19df135405a3b12b3c91c3b6de1181d8

This model is a fine-tuned version of Helsinki-NLP/opus-mt-en-ru on the Helsinki-NLP/opus_books [it-pt] dataset. It achieves the following results on the evaluation set:

• Loss: 2.9124
• Data Size: 1.0
• Epoch Runtime: 3.3139
• Bleu: 3.0603

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	7.8643	0	0.8366	0.0598
No log	1	29	7.0372	0.0078	1.0685	0.1592
No log	2	58	6.6368	0.0156	1.0917	0.1616
No log	3	87	6.4076	0.0312	1.1097	0.0946
No log	4	116	6.1124	0.0625	1.1707	0.0971
No log	5	145	5.6785	0.125	1.4770	0.2010
0.5756	6	174	5.1364	0.25	1.6681	0.2645
0.5756	7	203	4.5414	0.5	2.2428	0.8513
0.5756	8.0	232	4.0100	1.0	3.2256	1.2418
2.8626	9.0	261	3.7042	1.0	3.1127	1.6120
2.8626	10.0	290	3.5029	1.0	2.5964	1.9884
3.5225	11.0	319	3.3426	1.0	2.5789	2.2086
3.5225	12.0	348	3.2298	1.0	2.9337	2.3969
3.1276	13.0	377	3.1476	1.0	2.8596	2.5251
2.8205	14.0	406	3.1021	1.0	3.1079	2.4360
2.8205	15.0	435	3.0441	1.0	3.1088	2.5562
2.5682	16.0	464	2.9815	1.0	3.1119	2.7030
2.5682	17.0	493	2.9739	1.0	3.1831	2.6723
2.3601	18.0	522	2.9282	1.0	2.8826	2.7550
2.169	19.0	551	2.9331	1.0	2.9833	2.8079
2.169	20.0	580	2.9216	1.0	3.0446	2.7785
1.9848	21.0	609	2.9127	1.0	3.3704	2.8553
1.9848	22.0	638	2.9011	1.0	3.6254	2.9286
1.8372	23.0	667	2.9119	1.0	3.8259	2.9172
1.8372	24.0	696	2.9100	1.0	3.9413	2.9813
1.6768	25.0	725	2.9225	1.0	3.2220	2.9479
1.5527	26.0	754	2.9124	1.0	3.3139	3.0603

Framework versions

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