model-translate-ar-to-en-from-320k-dataset-ar-en-th2301191019

This model is a fine-tuned version of Helsinki-NLP/opus-mt-ar-en on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.2166
• Bleu: 37.7938

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: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
1.3931	1.0	18750	1.2483	36.9803
1.2587	2.0	37500	1.2246	37.5016
1.0638	3.0	56250	1.2166	37.7938

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.1+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2