449155f2bd6fa3fa3e8c0cd7862c37ba

This model is a fine-tuned version of facebook/mbart-large-cc25 on the Helsinki-NLP/opus_books [en-fi] dataset. It achieves the following results on the evaluation set:

• Loss: 3.0394
• Data Size: 1.0
• Epoch Runtime: 28.0594
• Bleu: 4.5346

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	8.8285	0	2.8129	0.1345
No log	1	91	5.6006	0.0078	3.2968	1.3733
No log	2	182	4.1664	0.0156	4.5960	2.0451
No log	3	273	3.9010	0.0312	5.7522	3.6416
No log	4	364	3.6790	0.0625	8.1714	5.0547
No log	5	455	3.3073	0.125	9.6245	7.5336
No log	6	546	2.9522	0.25	11.6486	7.1338
0.3487	7	637	2.6509	0.5	16.9396	5.5421
2.1967	8.0	728	2.4553	1.0	28.8635	4.2146
1.6412	9.0	819	2.5054	1.0	28.0762	4.9165
1.2418	10.0	910	2.6190	1.0	26.8029	4.5841
0.9185	11.0	1001	2.8003	1.0	27.5325	4.6813
0.6414	12.0	1092	3.0394	1.0	28.0594	4.5346

Framework versions

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