2c02ba7bd860f57460c06a7c69af0002

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

• Loss: 3.5088
• Data Size: 1.0
• Epoch Runtime: 23.5590
• Bleu: 6.0564

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	9.2428	0	2.3190	0.2601
No log	1	83	7.2409	0.0078	3.1635	0.7089
No log	2	166	6.0328	0.0156	3.6555	1.2021
No log	3	249	5.6793	0.0312	4.6396	1.7452
0.1899	4	332	5.3600	0.0625	6.2399	2.1872
0.1899	5	415	4.1511	0.125	8.0515	3.0644
0.1899	6	498	3.4630	0.25	10.3359	3.7444
0.6781	7	581	2.9374	0.5	14.6889	2.4826
3.0119	8.0	664	2.6274	1.0	25.8215	6.3077
2.0259	9.0	747	2.5787	1.0	25.8268	7.5728
1.3135	10.0	830	2.7438	1.0	23.7486	7.8539
0.896	11.0	913	3.0701	1.0	23.9186	6.3187
0.6429	12.0	996	3.3497	1.0	24.2604	6.4303
0.3468	13.0	1079	3.5088	1.0	23.5590	6.0564

Framework versions

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