3387290c87ceeb13343ac94fc9729a65

This model is a fine-tuned version of facebook/mbart-large-cc25 on the Helsinki-NLP/opus_books [en-it] dataset. It achieves the following results on the evaluation set:

• Loss: 2.4899
• Data Size: 1.0
• Epoch Runtime: 215.7689
• Bleu: 6.8512

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	9.8093	0	18.9657	0.1812
No log	1	808	3.6526	0.0078	21.3804	3.3729
No log	2	1616	3.2907	0.0156	22.6844	3.0720
No log	3	2424	2.9847	0.0312	26.1737	3.4109
0.1043	4	3232	2.6554	0.0625	33.5671	8.0032
2.7588	5	4040	2.4717	0.125	46.3061	7.3715
2.4344	6	4848	2.3054	0.25	71.9575	5.3641
9.2673	7	5656	7.5923	0.5	120.6124	0.0
2.0816	8.0	6464	2.0882	1.0	219.6691	6.3018
1.747	9.0	7272	2.0715	1.0	216.4037	6.3158
1.4949	10.0	8080	2.1297	1.0	216.7358	6.6091
1.2665	11.0	8888	2.1715	1.0	215.7898	6.6701
1.0301	12.0	9696	2.3285	1.0	217.1741	6.5242
0.8508	13.0	10504	2.4899	1.0	215.7689	6.8512

Framework versions

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