a777286ea74c92dbf51aedaa37cf5b54

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

• Loss: 2.0343
• Data Size: 1.0
• Epoch Runtime: 110.8831
• Bleu: 13.0225

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	4.9723	0	9.4915	5.2750
No log	1	437	2.7044	0.0078	11.5387	8.7819
No log	2	874	2.4462	0.0156	11.9691	7.8309
No log	3	1311	1.9974	0.0312	14.9892	9.9417
No log	4	1748	1.8580	0.0625	18.3916	12.6632
1.764	5	2185	1.7785	0.125	25.0846	14.3252
1.634	6	2622	1.6906	0.25	38.6214	13.4514
1.4863	7	3059	1.6098	0.5	62.3742	12.1987
1.2354	8.0	3496	1.5492	1.0	111.5669	15.3810
0.9381	9.0	3933	1.6111	1.0	111.9381	14.1359
0.6586	10.0	4370	1.7486	1.0	111.4121	16.5661
0.4636	11.0	4807	1.8970	1.0	111.3586	12.8779
0.3087	12.0	5244	2.0343	1.0	110.8831	13.0225

Framework versions

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