README.md

metadata

license: apache-2.0
language:
  - zh
  - en
model-index:
  - name: skNMT-zh-en-1.2
    results:
      - task:
          type: translation
        metrics:
          - name: BLEU
            type: BLEU
            value: 20.4218
          - name: chrF
            type: chrf
            value: 50.2827
        dataset:
          name: WMT 2019
          type: wmt/wmt19

skNMT-zh-en-1.2

The NMT (Neural Machine Translation) model trained by sparkastML for translating from Chinese to English.

This model use OpenNMT as its underlying structure.

Usage

We have already exported the model into CTranslate2-compatible format. You can download the necessary files (model.bin, config.json and shared_vocabulary.json), and start with the CTranslate2.

We alsow provide the training checkpoint and the sentencepiece model, so you can manually inference via OpenNMT.

Model Details

• Source Language: Chinese (Simplified)
• Target Language: English
• Training Time: Totally 11.3 hours, 46,500 steps (~1×10¹⁸ FLOPs)
• Training Device:
  • RTX 3080 (20GB): step 0-20,000
  • RTX 4070: step 20,000-46,500
• Corpus Size: Over 10 million sentences
• Validation BLEU Score: 21.28
• Validation Loss (Cross Entropy): 3.152