Commit
•
11d1a6f
1
Parent(s):
e5e9476
Update README.md
Browse files
README.md
CHANGED
|
@@ -16,6 +16,21 @@ The Token-Level Bias Classifier is a transformer-based model developed to detect
|
|
| 16 |
|
| 17 |
The model is built using the `distilbert-base-uncased` pretrained model, a smaller and faster version of BERT. It is fine-tuned on a custom dataset for the task of token-level bias classification. The model uses a Token Classification architecture, typically used for Named Entity Recognition (NER) tasks.
|
| 18 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
## Classes
|
| 20 |
|
| 21 |
The model identifies nine classes, including:
|
|
|
|
| 16 |
|
| 17 |
The model is built using the `distilbert-base-uncased` pretrained model, a smaller and faster version of BERT. It is fine-tuned on a custom dataset for the task of token-level bias classification. The model uses a Token Classification architecture, typically used for Named Entity Recognition (NER) tasks.
|
| 18 |
|
| 19 |
+
## Model Performance
|
| 20 |
+
|
| 21 |
+
| Metric | Value |
|
| 22 |
+
|------------------------|-------------------------|
|
| 23 |
+
| eval_loss | 0.03554883599281311 |
|
| 24 |
+
| eval_precision | 0.7868185694908753 |
|
| 25 |
+
| eval_recall | 0.7662314481801649 |
|
| 26 |
+
| eval_f1 | 0.7739129932274338 |
|
| 27 |
+
| eval_balanced accuracy | 0.7662314481801649 |
|
| 28 |
+
| eval_runtime | 4.5554 |
|
| 29 |
+
| eval_samples_per_second| 1196.818 |
|
| 30 |
+
| eval_steps_per_second | 74.856 |
|
| 31 |
+
| epoch | 6.0 |
|
| 32 |
+
|
| 33 |
+
|
| 34 |
## Classes
|
| 35 |
|
| 36 |
The model identifies nine classes, including:
|