2c20dc03484c9c7a25fbf299c79e7a20

This model is a fine-tuned version of Helsinki-NLP/opus-mt-tc-bible-big-deu_eng_fra_por_spa-mul on the Helsinki-NLP/opus_books [de-it] dataset. It achieves the following results on the evaluation set:

• Loss: 1.8916
• Data Size: 1.0
• Epoch Runtime: 45.0080
• Bleu: 4.3293

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	7.6228	0	3.8018	0.0412
No log	1	684	4.2279	0.0078	5.4240	0.9904
No log	2	1368	3.4820	0.0156	5.0020	1.3593
No log	3	2052	3.0334	0.0312	6.1478	1.9079
No log	4	2736	2.6682	0.0625	7.4365	2.3377
2.6341	5	3420	2.3814	0.125	9.9560	2.8127
2.3054	6	4104	2.1689	0.25	15.0634	3.2751
2.025	7	4788	1.9974	0.5	26.8253	3.6679
1.8067	8.0	5472	1.8389	1.0	46.5713	4.0737
1.5878	9.0	6156	1.7896	1.0	45.5217	4.2415
1.4214	10.0	6840	1.7723	1.0	45.4363	4.3850
1.2577	11.0	7524	1.7913	1.0	46.1656	4.3379
1.1413	12.0	8208	1.7902	1.0	46.5078	4.4271
1.0144	13.0	8892	1.8461	1.0	45.8650	4.3907
0.9216	14.0	9576	1.8916	1.0	45.0080	4.3293

Framework versions

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