<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
marian-finetuned-tgl-eng-netspeak-trial6
This model is a fine-tuned version of Helsinki-NLP/opus-mt-tl-en on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 2.3343
- Bleu: 28.3769
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: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 150
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu |
|---|---|---|---|---|
| 4.5424 | 1.0 | 57 | 3.8496 | 5.8759 |
| 3.6414 | 2.0 | 114 | 3.5073 | 9.0838 |
| 3.1777 | 3.0 | 171 | 3.2660 | 10.0099 |
| 2.7865 | 4.0 | 228 | 3.0646 | 12.0484 |
| 2.4638 | 5.0 | 285 | 2.9109 | 14.0737 |
| 2.1815 | 6.0 | 342 | 2.7853 | 16.3752 |
| 1.9426 | 7.0 | 399 | 2.6832 | 17.7971 |
| 1.725 | 8.0 | 456 | 2.6060 | 19.5480 |
| 1.5336 | 9.0 | 513 | 2.5433 | 20.8335 |
| 1.3571 | 10.0 | 570 | 2.4888 | 21.9160 |
| 1.2081 | 11.0 | 627 | 2.4424 | 21.7108 |
| 1.0733 | 12.0 | 684 | 2.4045 | 23.9640 |
| 0.9516 | 13.0 | 741 | 2.3940 | 24.4162 |
| 0.8487 | 14.0 | 798 | 2.3840 | 27.1127 |
| 0.7513 | 15.0 | 855 | 2.3563 | 27.3229 |
| 0.662 | 16.0 | 912 | 2.3501 | 25.8083 |
| 0.5835 | 17.0 | 969 | 2.3506 | 27.0424 |
| 0.5247 | 18.0 | 1026 | 2.3355 | 27.9392 |
| 0.4648 | 19.0 | 1083 | 2.3379 | 27.1880 |
| 0.4047 | 20.0 | 1140 | 2.3343 | 28.3769 |
| 0.3574 | 21.0 | 1197 | 2.3431 | 27.9125 |
| 0.3183 | 22.0 | 1254 | 2.3407 | 29.3798 |
| 0.2828 | 23.0 | 1311 | 2.3408 | 30.5316 |
| 0.2528 | 24.0 | 1368 | 2.3368 | 29.9854 |
| 0.2306 | 25.0 | 1425 | 2.3603 | 30.4071 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2
