d965486b5941b29a183921c577a9bd6d

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

• Loss: 1.5445
• Data Size: 1.0
• Epoch Runtime: 58.8049
• Bleu: 6.5367

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	7.4445	0	5.1063	0.0938
No log	1	1000	5.8756	0.0078	7.5546	0.1068
No log	2	2000	4.9859	0.0156	6.2499	0.2114
No log	3	3000	4.2372	0.0312	7.0806	0.4379
0.1632	4	4000	3.7105	0.0625	8.7523	0.9057
3.5828	5	5000	3.2649	0.125	12.2145	1.3495
0.1889	6	6000	2.8327	0.25	19.3382	2.1308
0.2328	7	7000	2.4261	0.5	33.2163	2.8094
2.165	8.0	8000	2.0497	1.0	61.8125	3.8672
1.924	9.0	9000	1.8628	1.0	61.2745	4.6907
1.7466	10.0	10000	1.7502	1.0	61.5088	5.1325
1.6322	11.0	11000	1.6712	1.0	62.1993	5.5876
1.5494	12.0	12000	1.6251	1.0	63.2371	5.6716
1.4597	13.0	13000	1.5923	1.0	61.2425	5.8973
1.386	14.0	14000	1.5602	1.0	60.1241	5.9398
1.2968	15.0	15000	1.5393	1.0	62.2988	6.0595
1.2484	16.0	16000	1.5326	1.0	63.1195	6.1564
1.1665	17.0	17000	1.5292	1.0	62.7886	6.1951
1.1624	18.0	18000	1.5221	1.0	61.7314	6.2969
1.1009	19.0	19000	1.5185	1.0	60.2493	6.4131
1.0542	20.0	20000	1.5220	1.0	61.7241	6.3698
1.0236	21.0	21000	1.5237	1.0	61.2245	6.4337
0.9892	22.0	22000	1.5305	1.0	58.5334	6.5112
0.934	23.0	23000	1.5445	1.0	58.8049	6.5367

Framework versions

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