7d99a3137d2a1e483ee0f9eb13b8250e

This model is a fine-tuned version of facebook/mbart-large-50-one-to-many-mmt on the Helsinki-NLP/opus_books [it-nl] dataset. It achieves the following results on the evaluation set:

• Loss: 3.1663
• Data Size: 1.0
• Epoch Runtime: 20.0388
• Bleu: 6.5272

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Bleu
No log	0	0	8.9717	0	1.9525	0.0883
No log	1	58	6.2651	0.0078	2.6200	0.8747
No log	2	116	5.4976	0.0156	3.1003	0.6217
No log	3	174	4.6468	0.0312	4.9905	0.8353
No log	4	232	4.0749	0.0625	6.3726	1.3796
No log	5	290	3.6770	0.125	8.0437	2.0889
0.3696	6	348	3.3378	0.25	9.9280	2.9668
0.464	7	406	3.0660	0.5	13.3932	3.4875
2.0464	8.0	464	2.8552	1.0	19.7944	4.2182
2.3879	9.0	522	2.7952	1.0	20.4597	5.0087
2.062	10.0	580	2.8244	1.0	20.3756	5.5366
1.8167	11.0	638	2.9068	1.0	18.6699	5.5536
1.5746	12.0	696	3.0349	1.0	19.1370	7.2565
1.1396	13.0	754	3.1663	1.0	20.0388	6.5272

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.2.0
• Tokenizers 0.22.1