826b89bd79e85f68edb38d7c003352e1

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

• Loss: 2.5765
• Data Size: 1.0
• Epoch Runtime: 173.3432
• Bleu: 6.5887

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	6.8678	0	14.8482	0.4622
No log	1	684	5.2122	0.0078	16.5556	0.7877
No log	2	1368	4.3973	0.0156	19.0116	1.3633
No log	3	2052	3.8859	0.0312	23.3773	2.0132
No log	4	2736	3.4578	0.0625	27.7830	2.8145
3.322	5	3420	3.0683	0.125	37.5707	3.6026
2.8854	6	4104	2.7191	0.25	57.7082	4.5074
2.4079	7	4788	2.4244	0.5	97.8405	5.7016
2.0508	8.0	5472	2.1817	1.0	177.5905	6.3958
1.6885	9.0	6156	2.1479	1.0	173.4006	7.7031
1.3677	10.0	6840	2.1994	1.0	174.1542	6.7265
1.1209	11.0	7524	2.2976	1.0	175.6237	7.2119
0.8887	12.0	8208	2.4181	1.0	173.4991	6.9422
0.6935	13.0	8892	2.5765	1.0	173.3432	6.5887

Framework versions

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