22a512cc45321a262d4b8162e41c0f59

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

• Loss: 3.4122
• Data Size: 1.0
• Epoch Runtime: 22.3931
• Bleu: 2.0316

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	7.8814	0	2.0258	0.3561
No log	1	70	5.2185	0.0078	2.4130	0.6406
No log	2	140	3.9773	0.0156	4.2047	1.4379
No log	3	210	3.4240	0.0312	5.4487	1.6175
No log	4	280	3.2574	0.0625	6.9062	1.8300
No log	5	350	3.1161	0.125	8.2313	2.0987
No log	6	420	2.9757	0.25	10.7578	2.2297
0.4627	7	490	2.8897	0.5	13.6793	2.4172
2.3313	8.0	560	2.7959	1.0	22.5926	2.5223
1.8863	9.0	630	2.8873	1.0	23.8936	1.8577
1.3473	10.0	700	3.0749	1.0	23.6989	1.9127
0.9662	11.0	770	3.2179	1.0	21.6525	1.4574
0.7848	12.0	840	3.4122	1.0	22.3931	2.0316

Framework versions

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