b8b3bff97fb1b6fa99fe84a726a23772

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

• Loss: 2.1345
• Data Size: 1.0
• Epoch Runtime: 111.9319
• Bleu: 10.1109

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	6.5234	0	9.5443	1.5698
No log	1	437	5.0694	0.0078	11.6079	4.7042
No log	2	874	4.7374	0.0156	11.9003	6.4231
No log	3	1311	3.2744	0.0312	14.1680	7.6423
No log	4	1748	2.5005	0.0625	17.8603	8.2596
4.0722	5	2185	21.4061	0.125	24.1977	0.0
7.6076	6	2622	5.8965	0.25	36.6845	1.3938
1.9355	7	3059	2.1748	0.5	61.4429	12.0880
1.4622	8.0	3496	1.6261	1.0	113.3700	13.5365
1.063	9.0	3933	1.6721	1.0	111.3725	13.2808
0.7611	10.0	4370	1.8179	1.0	111.2597	10.7188
0.5484	11.0	4807	1.9683	1.0	111.4251	11.3439
0.3714	12.0	5244	2.1345	1.0	111.9319	10.1109

Framework versions

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