7625b181d2e92ac5b1fc024c03df47c9

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

• Loss: 2.7958
• Data Size: 1.0
• Epoch Runtime: 5.5030
• Bleu: 1.2188

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	7.9482	0	1.0405	0.0175
No log	1	74	7.0727	0.0078	1.2689	0.0379
No log	2	148	6.4936	0.0156	1.3069	0.0400
0.2321	3	222	6.0663	0.0312	1.4860	0.0441
0.2321	4	296	5.5789	0.0625	1.5789	0.0654
0.3904	5	370	4.9782	0.125	1.9256	0.0736
0.3904	6	444	4.3889	0.25	2.4105	0.0475
0.9876	7	518	3.9695	0.5	3.5018	0.0651
2.5677	8.0	592	3.5792	1.0	5.7064	0.2648
3.525	9.0	666	3.3921	1.0	5.3753	0.4170
3.3565	10.0	740	3.2510	1.0	5.5686	0.5257
3.1458	11.0	814	3.1299	1.0	5.6335	0.6348
3.0346	12.0	888	3.0564	1.0	5.5839	0.7650
2.8539	13.0	962	2.9883	1.0	6.1253	0.7988
2.7784	14.0	1036	2.9352	1.0	5.4016	0.9171
2.6314	15.0	1110	2.9048	1.0	5.6046	0.9230
2.5463	16.0	1184	2.8656	1.0	5.3371	0.9870
2.4282	17.0	1258	2.8403	1.0	5.4282	0.9432
2.3614	18.0	1332	2.8179	1.0	5.5174	1.0111
2.2422	19.0	1406	2.8004	1.0	5.9597	1.1130
2.1763	20.0	1480	2.7917	1.0	5.3992	1.0947
2.075	21.0	1554	2.7932	1.0	5.9755	1.1365
1.9988	22.0	1628	2.8020	1.0	5.8151	1.1447
1.9175	23.0	1702	2.7943	1.0	5.5309	1.2010
1.8459	24.0	1776	2.7958	1.0	5.5030	1.2188

Framework versions

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