ae6228c485c3818b99944c041d029514

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

• Loss: 4.0600
• Data Size: 1.0
• Epoch Runtime: 23.9269
• Bleu: 2.4848

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	8.3242	0	2.3045	0.2318
No log	1	83	7.3548	0.0078	2.9914	0.2911
No log	2	166	6.6649	0.0156	3.6875	0.3709
No log	3	249	6.1721	0.0312	4.8621	0.3549
0.1945	4	332	5.5080	0.0625	6.1705	0.5063
0.1945	5	415	4.9975	0.125	8.2470	1.1509
0.1945	6	498	4.5544	0.25	10.4663	1.3295
0.79	7	581	4.1187	0.5	15.7499	1.7999
3.5345	8.0	664	3.7088	1.0	26.8734	2.2441
2.9393	9.0	747	3.6083	1.0	26.3814	2.2833
2.1116	10.0	830	3.6605	1.0	24.0157	2.3308
1.6044	11.0	913	3.7596	1.0	24.1972	2.4836
1.2482	12.0	996	3.8987	1.0	24.5207	2.3740
0.8067	13.0	1079	4.0600	1.0	23.9269	2.4848

Framework versions

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