<!-- 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. -->
lg-en-test-version
This model is a fine-tuned version of AI-Lab-Makerere/lg_en on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.5803
- Bleu: 31.3111
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: 9.687717341785184e-05
- train_batch_size: 15
- eval_batch_size: 2
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 4
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu |
|---|---|---|---|---|
| No log | 1.0 | 24 | 1.0100 | 28.5722 |
| No log | 2.0 | 48 | 0.7758 | 27.7506 |
| No log | 3.0 | 72 | 0.6459 | 40.3866 |
| No log | 4.0 | 96 | 0.5803 | 31.3111 |
Framework versions
- Transformers 4.21.2
- Pytorch 1.12.1+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1
