b3a91d559d3c26f4bd03fc21a2908c88

This model is a fine-tuned version of facebook/mbart-large-cc25 on the Helsinki-NLP/opus_books [en-pt] dataset. It achieves the following results on the evaluation set:

• Loss: 6.2673
• Data Size: 1.0
• Epoch Runtime: 14.2114
• Bleu: 5.9125

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	13.5250	0	1.7328	0.5234
No log	1	35	9.1423	0.0078	2.0933	1.0173
No log	2	70	7.9082	0.0156	3.4280	0.6023
No log	3	105	8.2509	0.0312	4.5618	1.2211
No log	4	140	6.7944	0.0625	5.5468	1.9113
No log	5	175	5.9146	0.125	6.8835	2.1030
No log	6	210	5.1021	0.25	8.6150	3.4554
No log	7	245	4.3407	0.5	11.1800	5.0762
1.024	8.0	280	3.0975	1.0	15.0432	8.6316
3.8213	9.0	315	18.1036	1.0	14.2664	0.0048
15.4101	10.0	350	7.9817	1.0	14.8119	0.1582
15.4101	11.0	385	5.9988	1.0	16.4920	0.3163
6.4508	12.0	420	6.2673	1.0	14.2114	5.9125

Framework versions

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