b6ae3735fc47c678c4fe21ef37c8e6c4

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

• Loss: 3.2930
• Data Size: 1.0
• Epoch Runtime: 22.4829
• Bleu: 2.7016

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	5.0240	0	2.0894	1.1714
No log	1	70	4.2125	0.0078	2.5303	1.5901
No log	2	140	3.5518	0.0156	4.2352	1.6501
No log	3	210	3.3092	0.0312	5.6056	2.0324
No log	4	280	3.1599	0.0625	6.7141	2.1936
No log	5	350	3.0061	0.125	9.2068	2.3065
No log	6	420	2.8878	0.25	11.2329	2.6471
0.4372	7	490	2.7796	0.5	13.9990	2.7439
2.2382	8.0	560	2.6987	1.0	23.9129	2.9733
1.7943	9.0	630	2.8077	1.0	23.3258	3.2043
1.2539	10.0	700	2.9686	1.0	22.0996	2.9753
0.8718	11.0	770	3.1302	1.0	21.9650	2.8229
0.6912	12.0	840	3.2930	1.0	22.4829	2.7016

Framework versions

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