647fe3da288ee2dba74e55d6843bf4d7

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

• Loss: 1.8024
• Data Size: 1.0
• Epoch Runtime: 117.2758
• Bleu: 10.9219

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	5.9666	0	9.8803	0.7704
No log	1	447	4.5187	0.0078	11.8300	1.7236
0.0793	2	894	3.4836	0.0156	12.8525	3.1679
0.0824	3	1341	2.9831	0.0312	14.8335	4.5718
0.1177	4	1788	2.6561	0.0625	18.0296	5.8346
0.1931	5	2235	2.3428	0.125	25.1397	6.9470
2.2172	6	2682	2.0652	0.25	38.0805	8.0879
1.7968	7	3129	1.8062	0.5	66.6981	9.4933
1.3694	8.0	3576	1.6164	1.0	118.5646	10.5113
1.07	9.0	4023	1.5784	1.0	117.8986	11.0174
0.7746	10.0	4470	1.6013	1.0	117.9308	10.8405
0.5413	11.0	4917	1.6573	1.0	118.1724	11.1652
0.3863	12.0	5364	1.7049	1.0	118.2122	11.2885
0.2563	13.0	5811	1.8024	1.0	117.2758	10.9219

Framework versions

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