094fd7098eac2b3a9905407d50c759a6

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

• Loss: 4.4439
• Data Size: 1.0
• Epoch Runtime: 22.5420
• Bleu: 2.0387

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	7.5944	0	2.0800	0.3199
No log	1	70	6.8697	0.0078	2.5215	0.3340
No log	2	140	6.3610	0.0156	4.0927	0.3510
No log	3	210	5.9572	0.0312	6.1886	0.4968
No log	4	280	5.3846	0.0625	7.4989	0.6250
No log	5	350	5.0034	0.125	9.4143	0.6358
No log	6	420	4.6157	0.25	10.9682	0.8674
0.7395	7	490	4.2437	0.5	13.5399	1.1599
3.7574	8.0	560	3.9353	1.0	22.5426	1.3493
3.1266	9.0	630	3.8738	1.0	22.3463	1.7505
2.4328	10.0	700	3.9355	1.0	21.1418	1.7014
1.8807	11.0	770	4.0712	1.0	22.5253	1.7181
1.6569	12.0	840	4.2781	1.0	22.8453	1.7616
1.1225	13.0	910	4.4439	1.0	22.5420	2.0387

Framework versions

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