metadata
tags:
- generated_from_trainer
metrics:
- bleu
model-index:
- name: lg-en-v3
results: []
lg-en-v3
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: 0.9289
- Bleu: 32.5138
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: 4.4271483249908667e-05
- train_batch_size: 14
- eval_batch_size: 6
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu |
|---|---|---|---|---|
| No log | 1.0 | 26 | 1.0323 | 32.6278 |
| No log | 2.0 | 52 | 0.9289 | 32.5138 |
Framework versions
- Transformers 4.21.2
- Pytorch 1.12.1+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1