tatoeba-ru-tok

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

• Loss: 0.5932
• Bleu: 47.6666

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
1.0515	1.0	1167	0.8539	37.3803
0.8186	2.0	2334	0.7284	41.5032
0.7002	3.0	3501	0.6803	43.5555
0.6501	4.0	4668	0.6485	45.0023
0.6091	5.0	5835	0.6302	45.6329
0.5778	6.0	7002	0.6180	45.8879
0.553	7.0	8169	0.6109	46.6945
0.533	8.0	9336	0.6041	46.6169
0.5128	9.0	10503	0.6002	47.0549
0.5015	10.0	11670	0.5961	47.2017
0.4851	11.0	12837	0.5962	47.5851
0.4795	12.0	14004	0.5939	47.5400
0.4659	13.0	15171	0.5932	47.6666
0.4608	14.0	16338	0.5939	47.6703
0.4593	15.0	17505	0.5936	47.6572

Framework versions

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