fc1675c452f6ade17e19d10e66fc1dc8

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

• Loss: 2.5109
• Data Size: 1.0
• Epoch Runtime: 107.4434
• Bleu: 5.3511

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	18.2126	0	9.7229	0.2821
No log	1	684	17.1843	0.0078	10.4601	0.2857
No log	2	1368	15.8314	0.0156	11.3541	0.3103
No log	3	2052	12.9768	0.0312	13.2571	0.3516
No log	4	2736	9.6351	0.0625	16.1274	0.3873
9.9321	5	3420	6.6460	0.125	22.3163	0.2937
6.6798	6	4104	4.6406	0.25	34.2147	1.1370
5.324	7	4788	3.9731	0.5	58.2977	0.9879
4.4751	8.0	5472	3.4211	1.0	105.3376	1.7759
4.144	9.0	6156	3.2472	1.0	107.5736	2.2586
3.9925	10.0	6840	3.1491	1.0	105.1886	2.5878
3.8374	11.0	7524	3.0726	1.0	105.8710	2.8416
3.7373	12.0	8208	3.0195	1.0	107.2988	3.0379
3.607	13.0	8892	2.9684	1.0	106.8523	3.2369
3.6247	14.0	9576	2.9359	1.0	106.6629	3.3520
3.5038	15.0	10260	2.8978	1.0	107.2863	3.4596
3.4472	16.0	10944	2.8660	1.0	107.9750	3.5886
3.381	17.0	11628	2.8427	1.0	108.8808	3.6989
3.3438	18.0	12312	2.8194	1.0	107.1700	3.7762
3.3045	19.0	12996	2.7957	1.0	107.2615	3.8943
3.2534	20.0	13680	2.7769	1.0	108.6581	4.0035
3.1617	21.0	14364	2.7602	1.0	108.4630	4.0996
3.1425	22.0	15048	2.7414	1.0	106.6954	4.1840
3.1421	23.0	15732	2.7296	1.0	106.6434	4.2296
3.1	24.0	16416	2.7118	1.0	107.8798	4.2947
3.0595	25.0	17100	2.6938	1.0	110.3338	4.3791
3.0455	26.0	17784	2.6815	1.0	106.4931	4.4342
3.0074	27.0	18468	2.6724	1.0	106.8790	4.4725
2.9806	28.0	19152	2.6584	1.0	110.1350	4.5494
2.9533	29.0	19836	2.6477	1.0	109.3596	4.5771
2.8919	30.0	20520	2.6434	1.0	107.8245	4.6223
2.89	31.0	21204	2.6337	1.0	107.7325	4.6883
2.9254	32.0	21888	2.6207	1.0	108.0603	4.7041
2.8883	33.0	22572	2.6148	1.0	110.0230	4.7649
2.8092	34.0	23256	2.6023	1.0	106.6470	4.8163
2.8076	35.0	23940	2.5957	1.0	107.1131	4.8667
2.8102	36.0	24624	2.5849	1.0	107.4508	4.8860
2.74	37.0	25308	2.5849	1.0	108.6266	4.9398
2.754	38.0	25992	2.5701	1.0	107.4365	4.9580
2.6915	39.0	26676	2.5653	1.0	107.4098	4.9882
2.682	40.0	27360	2.5630	1.0	108.5210	5.0310
2.6995	41.0	28044	2.5529	1.0	108.1337	5.0781
2.7046	42.0	28728	2.5460	1.0	107.4749	5.1326
2.6707	43.0	29412	2.5435	1.0	108.1177	5.1529
2.6692	44.0	30096	2.5349	1.0	108.4466	5.1935
2.6215	45.0	30780	2.5305	1.0	108.6350	5.1652
2.6401	46.0	31464	2.5295	1.0	107.5707	5.2422
2.6027	47.0	32148	2.5177	1.0	108.3829	5.2745
2.583	48.0	32832	2.5178	1.0	108.8079	5.3087
2.5528	49.0	33516	2.5134	1.0	108.3599	5.3373
2.5637	50.0	34200	2.5109	1.0	107.4434	5.3511

Framework versions

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