9d7adad6f2b3baa1865782aff0fd2b00

This model is a fine-tuned version of facebook/mbart-large-cc25 on the Helsinki-NLP/opus_books [fi-no] dataset. It achieves the following results on the evaluation set:

• Loss: 4.4856
• Data Size: 1.0
• Epoch Runtime: 26.0589
• Bleu: 3.1147

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	16.4234	0	2.7042	0.0619
No log	1	85	11.3680	0.0078	3.2041	0.1387
No log	2	170	8.9355	0.0156	3.8216	0.1334
No log	3	255	7.9401	0.0312	4.9067	0.1365
No log	4	340	7.4058	0.0625	6.6895	0.1800
0.5259	5	425	6.6628	0.125	8.0302	0.6464
0.5259	6	510	5.8132	0.25	11.0188	0.8900
3.0205	7	595	20.0273	0.5	16.0060	0.0018
9.7048	8.0	680	5.1119	1.0	29.0627	1.7884
4.4985	9.0	765	4.3440	1.0	28.9546	1.6974
3.7554	10.0	850	4.0683	1.0	26.2930	2.3931
3.2198	11.0	935	3.9265	1.0	26.5348	2.8568
6.1044	12.0	1020	3.9821	1.0	27.6091	2.9750
2.3897	13.0	1105	4.0579	1.0	25.6324	3.1879
2.132	14.0	1190	4.2298	1.0	25.8476	3.0551
1.5613	15.0	1275	4.4856	1.0	26.0589	3.1147

Framework versions

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