5bcc5396b07e513c56df8d227037bba1

This model is a fine-tuned version of google/umt5-base on the Helsinki-NLP/opus_books [fi-no] dataset. It achieves the following results on the evaluation set:

• Loss: 2.8014
• Data Size: 1.0
• Epoch Runtime: 22.4074
• Bleu: 5.4188

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	13.3193	0	2.3106	0.0333
No log	1	85	13.3120	0.0078	2.6831	0.0356
No log	2	170	13.0464	0.0156	3.6001	0.0370
No log	3	255	12.7136	0.0312	4.2199	0.0534
No log	4	340	12.1567	0.0625	5.2327	0.0324
1.1749	5	425	11.6050	0.125	6.5038	0.0318
1.1749	6	510	10.4120	0.25	8.6614	0.0432
4.5615	7	595	8.7903	0.5	13.4662	0.1144
10.1288	8.0	680	6.0192	1.0	22.9811	0.2258
6.163	9.0	765	4.0627	1.0	21.3025	2.5443
4.8451	10.0	850	3.4734	1.0	22.5832	1.9831
4.4945	11.0	935	3.2271	1.0	23.8550	2.8005
4.0415	12.0	1020	3.0752	1.0	21.7356	3.3590
3.768	13.0	1105	3.0154	1.0	22.7674	3.6688
3.6396	14.0	1190	2.9644	1.0	22.1751	3.9112
3.5757	15.0	1275	2.9374	1.0	21.2885	4.1408
3.42	16.0	1360	2.8955	1.0	22.3561	4.3539
3.3068	17.0	1445	2.8665	1.0	23.3444	4.5308
3.2127	18.0	1530	2.8577	1.0	24.0651	4.6450
3.1268	19.0	1615	2.8511	1.0	24.2527	4.7717
3.0505	20.0	1700	2.8305	1.0	21.8575	4.8608
2.991	21.0	1785	2.8234	1.0	22.0078	4.9905
2.9464	22.0	1870	2.8105	1.0	22.4925	4.9631
2.8467	23.0	1955	2.8010	1.0	22.3577	5.0732
2.8164	24.0	2040	2.7989	1.0	21.8246	5.1650
2.741	25.0	2125	2.7941	1.0	22.8675	5.1468
2.7071	26.0	2210	2.7833	1.0	23.4260	5.1629
2.6593	27.0	2295	2.7941	1.0	24.4490	5.3361
2.6095	28.0	2380	2.7955	1.0	24.7949	5.3050
2.5528	29.0	2465	2.7942	1.0	21.8180	5.3055
2.5013	30.0	2550	2.8014	1.0	22.4074	5.4188

Framework versions

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