80f52a103d0825e12f81c511acfb327c

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

• Loss: 1.7450
• Data Size: 1.0
• Epoch Runtime: 115.6653
• Bleu: 21.8918

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	8.5730	0	9.8693	0.3960
No log	1	447	3.0760	0.0078	10.9452	3.0275
0.0666	2	894	2.5225	0.0156	12.5995	4.9702
0.0583	3	1341	2.2336	0.0312	14.8824	6.4494
0.0866	4	1788	2.0329	0.0625	17.7833	7.9385
0.1456	5	2235	1.8403	0.125	24.9107	11.8712
1.7273	6	2682	1.6731	0.25	37.6486	22.4934
1.4629	7	3129	1.5201	0.5	64.9585	29.4014
1.1563	8.0	3576	1.4080	1.0	114.8667	30.7330
0.8858	9.0	4023	1.4160	1.0	114.6078	29.4591
0.617	10.0	4470	1.5177	1.0	115.9977	21.3742
0.4238	11.0	4917	1.6337	1.0	113.5079	23.9158
0.2923	12.0	5364	1.7450	1.0	115.6653	21.8918

Framework versions

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