bceb0337cb554f4c62214860f1373b26

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

• Loss: 3.1539
• Data Size: 1.0
• Epoch Runtime: 5.7632
• Bleu: 0.8213

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	7.6116	0	1.0686	0.0841
No log	1	83	6.9724	0.0078	1.7692	0.1276
No log	2	166	6.4382	0.0156	1.3506	0.1528
No log	3	249	6.0401	0.0312	1.5721	0.1538
0.1859	4	332	5.5715	0.0625	1.8596	0.1927
0.1859	5	415	5.0677	0.125	2.0680	0.2680
0.1859	6	498	4.5493	0.25	2.7410	0.3159
0.8075	7	581	4.0627	0.5	3.8577	0.4408
3.9199	8.0	664	3.6502	1.0	6.3081	0.7432
3.6771	9.0	747	3.4503	1.0	5.5533	0.7491
3.381	10.0	830	3.3299	1.0	5.5959	0.7642
3.1632	11.0	913	3.2270	1.0	5.6192	0.8688
3.0627	12.0	996	3.1704	1.0	5.7095	0.8179
2.8767	13.0	1079	3.1065	1.0	6.0457	0.9098
2.765	14.0	1162	3.0953	1.0	5.9379	0.8887
2.6574	15.0	1245	3.0498	1.0	5.7078	0.9819
2.5227	16.0	1328	3.0447	1.0	5.8356	0.8464
2.4066	17.0	1411	3.0448	1.0	5.6789	0.8952
2.3234	18.0	1494	3.0499	1.0	5.8090	0.7940
2.1864	19.0	1577	3.0465	1.0	5.5872	0.8076
2.0919	20.0	1660	3.0428	1.0	5.6298	0.8191
1.9912	21.0	1743	3.0776	1.0	5.7300	0.8730
1.8689	22.0	1826	3.0910	1.0	5.9598	0.8892
1.8023	23.0	1909	3.1180	1.0	5.7454	0.8601
1.7221	24.0	1992	3.1539	1.0	5.7632	0.8213

Framework versions

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