cantonese chinese translation

Introduction

The Cantonese-Chinese Translation model is a fine-tuned version of the fnlp/bart-base-chinese model, specifically designed for translating between Cantonese and Traditional Chinese. It was fine-tuned using the raptorkwok/cantonese-traditional-chinese-parallel-corpus dataset, achieving significant evaluation metrics such as a BLEU score of 62.1085.

Architecture

The model is based on the BART architecture, optimized for text-to-text generation tasks. It leverages the strengths of the fnlp/bart-base-chinese model, a robust foundation for Chinese language processing tasks.

Training

Training Parameters

The model was trained using the following hyperparameters:

• Learning Rate: 2e-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
• Number of Epochs: 30
• Mixed Precision Training: Native AMP

Training Results

The model demonstrated progressive improvements across training steps, with a final validation loss of 0.2258 and consistent performance metrics. The BLEU score and CHRF scores indicate high-quality translation outputs.

Framework Versions

• Transformers: 4.28.1
• PyTorch: 2.1.0+cu121
• Datasets: 2.14.6
• Tokenizers: 0.13.3

Guide: Running Locally

To run the Cantonese-Chinese Translation model locally, follow these steps:

1. Install Dependencies:

   pip install transformers datasets torch
   

2. Load the Model:

   from transformers import pipeline
   model = pipeline("translation", model="raptorkwok/cantonese-chinese-translation")
   

3. Run a Translation:

   result = model("Your Cantonese text here")
   print(result)
   

4. Cloud GPUs: For efficient processing, consider using cloud services like AWS, GCP, or Azure, which offer GPU instances.

License

The licensing terms for the Cantonese-Chinese Translation model are not explicitly stated in the provided document. It is recommended to review the model card on Hugging Face for detailed licensing information.