e587887b2e830a371f5849839cacad37

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

• Loss: 2.8832
• Data Size: 1.0
• Epoch Runtime: 27.6088
• Bleu: 5.5435

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	5.4409	0	2.4780	1.0501
No log	1	91	3.5981	0.0078	3.1606	2.9666
No log	2	182	3.0654	0.0156	4.4946	2.9876
No log	3	273	2.8838	0.0312	5.6420	3.2427
No log	4	364	2.7455	0.0625	7.2975	3.5273
No log	5	455	2.6249	0.125	8.8096	4.1640
No log	6	546	2.5067	0.25	11.0613	4.5614
0.253	7	637	2.4110	0.5	16.8851	4.6738
1.8308	8.0	728	2.3394	1.0	30.5210	5.0469
1.2947	9.0	819	2.3865	1.0	29.4193	5.6452
0.8467	10.0	910	2.5894	1.0	26.7900	5.8255
0.5515	11.0	1001	2.7316	1.0	26.8534	5.3681
0.3156	12.0	1092	2.8832	1.0	27.6088	5.5435

Framework versions

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