tatoeba-en-tok

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

• Loss: 0.4513
• Bleu: 49.2199

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: 2e-05
• train_batch_size: 64
• eval_batch_size: 64
• 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
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
0.8381	1.0	1167	0.6677	38.6270
0.6401	2.0	2334	0.5611	42.8112
0.5453	3.0	3501	0.5228	44.9041
0.5046	4.0	4668	0.4977	46.2278
0.474	5.0	5835	0.4806	47.2086
0.4466	6.0	7002	0.4723	47.5220
0.4274	7.0	8169	0.4662	48.3719
0.4134	8.0	9336	0.4587	48.4629
0.3949	9.0	10503	0.4593	48.8579
0.3864	10.0	11670	0.4537	48.6287
0.375	11.0	12837	0.4546	48.8812
0.3692	12.0	14004	0.4522	49.1093
0.3608	13.0	15171	0.4524	49.1794
0.3553	14.0	16338	0.4513	49.2199
0.3533	15.0	17505	0.4518	49.2096

Framework versions

• Transformers 4.52.4
• Pytorch 2.7.1+cu128
• Datasets 3.6.0
• Tokenizers 0.21.1