5a1300ad0688fcdf90b7e7134a1f0440

This model is a fine-tuned version of facebook/mbart-large-cc25 on the Helsinki-NLP/opus_books [de-fr] dataset. It achieves the following results on the evaluation set:

• Loss: 2.0775
• Data Size: 1.0
• Epoch Runtime: 228.4200
• Bleu: 8.4523

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	9.5185	0	19.7695	0.1139
No log	1	872	3.2724	0.0078	21.3004	3.5106
No log	2	1744	2.9019	0.0156	24.3693	4.3658
0.0573	3	2616	2.5503	0.0312	28.0127	5.8076
0.1716	4	3488	2.2643	0.0625	34.0810	6.1550
2.3441	5	4360	2.1034	0.125	47.2261	6.1790
2.1001	6	5232	1.9688	0.25	73.7344	6.6781
2.4045	7	6104	2.6920	0.5	126.0993	5.8348
1.7321	8.0	6976	1.7516	1.0	228.2525	7.9493
1.4357	9.0	7848	1.7042	1.0	227.9382	8.5248
1.2004	10.0	8720	1.7381	1.0	228.3942	8.6653
1.0102	11.0	9592	1.8187	1.0	228.2757	8.7123
0.801	12.0	10464	1.9144	1.0	227.8509	8.7149
0.6708	13.0	11336	2.0775	1.0	228.4200	8.4523

Framework versions

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