a2827bceab8438d5f59eb2f86e5b137c

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

• Loss: 2.0187
• Data Size: 1.0
• Epoch Runtime: 3.3276
• Bleu: 11.3535

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	7.5557	0	0.8580	0.1951
No log	1	35	6.9431	0.0078	1.2215	0.3184
No log	2	70	6.6133	0.0156	1.1762	0.3861
No log	3	105	6.2009	0.0312	1.1370	0.3448
No log	4	140	5.7919	0.0625	1.3712	0.3719
No log	5	175	5.3244	0.125	1.4944	1.4114
No log	6	210	4.7689	0.25	1.7733	1.8717
No log	7	245	4.0924	0.5	2.3587	2.9009
0.8971	8.0	280	3.4097	1.0	3.6034	4.2664
3.5184	9.0	315	2.9503	1.0	3.5442	5.9878
2.763	10.0	350	2.6559	1.0	3.5353	6.6779
2.763	11.0	385	2.4561	1.0	3.2018	7.5954
2.1787	12.0	420	2.2892	1.0	3.1464	8.4306
1.7665	13.0	455	2.1786	1.0	3.2968	8.7251
1.7665	14.0	490	2.0858	1.0	3.5047	9.5318
1.4479	15.0	525	2.0386	1.0	3.4783	10.2377
1.2102	16.0	560	2.0110	1.0	3.6395	10.9023
1.2102	17.0	595	1.9811	1.0	3.7202	10.4676
0.9834	18.0	630	1.9679	1.0	3.7819	10.7259
0.8031	19.0	665	1.9682	1.0	3.2968	10.8775
0.6847	20.0	700	1.9678	1.0	3.2164	11.0137
0.6847	21.0	735	1.9655	1.0	3.0988	11.2297
0.5418	22.0	770	1.9757	1.0	3.0876	11.1092
0.4529	23.0	805	1.9874	1.0	3.1412	11.3017
0.4529	24.0	840	2.0007	1.0	3.3145	11.3775
0.3737	25.0	875	2.0187	1.0	3.3276	11.3535

Framework versions

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