1760b2c0dbe599e29f33fae665f501aa

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

• Loss: 3.1941
• Data Size: 1.0
• Epoch Runtime: 12.1572
• Bleu: 9.3263

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	8.4568	0	1.2919	0.9534
No log	1	29	6.8655	0.0078	2.0146	1.1669
No log	2	58	5.7014	0.0156	3.3347	1.8853
No log	3	87	5.2410	0.0312	4.0915	2.5596
No log	4	116	4.9118	0.0625	5.1515	3.7748
No log	5	145	4.5979	0.125	7.1417	5.2022
0.4982	6	174	4.2284	0.25	9.2535	6.2000
0.4982	7	203	3.5552	0.5	10.4604	6.7708
0.4982	8.0	232	2.4911	1.0	13.0965	7.2000
1.8308	9.0	261	2.4808	1.0	12.2566	8.8940
1.8308	10.0	290	2.5922	1.0	13.3993	9.9278
1.2382	11.0	319	2.7697	1.0	13.3778	10.9352
1.2382	12.0	348	3.0291	1.0	12.1307	8.5481
0.6736	13.0	377	3.1941	1.0	12.1572	9.3263

Framework versions

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