a98ac3b3e7809d2139a3d2c0521d59c1

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

• Loss: 3.1980
• Data Size: 1.0
• Epoch Runtime: 25.2013
• Bleu: 3.4196

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	7.9907	0	2.3499	0.1770
No log	1	83	5.1980	0.0078	2.9879	0.9366
No log	2	166	3.8521	0.0156	3.6156	2.0034
No log	3	249	3.2828	0.0312	4.6977	1.8125
0.1298	4	332	3.1333	0.0625	6.3454	2.7626
0.1298	5	415	2.9971	0.125	8.2904	2.7503
0.1298	6	498	2.8842	0.25	11.1224	2.6912
0.4594	7	581	2.7587	0.5	15.5039	3.3800
2.1692	8.0	664	2.6554	1.0	27.2888	3.5051
1.743	9.0	747	2.7011	1.0	26.4307	3.6196
1.1263	10.0	830	2.8705	1.0	23.9929	3.4852
0.7659	11.0	913	3.1093	1.0	24.4701	3.5095
0.5523	12.0	996	3.1980	1.0	25.2013	3.4196

Framework versions

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