252e444b575eab5e6b6920703707115f

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

• Loss: 2.7048
• Data Size: 1.0
• Epoch Runtime: 107.9386
• Bleu: 4.1646

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	16.8185	0	9.7091	0.1638
No log	1	688	16.4396	0.0078	10.6734	0.2205
No log	2	1376	15.4148	0.0156	11.9081	0.1916
No log	3	2064	12.7371	0.0312	13.9575	0.2713
0.5767	4	2752	10.0688	0.0625	16.8220	0.2518
0.8794	5	3440	6.8944	0.125	22.8812	0.2819
6.4365	6	4128	4.6648	0.25	35.9062	1.0805
5.2	7	4816	3.9887	0.5	60.2414	0.9737
4.4643	8.0	5504	3.4851	1.0	111.0302	1.4972
4.2029	9.0	6192	3.3275	1.0	111.2567	1.8706
4.011	10.0	6880	3.2396	1.0	109.4068	2.1290
3.8989	11.0	7568	3.1762	1.0	110.6279	2.2698
3.7675	12.0	8256	3.1212	1.0	110.7316	2.4218
3.703	13.0	8944	3.0880	1.0	110.9528	2.5715
3.666	14.0	9632	3.0493	1.0	111.2139	2.6786
3.5767	15.0	10320	3.0264	1.0	110.0062	2.7794
3.5211	16.0	11008	2.9992	1.0	109.6109	2.8721
3.4903	17.0	11696	2.9718	1.0	111.9554	2.9183
3.4116	18.0	12384	2.9549	1.0	110.8191	2.9965
3.4001	19.0	13072	2.9316	1.0	109.8554	3.0525
3.383	20.0	13760	2.9168	1.0	110.4508	3.1468
3.3254	21.0	14448	2.9005	1.0	111.2983	3.2101
3.3037	22.0	15136	2.8937	1.0	111.4072	3.2799
3.2432	23.0	15824	2.8767	1.0	111.1446	3.3293
3.2361	24.0	16512	2.8614	1.0	112.3162	3.3471
3.2076	25.0	17200	2.8521	1.0	111.1131	3.3737
3.1398	26.0	17888	2.8393	1.0	112.0510	3.4616
3.1293	27.0	18576	2.8340	1.0	110.8477	3.4917
3.1118	28.0	19264	2.8205	1.0	111.7321	3.5628
3.1101	29.0	19952	2.8069	1.0	111.8174	3.5736
3.0583	30.0	20640	2.8023	1.0	111.9836	3.6254
3.0587	31.0	21328	2.7971	1.0	110.5331	3.6616
3.0171	32.0	22016	2.7849	1.0	110.4607	3.6894
3.0193	33.0	22704	2.7772	1.0	109.0230	3.7525
2.9894	34.0	23392	2.7695	1.0	107.6458	3.7653
2.9708	35.0	24080	2.7636	1.0	107.4148	3.7776
2.9556	36.0	24768	2.7605	1.0	108.0904	3.8465
2.9331	37.0	25456	2.7578	1.0	107.4862	3.8456
2.9557	38.0	26144	2.7503	1.0	107.4639	3.9086
2.8623	39.0	26832	2.7475	1.0	107.7396	3.9209
2.876	40.0	27520	2.7413	1.0	108.5188	3.9477
2.8569	41.0	28208	2.7377	1.0	107.9827	3.9703
2.838	42.0	28896	2.7346	1.0	107.2617	4.0063
2.8481	43.0	29584	2.7263	1.0	108.0689	4.0092
2.793	44.0	30272	2.7234	1.0	107.9407	4.0551
2.7863	45.0	30960	2.7221	1.0	108.2426	4.0762
2.7853	46.0	31648	2.7182	1.0	108.6926	4.0772
2.7543	47.0	32336	2.7112	1.0	107.8927	4.1124
2.7247	48.0	33024	2.7077	1.0	107.5613	4.1342
2.7427	49.0	33712	2.7111	1.0	107.9450	4.1599
2.7071	50.0	34400	2.7048	1.0	107.9386	4.1646

Framework versions

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