ddf53001e3038deb16ba0a0ca969b8e8

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

• Loss: 3.6326
• Data Size: 1.0
• Epoch Runtime: 21.1460
• Bleu: 4.2166

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	8.8398	0	2.0643	0.2033
No log	1	70	6.5074	0.0078	2.4343	0.2531
No log	2	140	4.5825	0.0156	4.2013	0.5182
No log	3	210	4.0555	0.0312	6.4094	0.5958
No log	4	280	3.6039	0.0625	7.7091	1.0033
No log	5	350	3.3705	0.125	9.1733	1.3584
No log	6	420	3.1925	0.25	10.8567	1.6576
0.5158	7	490	3.0430	0.5	13.2820	2.4639
2.6235	8.0	560	2.9413	1.0	22.1017	2.7695
2.1746	9.0	630	2.9691	1.0	22.7817	4.7339
1.6487	10.0	700	3.1687	1.0	20.7175	4.7652
1.2296	11.0	770	3.3322	1.0	20.6932	3.9598
1.0752	12.0	840	3.6326	1.0	21.1460	4.2166

Framework versions

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