13b6de449711a5d470d915b0847d3abe

This model is a fine-tuned version of facebook/mbart-large-cc25 on the Helsinki-NLP/opus_books [en-ru] dataset. It achieves the following results on the evaluation set:

• Loss: 2.0848
• Data Size: 1.0
• Epoch Runtime: 114.1406
• Bleu: 10.1693

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	11.1933	0	10.3223	0.2658
No log	1	437	4.1998	0.0078	11.6177	3.4195
No log	2	874	3.6857	0.0156	12.5071	5.3675
No log	3	1311	3.2341	0.0312	15.5816	5.1485
No log	4	1748	2.5896	0.0625	20.2271	5.7053
2.4569	5	2185	2.1577	0.125	27.3300	6.3891
2.1198	6	2622	8.7999	0.25	39.8441	0.0451
1.9349	7	3059	1.8195	0.5	64.4659	12.7740
1.5914	8.0	3496	1.6958	1.0	114.7476	11.1364
1.2797	9.0	3933	1.7328	1.0	114.1484	12.3817
0.9956	10.0	4370	1.8072	1.0	115.1733	11.2940
0.7789	11.0	4807	1.9274	1.0	117.0363	11.5555
0.5714	12.0	5244	2.0848	1.0	114.1406	10.1693

Framework versions

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