de4c754a1206f7011f5df265763330b1

This model is a fine-tuned version of google/umt5-base on the Helsinki-NLP/opus_books [it-nl] dataset. It achieves the following results on the evaluation set:

• Loss: 2.8579
• Data Size: 1.0
• Epoch Runtime: 17.3523
• Bleu: 4.5848

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	12.1927	0	1.9400	0.0177
No log	1	58	12.1568	0.0078	2.2327	0.0125
No log	2	116	12.0818	0.0156	2.4813	0.0170
No log	3	174	12.0587	0.0312	4.1362	0.0193
No log	4	232	11.9603	0.0625	4.7242	0.0237
No log	5	290	11.9706	0.125	6.8797	0.0163
1.6393	6	348	11.5785	0.25	8.6741	0.0234
2.3333	7	406	10.0356	0.5	12.3831	0.0344
9.1968	8.0	464	7.5373	1.0	20.3630	0.1810
9.6598	9.0	522	6.2582	1.0	16.1029	0.2735
7.2745	10.0	580	4.2682	1.0	16.4966	2.1169
5.4905	11.0	638	3.8061	1.0	17.0638	3.6345
4.8114	12.0	696	3.5253	1.0	17.5070	2.0219
4.2331	13.0	754	3.3572	1.0	17.4705	2.3782
4.057	14.0	812	3.2744	1.0	15.9843	2.7085
3.908	15.0	870	3.2033	1.0	16.3924	2.8857
3.7982	16.0	928	3.1446	1.0	17.6421	2.9933
3.7387	17.0	986	3.1115	1.0	18.3489	3.1300
3.6216	18.0	1044	3.0738	1.0	18.6649	3.1903
3.5197	19.0	1102	3.0507	1.0	16.6629	3.2863
3.4506	20.0	1160	3.0228	1.0	16.6362	3.4272
3.3894	21.0	1218	3.0067	1.0	17.2978	3.4488
3.356	22.0	1276	2.9862	1.0	17.2369	3.5777
3.3001	23.0	1334	2.9763	1.0	18.5782	3.6111
3.2678	24.0	1392	2.9550	1.0	16.7115	3.6258
3.1914	25.0	1450	2.9487	1.0	16.9935	3.7071
3.1448	26.0	1508	2.9270	1.0	17.5448	3.7992
3.1012	27.0	1566	2.9113	1.0	17.6353	3.8902
3.0836	28.0	1624	2.9047	1.0	18.0783	3.8852
3.0621	29.0	1682	2.8972	1.0	18.9103	3.9842
3.0107	30.0	1740	2.9011	1.0	17.3775	3.9823
2.9832	31.0	1798	2.8856	1.0	17.2965	4.0219
2.9152	32.0	1856	2.8870	1.0	17.3956	4.0612
2.8881	33.0	1914	2.8761	1.0	18.6104	4.0488
2.8558	34.0	1972	2.8761	1.0	19.3654	4.1790
2.833	35.0	2030	2.8713	1.0	16.9957	4.2046
2.7955	36.0	2088	2.8658	1.0	17.2385	4.3333
2.7774	37.0	2146	2.8642	1.0	18.4200	4.2782
2.7189	38.0	2204	2.8634	1.0	17.3501	4.3479
2.6733	39.0	2262	2.8667	1.0	17.7798	4.2935
2.675	40.0	2320	2.8621	1.0	17.9407	4.3301
2.6392	41.0	2378	2.8623	1.0	16.9314	4.3615
2.6231	42.0	2436	2.8601	1.0	17.2746	4.4111
2.5919	43.0	2494	2.8550	1.0	17.5905	4.4534
2.5534	44.0	2552	2.8528	1.0	17.5749	4.4871
2.52	45.0	2610	2.8556	1.0	17.8637	4.4545
2.5128	46.0	2668	2.8600	1.0	18.9311	4.5891
2.4698	47.0	2726	2.8610	1.0	16.8442	4.6278
2.4708	48.0	2784	2.8579	1.0	17.3523	4.5848

Framework versions

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