102eae1cee1956d5c64359560e7c0853

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

• Loss: 2.8760
• Data Size: 1.0
• Epoch Runtime: 27.1372
• Bleu: 10.5500

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	7.2052	0	2.4134	0.4940
No log	1	88	5.5232	0.0078	3.2676	1.4234
No log	2	176	4.8515	0.0156	4.5845	2.3376
No log	3	264	4.6335	0.0312	6.7771	2.8991
No log	4	352	4.3934	0.0625	8.3214	3.3140
No log	5	440	3.9892	0.125	10.2469	4.3544
0.3222	6	528	2.7994	0.25	12.6403	5.3824
0.9133	7	616	2.1150	0.5	16.3635	5.7457
1.7982	8.0	704	2.0925	1.0	28.1853	8.8922
1.2718	9.0	792	2.2151	1.0	28.7831	11.1707
0.8519	10.0	880	2.3905	1.0	25.1130	9.0682
0.5648	11.0	968	2.6723	1.0	26.4069	9.9033
0.3645	12.0	1056	2.8760	1.0	27.1372	10.5500

Framework versions

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