pokemon-finetuned-opus-mt-en-de

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

• Loss: 0.0554
• Exact Match: 0.9893

Model description

This model is similar to the Helsinki-NLP/opus-mt-en-de but it now properly translates Pokemon names.

Intended uses & limitations

This model is part of this tutorial repository. It is only intended as a proof-of-concept and is not intended for legitimate usage or deployment.

This model has not been tested to see if the fine-tuning process changed anything beyond a few Pokemon-related phrases.

Training and evaluation data

The model was purposely overfit toward the training data, which was a list of translated Pokemon names from this forum post

Training procedure

The evaluation and training sets were the same given a list of translated Pokemon names.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1