8d085b944d646ddeb456356a8cfd58b0

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

• Loss: 1.7779
• Data Size: 1.0
• Epoch Runtime: 53.3515
• Bleu: 4.6333

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	7.3253	0	4.7739	0.0681
No log	1	872	6.0345	0.0078	5.8041	0.0824
No log	2	1744	5.2928	0.0156	5.6406	0.1203
0.0905	3	2616	4.6313	0.0312	6.4192	0.1931
0.2952	4	3488	4.0727	0.0625	7.8977	0.4409
4.0354	5	4360	3.6077	0.125	10.8061	0.6865
3.4132	6	5232	3.1622	0.25	17.5762	1.1457
2.8767	7	6104	2.7416	0.5	29.2806	1.6228
2.4935	8.0	6976	2.3589	1.0	52.5426	2.2761
2.2324	9.0	7848	2.1594	1.0	52.8364	2.8621
2.0792	10.0	8720	2.0348	1.0	51.7051	3.1707
1.9309	11.0	9592	1.9477	1.0	52.2441	3.4644
1.7878	12.0	10464	1.8891	1.0	52.0549	3.7366
1.7101	13.0	11336	1.8439	1.0	51.9242	3.9138
1.607	14.0	12208	1.8182	1.0	51.6807	4.0923
1.5483	15.0	13080	1.7968	1.0	53.2312	4.1112
1.4871	16.0	13952	1.7713	1.0	53.5598	4.2782
1.394	17.0	14824	1.7606	1.0	54.2662	4.3508
1.3554	18.0	15696	1.7514	1.0	54.5283	4.3676
1.3169	19.0	16568	1.7524	1.0	53.5586	4.4339
1.2567	20.0	17440	1.7487	1.0	53.5895	4.4818
1.2419	21.0	18312	1.7565	1.0	53.2275	4.5257
1.1708	22.0	19184	1.7601	1.0	53.0871	4.5180
1.1108	23.0	20056	1.7625	1.0	54.0304	4.6402
1.067	24.0	20928	1.7779	1.0	53.3515	4.6333

Framework versions

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