bfd61325a448bf8c81c380e59ed64e6b

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

• Loss: 3.0120
• Data Size: 1.0
• Epoch Runtime: 13.9057
• Bleu: 9.0551

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	6.8265	0	1.3679	0.6095
No log	1	33	5.6527	0.0078	1.7965	0.8161
No log	2	66	4.6431	0.0156	3.3209	1.4655
No log	3	99	3.6840	0.0312	4.5415	3.5772
0.2537	4	132	3.3921	0.0625	6.3116	4.4263
0.2537	5	165	3.0902	0.125	7.6048	4.9428
0.2537	6	198	2.8140	0.25	9.9491	6.1224
0.5378	7	231	2.6034	0.5	11.5299	6.9438
1.5095	8.0	264	2.4368	1.0	15.0440	7.7531
1.5095	9.0	297	2.4602	1.0	14.5971	8.0104
1.6506	10.0	330	2.6133	1.0	14.5240	8.2361
0.9714	11.0	363	2.7772	1.0	14.8948	8.2189
0.9714	12.0	396	3.0120	1.0	13.9057	9.0551

Framework versions

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