b969bc867b90f56bc4433ff11dab7ae5

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on the Helsinki-NLP/opus_books [fr-no] dataset. It achieves the following results on the evaluation set:

• Loss: 3.8554
• Data Size: 1.0
• Epoch Runtime: 25.8744
• Bleu: 4.9002

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	9.5500	0	2.4574	0.1247
No log	1	86	8.0908	0.0078	3.0788	0.1868
No log	2	172	7.0738	0.0156	3.5624	0.4953
No log	3	258	6.5109	0.0312	4.8388	0.7828
No log	4	344	5.8899	0.0625	6.3969	1.0224
0.3345	5	430	5.1821	0.125	8.4454	1.6747
1.1946	6	516	4.4546	0.25	11.0254	2.0790
1.4495	7	602	3.9642	0.5	15.2643	2.7731
1.9859	8.0	688	3.5619	1.0	26.7082	3.7382
2.6583	9.0	774	3.4624	1.0	27.3083	4.0757
2.1107	10.0	860	3.4976	1.0	25.8223	4.3825
1.6913	11.0	946	3.5780	1.0	26.2844	4.6988
1.1866	12.0	1032	3.6965	1.0	26.6887	5.8384
0.9265	13.0	1118	3.8554	1.0	25.8744	4.9002

Framework versions

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