ec0656617830e1be131dbab8c38aa45c

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

• Loss: 2.1641
• Data Size: 1.0
• Epoch Runtime: 249.0948
• Bleu: 8.7838

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	5.0992	0	20.4278	2.0359
No log	1	966	2.9414	0.0078	22.4835	4.3322
No log	2	1932	2.6661	0.0156	26.2193	5.1031
0.0633	3	2898	2.5090	0.0312	29.6876	5.9133
2.4258	4	3864	2.3568	0.0625	37.4653	7.0149
2.2666	5	4830	2.2158	0.125	51.5678	8.1599
2.1221	6	5796	2.0845	0.25	79.6498	9.4095
1.8688	7	6762	1.9567	0.5	136.2487	10.7707
1.6574	8.0	7728	1.8533	1.0	248.8535	8.7222
1.394	9.0	8694	1.8611	1.0	247.6628	9.4444
1.1574	10.0	9660	1.9259	1.0	247.3566	9.8484
0.9065	11.0	10626	2.0348	1.0	248.7482	9.1454
0.727	12.0	11592	2.1641	1.0	249.0948	8.7838

Framework versions

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