0a99cc05f52de490ae6b6717f001fdd0

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on the Helsinki-NLP/opus_books [de-es] dataset. It achieves the following results on the evaluation set:

• Loss: 2.6594
• Data Size: 1.0
• Epoch Runtime: 174.3643
• Bleu: 5.6778

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	6.4553	0	14.8939	0.7732
No log	1	688	2.9888	0.0078	17.7059	3.3610
No log	2	1376	2.7713	0.0156	19.1442	3.9708
No log	3	2064	2.6373	0.0312	22.8170	4.2848
0.1006	4	2752	2.5330	0.0625	27.8764	4.5525
0.1988	5	3440	2.4375	0.125	38.1770	4.9978
2.3152	6	4128	2.3395	0.25	57.1517	5.3961
2.1185	7	4816	2.2405	0.5	97.1274	6.1037
1.9252	8.0	5504	2.1631	1.0	176.2259	6.4709
1.5818	9.0	6192	2.1849	1.0	176.0133	6.8717
1.2938	10.0	6880	2.2884	1.0	174.6970	6.9859
1.0192	11.0	7568	2.4454	1.0	174.3252	5.9712
0.7967	12.0	8256	2.6594	1.0	174.3643	5.6778

Framework versions

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