5a621a2ee9af95198a887713042d601b

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

• Loss: 3.0167
• Data Size: 1.0
• Epoch Runtime: 183.1962
• Bleu: 5.0547

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	7.0420	0	15.4678	0.4620
No log	1	721	5.4715	0.0078	17.0562	0.8030
No log	2	1442	4.6459	0.0156	19.2720	1.0898
0.0927	3	2163	4.1670	0.0312	23.2617	1.5126
0.2872	4	2884	3.7650	0.0625	28.6891	2.1195
3.6502	5	3605	3.4244	0.125	40.1512	2.7620
3.1881	6	4326	3.1032	0.25	60.2641	3.4779
2.851	7	5047	2.8161	0.5	101.9666	4.1364
2.4288	8.0	5768	2.5884	1.0	187.2732	5.3613
2.053	9.0	6489	2.5594	1.0	183.3586	5.5472
1.797	10.0	7210	2.6200	1.0	183.7364	5.2350
1.4898	11.0	7931	2.6558	1.0	184.6197	5.2496
1.2664	12.0	8652	2.8293	1.0	183.2790	4.9465
1.0396	13.0	9373	3.0167	1.0	183.1962	5.0547

Framework versions

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