9d349226367e9d18b78264d03d07015b

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

• Loss: 2.4124
• Data Size: 1.0
• Epoch Runtime: 241.3378
• Bleu: 10.2937

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	7.2589	0	20.3855	0.5802
No log	1	966	4.7902	0.0078	22.4682	3.0400
No log	2	1932	3.8371	0.0156	25.7186	3.4445
0.0954	3	2898	3.1302	0.0312	29.4700	4.4376
2.8464	4	3864	2.6920	0.0625	37.0910	5.0182
9.4064	5	4830	3.0577	0.125	51.2806	4.3941
3.1484	6	5796	2.4083	0.25	78.1883	11.8021
2.225	7	6762	2.1139	0.5	133.1155	15.6248
1.922	8.0	7728	1.9727	1.0	242.5346	12.9547
1.5996	9.0	8694	1.9480	1.0	242.2132	11.6672
1.3306	10.0	9660	1.9659	1.0	242.9703	9.5977
1.0594	11.0	10626	2.0788	1.0	243.4886	10.8914
0.8474	12.0	11592	2.2487	1.0	243.5826	9.3191
0.6643	13.0	12558	2.4124	1.0	241.3378	10.2937

Framework versions

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