64f6de9768b3c35984ae89792f287c42

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

• Loss: 2.4557
• Data Size: 1.0
• Epoch Runtime: 99.6508
• Bleu: 14.9950

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	6.5416	0	8.8485	0.7712
No log	1	390	3.2605	0.0078	10.1249	4.0407
No log	2	780	2.9410	0.0156	10.9217	5.3297
No log	3	1170	2.7271	0.0312	13.6924	5.9724
No log	4	1560	2.5655	0.0625	16.6201	6.7002
0.1517	5	1950	2.4337	0.125	22.9120	7.3128
0.3444	6	2340	2.2782	0.25	33.5483	9.3016
2.0661	7	2730	2.1516	0.5	54.9058	15.9019
1.7888	8.0	3120	2.0361	1.0	100.2726	10.1596
1.3944	9.0	3510	2.0683	1.0	99.2785	9.7060
1.0914	10.0	3900	2.1761	1.0	98.5484	8.7809
0.8515	11.0	4290	2.3118	1.0	98.7113	9.1502
0.6386	12.0	4680	2.4557	1.0	99.6508	14.9950

Framework versions

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