e4fdad87d127509bab083e4ff85358c5

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

• Loss: 1.7686
• Data Size: 1.0
• Epoch Runtime: 115.4577
• Bleu: 26.1737

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	6.8435	0	9.9026	1.1964
No log	1	447	5.0466	0.0078	12.0204	4.2724
0.0887	2	894	4.6620	0.0156	12.7230	5.6052
0.1049	3	1341	4.0533	0.0312	14.6503	6.8227
0.1571	4	1788	5.2419	0.0625	17.9297	0.8969
0.2144	5	2235	1.8335	0.125	24.6933	8.5775
1.743	6	2682	1.5809	0.25	37.5709	17.8514
1.4079	7	3129	1.4572	0.5	64.3133	22.2166
1.1106	8.0	3576	1.3524	1.0	114.7100	21.9254
1.1181	9.0	4023	2.8708	1.0	115.0033	8.7383
0.6107	10.0	4470	1.4992	1.0	115.2602	27.9674
0.4165	11.0	4917	1.6522	1.0	115.2928	24.0041
0.2971	12.0	5364	1.7686	1.0	115.4577	26.1737

Framework versions

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