mBART-jp-ch-anime-finetune

This is a Japanese-to-Chinese translation model finetuned based on mbart-large-50-many-to-many-mmt. The model is specifically optimized for translating anime-related content, ensuring accurate and context-aware translations between Japanese and Chinese.

---

Overview

• Base Model : mbart-large-50-many-to-many-mmt
• Fine-tuning Domain : Anime-related text (e.g., subtitles, dialogues, descriptions)
• Training Script : run.py
• Testing Script : test.py
• Dataset : Located in split_datasets/
• Generated Data : Stored in data_test/

---

Features

• High-Quality Translations : Tailored for anime-related content, capturing nuances and cultural context.
• Efficient Fine-Tuning : Leverages the multilingual capabilities of mBART-large-50.
• Easy to Use : Simple scripts for training and testing.

---

Usage

Training

To fine-tune the model, run the following command:

python run.py

The training data should be placed in the split_datasets/ directory.

Testing

To evaluate the model, use the test script:

python test.py

The generated translations will be saved in the data_test/ directory.

Application

To generate subtitles for video, run

python video_to_audio.py path_to_your_video

---

License

This project is licensed under the same terms as the base model, mbart-large-50-many-to-many-mmt. Refer to the original license for more details.

---

Acknowledgments

• Thanks to Facebook AI for the mbart-large-50-many-to-many-mmt model.
• Special thanks to the contributors of the anime dataset used for fine-tuning.

---

For questions or contributions, please open an issue or submit a pull request.