Updating stylization
Browse files
README.md
CHANGED
|
@@ -11,19 +11,6 @@ A tiny comment toxicity classifier model at only 2M parameters. With only ~10MB
|
|
| 11 |
A paper on this model is being released soon.
|
| 12 |
|
| 13 |
|
| 14 |
-
## Benchmarks
|
| 15 |
-
|
| 16 |
-
The Tiny-Toxic-Detector achieves an impressive 90.26% on the Toxigen benchmark and 87.34% on the Jigsaw-Toxic-Comment-Classification-Challenge. Here we compare our results against other toxic classification models:
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
| Model | Size (parameters) | Toxigen (%) | Jigsaw (%) |
|
| 20 |
-
| --------------------------------- | ----------------- | ----------- | ---------- |
|
| 21 |
-
| lmsys/toxicchat-t5-large-v1.0 | 738M | 72.67 | 88.82 |
|
| 22 |
-
| s-nlp/roberta toxicity classifier | 124M | 88.41 | **94.92** |
|
| 23 |
-
| mohsenfayyaz/toxicity-classifier | 109M | 81.50 | 83.31 |
|
| 24 |
-
| martin-ha/toxic-comment-model | 67M | 68.02 | 91.56 |
|
| 25 |
-
| **Tiny-toxic-detector** | **2M** | **90.97** | 86.98 |
|
| 26 |
-
|
| 27 |
|
| 28 |
## Usage
|
| 29 |
This model uses custom architecture and requires some extra custom code to work. Below you can find the architecture and a fully-usable example.
|
|
@@ -204,6 +191,20 @@ with torch.no_grad():
|
|
| 204 |
prediction = "Toxic" if logits > 0.5 else "Not Toxic"
|
| 205 |
```
|
| 206 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 207 |
|
| 208 |
## Usage and Limitations
|
| 209 |
|
|
|
|
| 11 |
A paper on this model is being released soon.
|
| 12 |
|
| 13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
|
| 15 |
## Usage
|
| 16 |
This model uses custom architecture and requires some extra custom code to work. Below you can find the architecture and a fully-usable example.
|
|
|
|
| 191 |
prediction = "Toxic" if logits > 0.5 else "Not Toxic"
|
| 192 |
```
|
| 193 |
|
| 194 |
+
## Benchmarks
|
| 195 |
+
|
| 196 |
+
The Tiny-Toxic-Detector achieves an impressive 90.26% on the Toxigen benchmark and 87.34% on the Jigsaw-Toxic-Comment-Classification-Challenge. Here we compare our results against other toxic classification models:
|
| 197 |
+
|
| 198 |
+
|
| 199 |
+
| Model | Size (parameters) | Toxigen (%) | Jigsaw (%) |
|
| 200 |
+
| --------------------------------- | ----------------- | ----------- | ---------- |
|
| 201 |
+
| lmsys/toxicchat-t5-large-v1.0 | 738M | 72.67 | 88.82 |
|
| 202 |
+
| s-nlp/roberta toxicity classifier | 124M | 88.41 | **94.92** |
|
| 203 |
+
| mohsenfayyaz/toxicity-classifier | 109M | 81.50 | 83.31 |
|
| 204 |
+
| martin-ha/toxic-comment-model | 67M | 68.02 | 91.56 |
|
| 205 |
+
| **Tiny-toxic-detector** | **2M** | **90.97** | 86.98 |
|
| 206 |
+
|
| 207 |
+
|
| 208 |
|
| 209 |
## Usage and Limitations
|
| 210 |
|