---
license: cc-by-sa-4.0
language:
- vi
library_name: transformers
pipeline_tag: token-classification
tags:
- capitalization
- punctuation
- capu
- vietnamese
---

# vibert-capu

Consolidated CAPU model package for `vit-stt`.

This repository combines:

- CAPU fine-tuned model files from `dragonSwing/vibert-capu`
- ViBERT base tokenizer/config files from `FPTAI/vibert-base-cased`, placed under `base_model/`

The consolidation keeps the runtime model swappable with a single model
directory. `config.json` points to `base_model` through
`pretrained_name_or_path`.

## Intended use

Vietnamese capitalization and punctuation restoration for `vit-stt`
postprocessing.

## Provenance

- Source CAPU model: https://huggingface.co/dragonSwing/vibert-capu
- Source base model: https://huggingface.co/FPTAI/vibert-base-cased

The source CAPU model reports `cc-by-sa-4.0` metadata on Hugging Face. The base
model repository did not expose license metadata through the Hugging Face API at
upload time.