679b42752bb5caf0424fe54865d7b56a

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

• Loss: 2.2585
• Data Size: 1.0
• Epoch Runtime: 247.6036
• Bleu: 11.3291

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	4.5453	0	20.4632	6.9282
No log	1	966	3.2934	0.0078	22.5842	6.1692
No log	2	1932	2.6798	0.0156	25.8940	10.5486
0.0655	3	2898	2.5016	0.0312	29.6719	8.7352
2.4084	4	3864	2.3449	0.0625	36.7494	11.1081
2.2468	5	4830	2.2009	0.125	50.8562	13.2254
2.0981	6	5796	2.0725	0.25	78.1799	10.1385
1.8541	7	6762	1.9510	0.5	133.1720	12.9979
1.6479	8.0	7728	1.8467	1.0	243.3393	10.1086
1.3849	9.0	8694	1.8596	1.0	248.7283	9.0653
1.1489	10.0	9660	1.9468	1.0	250.4793	11.0431
0.8969	11.0	10626	2.0601	1.0	251.6170	10.2699
0.7249	12.0	11592	2.2585	1.0	247.6036	11.3291

Framework versions

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