9a62f488670de53073ebb3ef20f05122

This model is a fine-tuned version of facebook/mbart-large-50-many-to-one-mmt on the Helsinki-NLP/opus_books [fi-no] dataset. It achieves the following results on the evaluation set:

• Loss: 4.2683
• Data Size: 1.0
• Epoch Runtime: 24.0280
• Bleu: 6.6626

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	8.5975	0	2.4251	0.3223
No log	1	85	7.8556	0.0078	2.7591	0.3776
No log	2	170	7.4122	0.0156	3.6302	0.3743
No log	3	255	6.8551	0.0312	4.7711	0.5837
No log	4	340	6.0538	0.0625	6.2807	0.6392
0.4097	5	425	5.5405	0.125	8.2002	1.1423
0.4097	6	510	4.9052	0.25	10.7034	1.5576
1.4869	7	595	4.3737	0.5	14.8912	3.1569
3.9144	8.0	680	3.9527	1.0	26.8421	2.9492
3.0946	9.0	765	3.7947	1.0	26.0129	5.6238
2.5772	10.0	850	3.8332	1.0	23.9638	6.3281
2.0857	11.0	935	3.9019	1.0	24.9219	7.2568
1.6424	12.0	1020	4.0640	1.0	25.3270	5.2450
1.3056	13.0	1105	4.2683	1.0	24.0280	6.6626

Framework versions

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