85e7f5e9fd9f27aaa06ad576c111ec0e

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

• Loss: 3.4320
• Data Size: 1.0
• Epoch Runtime: 21.8610
• Bleu: 2.7968

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	8.8652	0	2.1078	0.3527
No log	1	70	5.6598	0.0078	2.6090	0.4955
No log	2	140	4.2025	0.0156	4.3660	1.3683
No log	3	210	3.4679	0.0312	6.4246	1.5424
No log	4	280	3.3166	0.0625	8.0515	2.3667
No log	5	350	3.1599	0.125	9.2250	2.5596
No log	6	420	3.0531	0.25	10.5494	3.4340
0.4761	7	490	2.9127	0.5	13.1420	2.3470
2.4158	8.0	560	2.8439	1.0	22.0134	2.5924
1.9131	9.0	630	2.9053	1.0	22.5572	2.5951
1.3774	10.0	700	3.0882	1.0	20.3426	2.8000
0.9683	11.0	770	3.2552	1.0	21.1639	2.5319
0.8228	12.0	840	3.4320	1.0	21.8610	2.7968

Framework versions

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