README.md

metadata

language: nl
license: mit
tags:
  - dutch
  - bert
  - regression
  - text-quality
  - transformers
datasets: []
metrics:
  - rmse
  - r2
model-index:
  - name: transformer_y_quality
    results:
      - task:
          type: text-regression
          name: Text Regression
        dataset:
          name: Custom Dataset
          type: text
        metrics:
          - name: RMSE
            type: rmse
            value: 0.078
          - name: R²
            type: r2
            value: 0.8367

🚀 transformer_y_quality — Dutch BERT for Text Quality Regression

Model: GroNLP/bert-base-dutch-cased finetuned for regression
Task: Predict y_quality_simple (text quality score)
Language: Dutch 🇳🇱
Problem type: Single output regression

---

📈 Performance

Epoch	Train Loss	Val Loss	RMSE	R²
1	0.007200	0.007010	0.0837	0.8117
2	0.005500	0.006300	0.0794	0.8307
3	0.004600	0.006079	0.0780	0.8367
4	0.003300	0.006122	0.0782	0.8355
5	0.002600	0.006891	0.0830	0.8149

✅ Final Test Metrics:

• RMSE: 0.0780
• R²: 0.8367

---

⚙️ How to use

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("YourUsername/transformer_y_quality")
model = AutoModelForSequenceClassification.from_pretrained("YourUsername/transformer_y_quality")