835dffafccf8eac80deb1faa6ec7181b

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

• Loss: 1.8818
• Data Size: 1.0
• Epoch Runtime: 113.4524
• Bleu: 10.0707

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	4.2182	0	9.6689	2.5533
No log	1	434	2.2424	0.0078	11.6986	5.9236
No log	2	868	2.1050	0.0156	12.8396	8.4874
No log	3	1302	1.9950	0.0312	14.7262	8.5035
No log	4	1736	1.9164	0.0625	17.7558	7.3701
0.0789	5	2170	1.8304	0.125	24.2278	8.4135
1.6796	6	2604	1.7328	0.25	36.8459	9.7528
1.5086	7	3038	1.6272	0.5	62.4130	10.1015
1.2723	8.0	3472	1.5584	1.0	113.2186	11.0803
0.9168	9.0	3906	1.5856	1.0	112.3829	12.5470
0.6475	10.0	4340	1.6735	1.0	112.7170	12.9360
0.4694	11.0	4774	1.7640	1.0	112.4137	11.5547
0.3135	12.0	5208	1.8818	1.0	113.4524	10.0707

Framework versions

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