en-ta-marian

This model is a fine-tuned version of Helsinki-NLP/opus-mt-en-dra on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2338
• Bleu: 80.6659

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: 5e-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: 16
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
No log	1.0	32	2.4300	2.2293
No log	2.0	64	1.8038	6.3360
No log	3.0	96	1.4173	19.6274
No log	4.0	128	1.1478	27.8938
No log	5.0	160	0.9330	35.1550
No log	6.0	192	0.7652	42.3234
No log	7.0	224	0.6344	48.9213
No log	8.0	256	0.5253	55.6715
No log	9.0	288	0.4450	61.4338
No log	10.0	320	0.3787	66.4522
No log	11.0	352	0.3337	70.3115
No log	12.0	384	0.2961	74.5026
No log	13.0	416	0.2663	77.6865
No log	14.0	448	0.2491	79.4044
No log	15.0	480	0.2375	80.5948
1.4043	16.0	512	0.2338	80.6659

Framework versions

• Transformers 4.40.2
• Pytorch 2.1.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1