62f78687a63bd0163ea6ae2a88f5f9ea

This model is a fine-tuned version of facebook/mbart-large-en-ro on the Helsinki-NLP/opus_books [de-pt] dataset. It achieves the following results on the evaluation set:

• Loss: 3.7935
• Data Size: 1.0
• Epoch Runtime: 13.2847
• Bleu: 6.1648

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	9.3029	0	1.5384	0.3753
No log	1	27	7.9261	0.0078	1.8883	0.4420
No log	2	54	7.7014	0.0156	2.3801	0.3426
No log	3	81	7.1754	0.0312	3.2272	0.4821
No log	4	108	6.5874	0.0625	5.1551	0.9791
No log	5	135	5.8621	0.125	6.8709	1.4381
No log	6	162	5.2101	0.25	9.1331	1.8087
No log	7	189	4.4697	0.5	10.2200	2.6981
1.0836	8.0	216	3.7250	1.0	12.8522	4.3788
1.0836	9.0	243	3.4973	1.0	12.4312	5.5047
3.1573	10.0	270	3.4011	1.0	12.5773	6.2735
3.1573	11.0	297	3.4240	1.0	12.8577	5.6236
2.096	12.0	324	3.5377	1.0	13.3776	6.0582
1.3877	13.0	351	3.6592	1.0	12.2628	6.1877
1.3877	14.0	378	3.7935	1.0	13.2847	6.1648

Framework versions

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