c11299132795dddc922e3afc1c391c93

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

• Loss: 2.4782
• Data Size: 1.0
• Epoch Runtime: 14.1248
• Bleu: 30.5773

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	6.7226	0	1.3572	0.8420
No log	1	35	6.0182	0.0078	1.7719	1.1671
No log	2	70	5.6072	0.0156	3.3959	1.3022
No log	3	105	5.0054	0.0312	4.6290	2.2814
No log	4	140	4.5388	0.0625	5.5968	2.9079
No log	5	175	4.0266	0.125	7.3799	3.7518
No log	6	210	3.4877	0.25	9.9524	5.0411
No log	7	245	2.8832	0.5	10.7647	7.5047
0.6501	8.0	280	2.4228	1.0	13.8585	11.3245
2.019	9.0	315	2.2550	1.0	13.4006	14.1053
1.2677	10.0	350	2.2912	1.0	13.3583	27.0282
1.2677	11.0	385	2.3061	1.0	14.4466	28.5896
0.7065	12.0	420	2.3487	1.0	13.4923	31.4395
0.4309	13.0	455	2.4782	1.0	14.1248	30.5773

Framework versions

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