bart-base-cantonese / README.md
indiejoseph's picture
Update README.md
4c9cbd8
|
raw
history blame
1.61 kB
metadata
base_model: fnlp/bart-base-chinese
tags:
  - generated_from_trainer
model-index:
  - name: bart-base-cantonese
    results: []
datasets:
  - indiejoseph/wikipedia-zh-yue-filtered
  - indiejoseph/cc100-yue
  - indiejoseph/ted-transcriptions-cantonese
  - indiejoseph/c4-cantonese-filtered
  - mozilla-foundation/common_voice_13_0
  - jed351/rthk_news
  - jed351/shikoto_zh_hk
widget:
  - text: 今日去咗旺角[MASK]
    example_title: Mong Kok
  - text: 今時今日香港係一個[MASK]。
    example_title: Hong Kong

bart-base-cantonese

This model is a continue pre-train version of fnlp/bart-base-chinese on filtered Cantonese common crawl dataset with 950M tokens.

This tokenizer has extended the Bert tokenizer from fnlp/bart-base-chinese with 500 more Chinese characters commonly found in Cantonese

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 1.0

Training results

Framework versions

  • Transformers 4.35.2
  • Pytorch 2.1.1+cu121
  • Datasets 2.14.6
  • Tokenizers 0.14.1