a003c453d3261decfeda85bb22b3551f

This model is a fine-tuned version of google/mt5-base on the Helsinki-NLP/opus_books [fi-no] dataset. It achieves the following results on the evaluation set:

• Loss: 2.5250
• Data Size: 1.0
• Epoch Runtime: 21.4854
• Bleu: 4.1715

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	14.9863	0	2.3900	0.0036
No log	1	85	14.7708	0.0078	2.7604	0.0039
No log	2	170	14.3652	0.0156	3.9297	0.0039
No log	3	255	14.2351	0.0312	5.4341	0.0012
No log	4	340	13.0767	0.0625	6.6130	0.0026
1.0167	5	425	11.7008	0.125	8.1159	0.0029
1.0167	6	510	10.4538	0.25	11.1138	0.0028
3.839	7	595	8.1347	0.5	14.5761	0.0039
7.8471	8.0	680	4.1510	1.0	23.7922	0.1332
4.5158	9.0	765	3.3324	1.0	21.0985	0.8412
3.9229	10.0	850	3.0150	1.0	21.8203	1.6279
3.7322	11.0	935	2.8910	1.0	24.0187	1.9297
3.4665	12.0	1020	2.7405	1.0	20.8419	2.6863
3.3055	13.0	1105	2.6915	1.0	21.2144	2.9196
3.1974	14.0	1190	2.6553	1.0	20.7735	2.9431
3.0983	15.0	1275	2.6283	1.0	20.7107	3.3418
2.9776	16.0	1360	2.5948	1.0	20.4475	3.4987
2.8959	17.0	1445	2.5797	1.0	21.3028	3.4890
2.8235	18.0	1530	2.5554	1.0	22.5032	3.8153
2.7529	19.0	1615	2.5545	1.0	22.4183	3.6873
2.6671	20.0	1700	2.5279	1.0	20.6834	3.8222
2.6083	21.0	1785	2.5273	1.0	21.1201	3.8469
2.5646	22.0	1870	2.5159	1.0	22.0378	3.9009
2.4816	23.0	1955	2.5207	1.0	22.7214	4.0061
2.4394	24.0	2040	2.5110	1.0	20.8485	4.0459
2.3732	25.0	2125	2.5241	1.0	21.0050	4.0822
2.3486	26.0	2210	2.5189	1.0	21.6269	4.2444
2.2982	27.0	2295	2.5213	1.0	22.6163	4.1839
2.255	28.0	2380	2.5250	1.0	21.4854	4.1715

Framework versions

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