ba3d7920739f073c6e73ff735053bd96

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

• Loss: 2.0515
• Data Size: 1.0
• Epoch Runtime: 250.0475
• Bleu: 15.7698

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	8.1709	0	21.0008	0.4096
No log	1	1000	4.6603	0.0078	23.8425	3.4143
No log	2	2000	4.0562	0.0156	25.6754	4.8792
No log	3	3000	2.6733	0.0312	29.8638	7.8679
0.1127	4	4000	2.2054	0.0625	36.8461	9.8023
11.3037	5	5000	3.9375	0.125	52.6720	2.2594
0.1678	6	6000	2.9874	0.25	80.3009	11.1655
0.1589	7	7000	1.7727	0.5	137.9458	16.8958
1.4839	8.0	8000	1.6009	1.0	252.8147	17.9767
1.2031	9.0	9000	1.5979	1.0	252.2794	13.9304
1.1558	10.0	10000	1.7806	1.0	251.6586	14.7671
0.9149	11.0	11000	1.8045	1.0	254.2162	16.9725
0.642	12.0	12000	1.8849	1.0	250.0821	14.9369
0.4824	13.0	13000	2.0515	1.0	250.0475	15.7698

Framework versions

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