38589364f51cabf2923978b3e42ae5b3

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

• Loss: 3.4488
• Data Size: 1.0
• Epoch Runtime: 24.3730
• Bleu: 3.4661

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	10.9149	0	2.7893	0.0293
No log	1	83	6.8834	0.0078	3.0049	0.1750
No log	2	166	4.6365	0.0156	3.9639	0.7971
No log	3	249	4.1988	0.0312	5.4995	1.3900
0.1711	4	332	3.9959	0.0625	7.3553	2.3859
0.1711	5	415	3.5855	0.125	9.2971	2.2028
0.1711	6	498	3.3341	0.25	11.4200	2.3001
0.5978	7	581	2.9201	0.5	15.2931	3.8195
2.6328	8.0	664	2.7256	1.0	26.2070	3.5757
2.1443	9.0	747	2.7218	1.0	26.5309	3.5216
1.5085	10.0	830	2.8476	1.0	24.0704	3.9500
1.1454	11.0	913	3.0918	1.0	24.5482	4.3021
0.8967	12.0	996	3.2989	1.0	24.6792	3.5121
0.6196	13.0	1079	3.4488	1.0	24.3730	3.4661

Framework versions

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