68abbedd67b3ad9e6c547aaedf59fefe

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

• Loss: 4.1349
• Data Size: 1.0
• Epoch Runtime: 24.7812
• Bleu: 3.9231

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	9.4228	0	2.4525	0.1556
No log	1	86	8.4791	0.0078	2.9417	0.1234
No log	2	172	7.5451	0.0156	3.7800	0.3593
No log	3	258	6.9953	0.0312	4.9257	0.5513
No log	4	344	6.3182	0.0625	6.4235	0.7356
0.3564	5	430	5.6638	0.125	8.7085	1.1451
1.3142	6	516	4.8322	0.25	10.7702	1.7293
1.576	7	602	4.2068	0.5	14.9017	2.3929
2.1659	8.0	688	3.7897	1.0	26.4596	3.3869
3.0004	9.0	774	3.6596	1.0	26.4515	3.7408
2.4808	10.0	860	3.6951	1.0	25.0917	3.6660
2.0624	11.0	946	3.8219	1.0	25.2787	4.4714
1.532	12.0	1032	3.9705	1.0	26.3208	4.5332
1.2515	13.0	1118	4.1349	1.0	24.7812	3.9231

Framework versions

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