981afccec01bf88a81fb00932c25daf9

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

• Loss: 4.4884
• Data Size: 1.0
• Epoch Runtime: 22.2813
• Bleu: 1.9726

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	7.4919	0	2.0229	0.2508
No log	1	70	7.2982	0.0078	2.7890	0.3621
No log	2	140	6.3698	0.0156	4.4398	0.2253
No log	3	210	5.9575	0.0312	5.5643	0.2653
No log	4	280	5.3716	0.0625	7.5713	0.4389
No log	5	350	4.9457	0.125	9.8260	0.6113
No log	6	420	4.5837	0.25	12.0081	0.7620
0.7385	7	490	4.2412	0.5	14.6100	1.2364
3.6947	8.0	560	3.8733	1.0	23.6570	1.4282
3.0876	9.0	630	3.8536	1.0	22.8790	1.4831
2.3822	10.0	700	3.9193	1.0	21.6953	1.7791
1.8726	11.0	770	4.0480	1.0	21.3844	1.9155
1.5881	12.0	840	4.2881	1.0	21.8703	1.9862
1.0711	13.0	910	4.4884	1.0	22.2813	1.9726

Framework versions

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