t5-small-finetuned-de-en-epochs5
This model is a fine-tuned version of t5-small on the wmt14 dataset. It achieves the following results on the evaluation set:
- Loss: 2.2040
- Bleu: 5.8913
- Gen Len: 17.5408
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: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu | Gen Len |
|---|---|---|---|---|---|
| No log | 1.0 | 188 | 2.3366 | 2.8075 | 17.8188 |
| No log | 2.0 | 376 | 2.2557 | 4.8765 | 17.626 |
| 2.6928 | 3.0 | 564 | 2.2246 | 5.5454 | 17.5534 |
| 2.6928 | 4.0 | 752 | 2.2086 | 5.8511 | 17.5461 |
| 2.6928 | 5.0 | 940 | 2.2040 | 5.8913 | 17.5408 |
Framework versions
- Transformers 4.12.5
- Pytorch 1.10.0+cu111
- Datasets 1.16.1
- Tokenizers 0.10.3