679a334fad065e821a8b103d3efb2654

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

• Loss: 3.3991
• Data Size: 1.0
• Epoch Runtime: 12.5562
• Bleu: 11.6186

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	7.9873	0	1.3158	0.8536
No log	1	29	6.7276	0.0078	1.9904	0.7781
No log	2	58	5.7029	0.0156	3.3846	1.1791
No log	3	87	4.8670	0.0312	4.4419	1.7688
No log	4	116	4.1832	0.0625	5.3234	3.3335
No log	5	145	3.6579	0.125	7.4559	3.7247
0.4177	6	174	3.2451	0.25	9.1291	5.1020
0.4177	7	203	2.9118	0.5	10.5136	6.0127
0.4177	8.0	232	2.7107	1.0	14.0670	7.3544
1.5622	9.0	261	2.6912	1.0	13.2080	9.8501
1.5622	10.0	290	2.7785	1.0	13.5425	9.9462
1.3654	11.0	319	2.9761	1.0	13.8556	10.4719
1.3654	12.0	348	3.1769	1.0	14.8348	12.7115
0.7984	13.0	377	3.3991	1.0	12.5562	11.6186

Framework versions

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