3bb9500fd174ca3632fbff99561c102f

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

• Loss: 2.5831
• Data Size: 1.0
• Epoch Runtime: 24.7380
• Bleu: 23.9221

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	4.5533	0	2.2762	8.3962
No log	1	77	3.3074	0.0078	2.6959	14.8220
No log	2	154	3.0184	0.0156	4.5366	12.0333
No log	3	231	2.8836	0.0312	6.3037	9.0668
No log	4	308	2.7240	0.0625	7.9018	8.8327
No log	5	385	2.4019	0.125	10.4028	9.3552
0.2323	6	462	2.1517	0.25	11.9249	10.5837
0.8154	7	539	2.0405	0.5	15.1691	12.0843
1.5338	8.0	616	1.9510	1.0	24.8324	23.6657
1.1974	9.0	693	2.0065	1.0	23.9615	25.0450
0.6839	10.0	770	2.1927	1.0	23.6039	25.4024
0.5332	11.0	847	2.3475	1.0	24.5288	23.6693
0.2517	12.0	924	2.5831	1.0	24.7380	23.9221

Framework versions

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