ecad43bcce3ca820d76ba3d35a0e5fae

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

• Loss: 3.4675
• Data Size: 1.0
• Epoch Runtime: 22.4097
• Bleu: 4.2961

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	6.0718	0	2.0394	0.7229
No log	1	70	4.8042	0.0078	2.5849	2.8342
No log	2	140	4.3469	0.0156	4.2470	2.1394
No log	3	210	3.9240	0.0312	5.5450	2.1730
No log	4	280	3.5928	0.0625	6.5012	2.6214
No log	5	350	3.2208	0.125	9.1357	2.7766
No log	6	420	2.8480	0.25	11.0888	2.9352
0.4604	7	490	2.7168	0.5	13.4212	3.3076
2.2123	8.0	560	2.6395	1.0	22.2576	5.6053
1.7942	9.0	630	2.7723	1.0	21.8267	5.6059
1.2663	10.0	700	2.9940	1.0	21.0697	4.7295
0.9044	11.0	770	3.2086	1.0	22.1088	4.7528
0.7006	12.0	840	3.4675	1.0	22.4097	4.2961

Framework versions

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