helsinki_new_ver6.1

This model is a fine-tuned version of Helsinki-NLP/opus-mt-en-ZH on the sarahwei/Taiwanese-Minnan-Sutiau dataset. It achieves the following results on the evaluation set:

• Loss: 0.6244
• Bleu: 0.9767
• Ter: 97.2786

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-06
• train_batch_size: 8
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 14000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Ter
0.7463	0.4230	1000	0.7283	0.8156	100.0990
0.7462	0.8460	2000	0.7027	0.8965	99.9010
0.7074	1.2690	3000	0.6837	1.2326	99.4557
0.6903	1.6920	4000	0.6676	0.9807	98.8125
0.6701	2.1151	5000	0.6540	0.9098	98.3177
0.6633	2.5381	6000	0.6416	1.0929	97.9713
0.6627	2.9611	7000	0.6327	0.9645	97.6744
0.6328	3.3841	8000	0.6244	0.9767	97.2786

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.1
• Tokenizers 0.21.1

@misc{sarahwei_taiwanese_minnan2024, author = {Chieh Hui Wei}, title = {Taiwanese Minnan Example Sentences}, year = {2024}, publisher = {Hugging Face}, howpublished = {\url{https://huggingface.co/datasets/sarahwei/Taiwanese-Minnan-Example-Sentences}} }