tatoeba-tok-en

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: 1.4103
  • Bleu: 24.8720

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
1.9475 1.0 1167 1.7238 18.0675
1.6448 2.0 2334 1.5780 20.8358
1.4672 3.0 3501 1.5123 22.5640
1.3647 4.0 4668 1.4765 23.4342
1.2831 5.0 5835 1.4523 24.3822
1.2213 6.0 7002 1.4393 24.6540
1.1609 7.0 8169 1.4295 25.0382
1.1231 8.0 9336 1.4235 24.8617
1.076 9.0 10503 1.4165 23.6289
1.0489 10.0 11670 1.4162 25.8471
1.0174 11.0 12837 1.4112 25.1640
0.9961 12.0 14004 1.4114 22.6132
0.9735 13.0 15171 1.4110 25.1948
0.9597 14.0 16338 1.4103 23.6949
0.9519 15.0 17505 1.4107 24.2304

Framework versions

  • Transformers 4.52.4
  • Pytorch 2.7.1+cu128
  • Datasets 3.6.0
  • Tokenizers 0.21.1
Downloads last month
8
Safetensors
Model size
76.2M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for NetherQuartz/tatoeba-tok-en

Finetuned
(3)
this model

Dataset used to train NetherQuartz/tatoeba-tok-en

Collection including NetherQuartz/tatoeba-tok-en