xlmr-zu-sentiment-analysis
Fine-tuned XLM-R for isiZulu sentiment (negative / neutral / positive).
Labels
{0: 'negative', 1: 'neutral', 2: 'positive'}
Training
- Split: 80/10/10 from in-domain data
- Epochs: 10
- Early stopping: yes
In-domain Test Metrics
- Accuracy: 0.8741976893453145
- Precision (macro): 0.8712608559194496
- Recall (macro): 0.8746246046623661
- F1 (macro): 0.8727454664711543
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
๐
Ask for provider support