601532e09c696271fb088bd22914dd73

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

• Loss: 2.8308
• Data Size: 1.0
• Epoch Runtime: 97.6435
• Bleu: 8.5806

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	6.6868	0	8.5020	0.6128
No log	1	390	5.5542	0.0078	9.8223	0.8411
No log	2	780	4.6723	0.0156	10.9178	1.2933
No log	3	1170	4.1179	0.0312	13.5088	2.5588
No log	4	1560	3.6680	0.0625	16.9016	3.5287
0.2223	5	1950	3.3094	0.125	22.2548	4.6092
0.4713	6	2340	2.9458	0.25	32.7358	5.6265
2.6286	7	2730	2.6163	0.5	54.9287	6.5900
2.1805	8.0	3120	2.3596	1.0	99.4294	8.1377
1.7441	9.0	3510	2.2978	1.0	98.3983	8.3482
1.4148	10.0	3900	2.3557	1.0	98.5713	8.3169
1.1456	11.0	4290	2.4745	1.0	98.5585	9.5080
0.8757	12.0	4680	2.6257	1.0	100.1076	8.5862
0.6826	13.0	5070	2.8308	1.0	97.6435	8.5806

Framework versions

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