142857143, |
|
|
"f1-score": 0.7936507936507936, |
|
|
"support": 35.0 |
|
|
}, |
|
|
"PASS": { |
|
|
"precision": 0.7619047619047619, |
|
|
"recall": 0.9142857142857143, |
|
|
"f1-score": 0.8311688311688312, |
|
|
"support": 35.0 |
|
|
}, |
|
|
"accuracy": 0.8142857142857143, |
|
|
"macro avg": { |
|
|
"precision": 0.8273809523809523, |
|
|
"recall": 0.8142857142857143, |
|
|
"f1-score": 0.8124098124098125, |
|
|
"support": 70.0 |
|
|
}, |
|
|
"weighted avg": { |
|
|
"precision": 0.8273809523809523, |
|
|
"recall": 0.8142857142857143, |
|
|
"f1-score": 0.8124098124098124, |
|
|
"support": 70.0 |
|
|
} |
|
|
} |
|
|
``` |
|
|
</details> |
|
|
|
|
|
|
|
|
<details> |
|
|
<summary><b>Sample Predictions</b></summary> |
|
|
|
|
|
| Text | True Label | Predicted Label | |
|
|
|------|------------|-----------------| |
|
|
| *Mojangsta rap\n\nFTFY | PASS | PASS | |
|
|
| RASHEEEED WALLACE^WALLACE^WALLACE^WALLACE | PASS | PASS | |
|
|
| I've thought of that before a few times. Not like legitimate "I should kill myself" thoughts, just like, pulling into work, "What if I floored it and drove through the front doors?" | FAIL | FAIL | |
|
|
| There is no average. I get all men and women under the sun. Old, young, hot, not so hot, short, tall, wheelchair, celebrities, women.... | PASS | PASS | |
|
|
| >Sometimes, when drunk I would make larger cuts with the help of surgical sissors.\n\naaaand that's enough internet for today. | FAIL | PASS | |
|
|
| *Mojangsta rap\n\nFTFY | PASS | PASS | |
|
|
</details> |
|
|
|
|
|
|
|
|
<details> |
|
|
<summary><b>Prediction Speed Benchmarks</b></summary> |
|
|
|
|
|
| Dataset Size | Time (seconds) | Predictions/Second | |
|
|
|--------------|----------------|---------------------| |
|
|
| 1 | 0.0002 | 5809.29 | |
|
|
| 70 | 0.0073 | 9606.11 | |
|
|
| 70 | 0.0054 | 12990.06 | |
|
|
</details> |
|
|
|
|
|
|
|
|
## Other model variants |
|
|
|
|
|
Below is a general overview of the best-performing models for each dataset variant. |
|
|
|
|
|
| Classifies | Model | Precision | Recall | F1 | |
|
|
| --- | --- | --- | --- | --- | |
|
|
| prompt-harassment-binary | [enguard/tiny-guard-2m-en-prompt-harassment-binary-moderation](https://huggingface.co/enguard/tiny-guard-2m-en-prompt-harassment-binary-moderation) | 0.8788 | 0.7180 | 0.7903 | |
|
|
| prompt-harmfulness-binary | [enguard/tiny-guard-2m-en-prompt-harmfulness-binary-moderation](https://huggingface.co/enguard/tiny-guard-2m-en-prompt-harmfulness-binary-moderation) | 0.8543 | 0.7256 | 0.7847 | |
|
|
| prompt-harmfulness-multilabel | [enguard/tiny-guard-2m-en-prompt-harmfulness-multilabel-moderation](https://huggingface.co/enguard/tiny-guard-2m-en-prompt-harmfulness-multilabel-moderation) | 0.7687 | 0.5006 | 0.6064 | |
|
|
| prompt-hate-speech-binary | [enguard/tiny-guard-2m-en-prompt-hate-speech-binary-moderation](https://huggingface.co/enguard/tiny-guard-2m-en-prompt-hate-speech-binary-moderation) | 0.9141 | 0.7269 | 0.8098 | |
|
|
| prompt-self-harm-binary | [enguard/tiny-guard-2m-en-prompt-self-harm-binary-moderation](https://huggingface.co/enguard/tiny-guard-2m-en-prompt-self-harm-binary-moderation) | 0.8929 | 0.7143 | 0.7937 | |
|
|
| prompt-sexual-content-binary | [enguard/tiny-guard-2m-en-prompt-sexual-content-binary-moderation](https://huggingface.co/enguard/tiny-guard-2m-en-prompt-sexual-content-binary-moderation) | 0.9256 | 0.8141 | 0.8663 | |
|
|
| prompt-violence-binary | [enguard/tiny-guard-2m-en-prompt-violence-binary-moderation](https://huggingface.co/enguard/tiny-guard-2m-en-prompt-violence-binary-moderation) | 0.9017 | 0.7645 | 0.8275 | |
|
|
| prompt-harassment-binary | [enguard/tiny-guard-4m-en-prompt-harassment-binary-moderation](https://huggingface.co/enguard/tiny-guard-4m-en-prompt-harassment-binary-moderation) | 0.8895 | 0.7160 | 0.7934 | |
|
|
| prompt-harmfulness-binary | [enguard/tiny-guard-4m-en-prompt-harmfulness-binary-moderation](https://huggingface.co/enguard/tiny-guard-4m-en-prompt-harmfulness-binary-moderation) | 0.8565 | 0.7540 | 0.8020 | |
|
|
| prompt-harmfulness-multilabel | [enguard/tiny-guard-4m-en-prompt-harmfulness-multilabel-moderation](https://huggingface.co/enguard/tiny-guard-4m-en-prompt-harmfulness-multilabel-moderation) | 0.7924 | 0.5663 | 0.6606 | |
|
|
| prompt-hate-speech-binary | [enguard/tiny-guard-4m-en-prompt-hate-speech-binary-moderation](https://huggingface.co/enguard/tiny-guard-4m-en-prompt-hate-speech-binary-moderation) | 0.9198 | 0.7831 | 0.8460 | |
|
|
| prompt-self-harm-binary | [enguard/tiny-guard-4m-en-prompt-self-harm-binary-moderation](https://huggingface.co/enguard/tiny-guard-4m-en-prompt-self-harm-binary-moderation) | 0.9062 | 0.8286 | 0.8657 | |
|
|
| prompt-sexual-content-binary | [enguard/tiny-guard-4m-en-prompt-sexual-content-binary-moderation](https://huggingface.co/enguard/tiny-guard-4m-en-prompt-sexual-content-binary-moderation) | 0.9371 | 0.8468 | 0.8897 | |
|
|
| prompt-violence-binary | [enguard/tiny-guard-4m-en-prompt-violence-binary-moderation](https://huggingface.co/enguard/tiny-guard-4m-en-prompt-violence-binary-moderation) | 0.8851 | 0.8370 | 0.8603 | |
|
|
| prompt-harassment-binary | [enguard/tiny-guard-8m-en-prompt-harassment-binary-moderation](https://huggingface.co/enguard/tiny-guard-8m-en-prompt-harassment-binary-moderation) | 0.8895 | 0.7767 | 0.8292 | |
|
|
| prompt-harmfulness-binary | [enguard/tiny-guard-8m-en-prompt-harmfulness-binary-moderation](https://huggingface.co/enguard/tiny-guard-8m-en-prompt-harmfulness-binary-moderation) | 0.8627 | 0.7912 | 0.8254 | |
|
|
| prompt-harmfulness-multilabel | [enguard/tiny-guard-8m-en-prompt-harmfulness-multilabel-moderation](https://huggingface.co/enguard/tiny-guard-8m-en-prompt-harmfulness-multilabel-moderation) | 0.7902 | 0.5926 | 0.6773 | |
|
|
| prompt-hate-speech-binary | [enguard/tiny-guard-8m-en-prompt-hate-speech-binary-moderation](https://huggingface.co/enguard/tiny-guard-8m-en-prompt-hate-speech-binary-moderation) | 0.9152 | 0.8233 | 0.8668 | |
|
|
| prompt-self-harm-binary | [enguard/tiny-guard-8m-en-prompt-self-harm-binary-moderation](https://huggingface.co/enguard/tiny-guard-8m-en-prompt-self-harm-binary-moderation) | 0.9667 | 0.8286 | 0.8923 | |
|
|
| prompt-sexual-content-binary | [enguard/tiny-guard-8m-en-prompt-sexual-content-binary-moderation](https://huggingface.co/enguard/tiny-guard-8m-en-prompt-sexual-content-binary-moderation) | 0.9382 | 0.8881 | 0.9125 | |
|
|
| prompt-violence-binary | [enguard/tiny-guard-8m-en-prompt-violence-binary-moderation](https://huggingface.co/enguard/tiny-guard-8m-en-prompt-violence-binary-moderation) | 0.9042 | 0.8551 | 0.8790 | |
|
|
| prompt-harassment-binary | [enguard/small-guard-32m-en-prompt-harassment-binary-moderation](https://huggingface.co/enguard/small-guard-32m-en-prompt-harassment-binary-moderation) | 0.8809 | 0.7964 | 0.8365 | |
|
|
| prompt-harmfulness-binary | [enguard/small-guard-32m-en-prompt-harmfulness-binary-moderation](https://huggingface.co/enguard/small-guard-32m-en-prompt-harmfulness-binary-moderation) | 0.8548 | 0.8239 | 0.8391 | |
|
|
| prompt-harmfulness-multilabel | [enguard/small-guard-32m-en-prompt-harmfulness-multilabel-moderation](https://huggingface.co/enguard/small-guard-32m-en-prompt-harmfulness-multilabel-moderation) | 0.8065 | 0.6494 | 0.7195 | |
|
|
| prompt-hate-speech-binary | [enguard/small-guard-32m-en-prompt-hate-speech-binary-moderation](https://huggingface.co/enguard/small-guard-32m-en-prompt-hate-speech-binary-moderation) | 0.9207 | 0.8394 | 0.8782 | |
|
|
| prompt-self-harm-binary | [enguard/small-guard-32m-en-prompt-self-harm-binary-moderation](https://huggingface.co/enguard/small-guard-32m-en-prompt-self-harm-binary-moderation) | 0.9333 | 0.8000 | 0.8615 | |
|
|
| prompt-sexual-content-binary | [enguard/small-guard-32m-en-prompt-sexual-content-binary-moderation](https://huggingface.co/enguard/small-guard-32m-en-prompt-sexual-content-binary-moderation) | 0.9328 | 0.8847 | 0.9081 | |
|
|
| prompt-violence-binary | [enguard/small-guard-32m-en-prompt-violence-binary-moderation](https://huggingface.co/enguard/small-guard-32m-en-prompt-violence-binary-moderation) | 0.9077 | 0.8913 | 0.8995 | |
|
|
| prompt-harassment-binary | [enguard/medium-guard-128m-xx-prompt-harassment-binary-moderation](https://huggingface.co/enguard/medium-guard-128m-xx-prompt-harassment-binary-moderation) | 0.8660 | 0.8034 | 0.8336 | |
|
|
| prompt-harmfulness-binary | [enguard/medium-guard-128m-xx-prompt-harmfulness-binary-moderation](https://huggingface.co/enguard/medium-guard-128m-xx-prompt-harmfulness-binary-moderation) | 0.8457 | 0.8074 | 0.8261 | |
|
|
| prompt-harmfulness-multilabel | [enguard/medium-guard-128m-xx-prompt-harmfulness-multilabel-moderation](https://huggingface.co/enguard/medium-guard-128m-xx-prompt-harmfulness-multilabel-moderation) | 0.7795 | 0.6516 | 0.7098 | |
|
|
| prompt-hate-speech-binary | [enguard/medium-guard-128m-xx-prompt-hate-speech-binary-moderation](https://huggingface.co/enguard/medium-guard-128m-xx-prompt-hate-speech-binary-moderation) | 0.8826 | 0.8153 | 0.8476 | |
|
|
| prompt-self-harm-binary | [enguard/medium-guard-128m-xx-prompt-self-harm-binary-moderation](https://huggingface.co/enguard/medium-guard-128m-xx-prompt-self-harm-binary-moderation) | 0.9375 | 0.8571 | 0.8955 | |
|
|
| prompt-sexual-content-binary | [enguard/medium-guard-128m-xx-prompt-sexual-content-binary-moderation](https://huggingface.co/enguard/medium-guard-128m-xx-prompt-sexual-content-binary-moderation) | 0.9153 | 0.8744 | 0.8944 | |
|
|
| prompt-violence-binary | [enguard/medium-guard-128m-xx-prompt-violence-binary-moderation](https://huggingface.co/enguard/medium-guard-128m-xx-prompt-violence-binary-moderation) | 0.8821 | 0.8406 | 0.8609 | |
|
|
|
|
|
## Resources |
|
|
|
|
|
- Awesome AI Guardrails: <https://github.com/enguard-ai/awesome-ai-guardails> |
|
|
- Model2Vec: https://github.com/MinishLab/model2vec |
|
|
- Docs: https://minish.ai/packages/model2vec/introduction |
|
|
|
|
|
## Citation |
|
|
|
|
|
If you use this model, please cite Model2Vec: |
|
|
|
|
|
``` |
|
|
@software{minishlab2024model2vec, |
|
|
author = {Stephan Tulkens and {van Dongen}, Thomas}, |
|
|
title = {Model2Vec: Fast State-of-the-Art Static Embeddings}, |
|
|
year = {2024}, |
|
|
publisher = {Zenodo}, |
|
|
doi = {10.5281/zenodo.17270888}, |
|
|
url = {https://github.com/MinishLab/model2vec}, |
|
|
license = {MIT} |
|
|
} |
|
|
``` |
