Upload ASTRA trained model

README.md

@@ -0,0 +1,59 @@
+---
+license: mit
+tags:
+- text-normalization
+- vietnamese
+- lexical-normalization
+- astra
+- visobert
+pipeline_tag: text2text-generation
+---
+
+# hadung1802/visobert-normalizer
+
+This model is a Vietnamese text normalization model trained using the ASTRA framework with VISOBERT architecture.
+
+## Model Description
+
+This model performs lexical normalization for Vietnamese text, converting informal text to standard Vietnamese. It was trained using the ASTRA (Self-training with Weak Supervision) framework.
+
+## Performance
+
+
+## Training Configuration
+
+- **Student Model**: VISOBERT
+- **Training Mode**: weakly_supervised
+- **Learning Rate**: 0.001
+- **Epochs**: 10
+- **Batch Size**: 16
+
+## Usage
+
+```python
+from transformers import AutoTokenizer, AutoModel
+import torch
+
+# Load model and tokenizer
+tokenizer = AutoTokenizer.from_pretrained("hadung1802/visobert-normalizer")
+model = AutoModel.from_pretrained("hadung1802/visobert-normalizer")
+
+# Example usage
+text = "toi di hoc"
+inputs = tokenizer(text, return_tensors="pt")
+with torch.no_grad():
+    outputs = model(**inputs)
+```
+
+## Citation
+
+If you use this model, please cite the ASTRA paper:
+
+```bibtex
+@article{astra2024,
+  title={ASTRA: Self-training with Weak Supervision for Vietnamese Text Normalization},
+  author={Your Name},
+  journal={arXiv preprint},
+  year={2024}
+}
+```