dafa4e8eb1d17852f330c14c5e2a3b7a

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

• Loss: 3.0946
• Data Size: 1.0
• Epoch Runtime: 27.9420
• Bleu: 11.9634

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	5.4802	0	2.4402	3.2170
No log	1	91	3.7203	0.0078	3.0760	6.6129
No log	2	182	3.3913	0.0156	4.6211	5.0636
No log	3	273	3.2828	0.0312	5.8106	4.3378
No log	4	364	3.1267	0.0625	7.3266	4.7021
No log	5	455	2.7582	0.125	8.8637	6.9631
No log	6	546	2.4730	0.25	11.0304	6.3279
0.2683	7	637	2.3649	0.5	16.8135	6.6146
1.8331	8.0	728	2.2818	1.0	29.5321	11.8647
1.2922	9.0	819	2.3430	1.0	29.2260	13.3361
0.8527	10.0	910	2.5722	1.0	26.4292	13.8038
0.5482	11.0	1001	2.8424	1.0	26.6330	11.8096
0.3198	12.0	1092	3.0946	1.0	27.9420	11.9634

Framework versions

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