e0c969c2033434e1b2e6b5df430d6685

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

• Loss: 1.9542
• Data Size: 1.0
• Epoch Runtime: 115.2325
• Bleu: 9.4790

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	9.7331	0	10.1980	0.2355
No log	1	437	5.8949	0.0078	11.3470	0.5734
No log	2	874	3.4754	0.0156	12.4938	2.9537
No log	3	1311	2.6557	0.0312	15.4939	4.6738
No log	4	1748	2.2887	0.0625	19.3210	6.2287
2.272	5	2185	2.1031	0.125	26.3551	7.1351
2.0705	6	2622	1.9476	0.25	39.2982	7.5199
1.8473	7	3059	1.8188	0.5	65.0606	8.1086
1.5448	8.0	3496	1.7166	1.0	114.5274	8.7691
1.2679	9.0	3933	1.7472	1.0	115.5967	9.3830
0.9914	10.0	4370	1.7783	1.0	116.5954	9.2402
0.7985	11.0	4807	1.8587	1.0	116.0882	9.3621
0.6148	12.0	5244	1.9542	1.0	115.2325	9.4790

Framework versions

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