ruby-rails-fine-tuned-ja-en

This model is a fine-tuned version of Helsinki-NLP/opus-mt-ja-en. It achieves the following results on the evaluation set:

• Loss: 1.6714
• Bleu: 0.3300
• Chrf: 64.8493

Model description

Fine-tuned for Japanese to English translation of the Ruby and Rails documentation.

Intended uses & limitations

The training dataset (see below) is still very small, so only minor improvements are expected.

Training and evaluation data

Trained on a custom dataset specifically based on Ruby and Ruby on Rails documentation: https://huggingface.co/datasets/morinoko-inari/ruby-rails-ja-en

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Chrf
No log	1.0	23	1.7593	0.3283	64.8523
No log	2.0	46	1.6915	0.3260	64.6705
No log	3.0	69	1.6714	0.3300	64.8493

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0
• Datasets 3.5.0
• Tokenizers 0.21.1