346672dd1e3db337fadf2f41c73126cd

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

• Loss: 2.0344
• Data Size: 1.0
• Epoch Runtime: 112.1518
• Bleu: 23.7390

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	6.0251	0	9.6077	0.4803
No log	1	434	3.0267	0.0078	10.6128	2.5907
No log	2	868	2.6114	0.0156	13.0966	4.2078
No log	3	1302	2.4027	0.0312	14.7499	4.9749
No log	4	1736	2.2267	0.0625	18.0876	5.8878
0.0962	5	2170	2.0601	0.125	24.4411	6.7591
1.9004	6	2604	1.9152	0.25	37.6751	8.0561
1.6738	7	3038	1.7625	0.5	62.2125	11.9116
1.4059	8.0	3472	1.6533	1.0	112.6699	21.6430
1.0486	9.0	3906	1.6618	1.0	111.2092	25.2220
0.77	10.0	4340	1.7649	1.0	112.7545	24.7275
0.5547	11.0	4774	1.8952	1.0	111.6849	22.5717
0.3818	12.0	5208	2.0344	1.0	112.1518	23.7390

Framework versions

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