63e8f5a6116a6e8efeea42f20b6a3215

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

• Loss: 3.2574
• Data Size: 1.0
• Epoch Runtime: 20.4562
• Bleu: 4.8740

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	7.8433	0	1.9679	0.1627
No log	1	58	5.8522	0.0078	2.7659	0.2789
No log	2	116	4.9739	0.0156	3.0207	1.0557
No log	3	174	4.6023	0.0312	5.1266	1.5034
No log	4	232	4.3223	0.0625	6.4048	2.2412
No log	5	290	4.0106	0.125	8.3149	2.6710
0.3849	6	348	3.6638	0.25	10.0080	3.4947
0.5056	7	406	2.9744	0.5	13.5005	3.8095
1.9979	8.0	464	3.2876	1.0	20.3764	3.8362
10.0914	9.0	522	4.9917	1.0	20.4860	0.2755
5.7253	10.0	580	2.6004	1.0	20.5950	4.6070
1.9679	11.0	638	2.6251	1.0	19.0830	4.7190
1.6579	12.0	696	2.8071	1.0	19.0702	4.7664
1.0424	13.0	754	2.9515	1.0	19.9845	4.7910
0.8032	14.0	812	3.2574	1.0	20.4562	4.8740

Framework versions

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