8a681bd5e48b1876819df01f82399e28

This model is a fine-tuned version of google/mt5-small on the Helsinki-NLP/opus_books [fi-pl] dataset. It achieves the following results on the evaluation set:

• Loss: 3.1186
• Data Size: 1.0
• Epoch Runtime: 12.7358
• Bleu: 0.7448

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	29.2432	0	1.6355	0.0040
No log	1	70	29.0176	0.0078	2.7695	0.0080
No log	2	140	26.7250	0.0156	1.9340	0.0071
No log	3	210	24.7823	0.0312	2.4169	0.0096
No log	4	280	22.9952	0.0625	2.5776	0.0068
No log	5	350	19.9288	0.125	3.4689	0.0101
No log	6	420	15.9629	0.25	4.9484	0.0116
3.181	7	490	12.7236	0.5	7.0200	0.0051
12.7957	8.0	560	8.4316	1.0	12.1829	0.0183
9.9094	9.0	630	5.9163	1.0	12.0417	0.0204
6.3213	10.0	700	4.2825	1.0	11.0516	0.0636
5.4421	11.0	770	3.8785	1.0	11.8811	0.1150
5.1096	12.0	840	3.7124	1.0	11.1901	0.1584
4.7232	13.0	910	3.6135	1.0	11.4277	0.2136
4.5929	14.0	980	3.5475	1.0	11.3017	0.3249
4.4352	15.0	1050	3.5057	1.0	11.6904	0.3929
4.3344	16.0	1120	3.4632	1.0	11.9123	0.3688
4.3018	17.0	1190	3.4292	1.0	11.8033	0.3894
4.1827	18.0	1260	3.3994	1.0	12.7377	0.4141
4.15	19.0	1330	3.3746	1.0	11.3964	0.4392
4.0598	20.0	1400	3.3483	1.0	11.1041	0.5399
3.9852	21.0	1470	3.3295	1.0	11.9825	0.4914
3.9972	22.0	1540	3.3044	1.0	11.4271	0.5344
3.8871	23.0	1610	3.2907	1.0	11.6221	0.5166
3.881	24.0	1680	3.2773	1.0	12.4179	0.5094
3.8228	25.0	1750	3.2656	1.0	12.6361	0.5314
3.7924	26.0	1820	3.2531	1.0	12.9146	0.5420
3.7602	27.0	1890	3.2372	1.0	11.2274	0.5487
3.7062	28.0	1960	3.2300	1.0	11.4544	0.5856
3.69	29.0	2030	3.2221	1.0	11.5363	0.6032
3.6605	30.0	2100	3.2113	1.0	12.1520	0.6290
3.6333	31.0	2170	3.2026	1.0	12.5515	0.6006
3.6323	32.0	2240	3.1991	1.0	12.3393	0.6086
3.5883	33.0	2310	3.1877	1.0	12.6489	0.6431
3.5566	34.0	2380	3.1791	1.0	12.5892	0.6643
3.5366	35.0	2450	3.1717	1.0	12.8103	0.6764
3.5038	36.0	2520	3.1706	1.0	11.6256	0.7019
3.4822	37.0	2590	3.1654	1.0	11.5844	0.6915
3.4731	38.0	2660	3.1596	1.0	11.6021	0.6762
3.4375	39.0	2730	3.1597	1.0	12.1233	0.6931
3.4068	40.0	2800	3.1518	1.0	12.4407	0.6922
3.3934	41.0	2870	3.1518	1.0	13.1373	0.6849
3.3942	42.0	2940	3.1437	1.0	13.3477	0.7083
3.3409	43.0	3010	3.1393	1.0	13.4745	0.7043
3.3472	44.0	3080	3.1379	1.0	13.1708	0.7050
3.3113	45.0	3150	3.1343	1.0	11.7511	0.7387
3.2914	46.0	3220	3.1290	1.0	12.0681	0.7506
3.2856	47.0	3290	3.1196	1.0	12.2375	0.7463
3.269	48.0	3360	3.1260	1.0	12.4927	0.7377
3.2462	49.0	3430	3.1207	1.0	12.5982	0.7650
3.2483	50.0	3500	3.1186	1.0	12.7358	0.7448

Framework versions

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