3462875daa196a20fa9bf2da2177d375

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

• Loss: 4.3450
• Data Size: 1.0
• Epoch Runtime: 22.6952
• Bleu: 4.5011

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	7.8738	0	2.2483	0.1683
No log	1	75	7.2352	0.0078	2.6356	0.2798
No log	2	150	6.6912	0.0156	4.4592	0.3196
No log	3	225	6.2317	0.0312	6.2281	0.3211
No log	4	300	5.7579	0.0625	7.6936	0.5354
No log	5	375	5.1046	0.125	10.1824	0.7608
No log	6	450	4.6366	0.25	11.5660	1.4243
No log	7	525	4.2187	0.5	14.9105	2.1300
3.5524	8.0	600	3.8210	1.0	24.7847	2.9911
2.9283	9.0	675	3.7210	1.0	24.3817	3.6355
2.3344	10.0	750	3.7195	1.0	22.5836	4.1201
1.8345	11.0	825	3.8664	1.0	23.2775	3.6752
1.4425	12.0	900	4.0096	1.0	23.4431	4.3488
1.1082	13.0	975	4.2418	1.0	22.5425	4.6893
0.8585	14.0	1050	4.3450	1.0	22.6952	4.5011

Framework versions

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