a6b4afc22cf2f6175e5bd2c75f920cf2

This model is a fine-tuned version of facebook/mbart-large-cc25 on the Helsinki-NLP/opus_books [en-nl] dataset. It achieves the following results on the evaluation set:

• Loss: 2.1829
• Data Size: 1.0
• Epoch Runtime: 252.9514
• Bleu: 7.9792

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	9.8204	0	21.7157	0.1674
No log	1	966	3.8104	0.0078	23.6025	2.6623
No log	2	1932	3.3458	0.0156	26.9259	3.0940
0.086	3	2898	2.9537	0.0312	30.6678	3.7471
2.9327	4	3864	2.5997	0.0625	38.4073	4.1741
2.5989	5	4830	2.3820	0.125	52.0540	4.9030
2.3683	6	5796	2.2250	0.25	80.4614	6.3588
2.1061	7	6762	2.0845	0.5	137.5801	6.6802
1.8663	8.0	7728	1.9650	1.0	252.3733	7.4014
1.8863	9.0	8694	1.9836	1.0	251.2755	10.5893
1.4893	10.0	9660	2.0022	1.0	252.0327	8.6716
1.2003	11.0	10626	2.0552	1.0	253.1820	9.1781
1.0212	12.0	11592	2.1829	1.0	252.9514	7.9792

Framework versions

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