3c58d0bab7c50f6e8dfc30ee52147218

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

• Loss: 2.6168
• Data Size: 1.0
• Epoch Runtime: 62.5072
• Bleu: 6.9745

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	15.6686	0	5.8408	0.2469
No log	1	390	15.3559	0.0078	6.4347	0.2206
No log	2	780	14.3436	0.0156	7.0920	0.2455
No log	3	1170	13.0057	0.0312	8.1606	0.3001
No log	4	1560	11.1104	0.0625	10.1428	0.3368
0.8431	5	1950	8.5109	0.125	13.8434	0.4086
1.4743	6	2340	5.8619	0.25	21.4355	0.4794
6.3219	7	2730	4.3964	0.5	33.9841	3.2816
4.9917	8.0	3120	3.7946	1.0	61.1269	2.4625
4.4862	9.0	3510	3.4671	1.0	60.8731	3.3469
4.2146	10.0	3900	3.3099	1.0	59.8231	3.8854
4.0429	11.0	4290	3.2172	1.0	60.1998	4.2481
3.913	12.0	4680	3.1442	1.0	61.5796	4.4836
3.8373	13.0	5070	3.0960	1.0	61.2461	4.6702
3.7399	14.0	5460	3.0555	1.0	60.0812	4.8506
3.6047	15.0	5850	3.0130	1.0	60.4809	5.0004
3.5615	16.0	6240	2.9855	1.0	59.5285	5.1945
3.5496	17.0	6630	2.9604	1.0	59.1054	5.2738
3.4475	18.0	7020	2.9298	1.0	60.1827	5.4059
3.3998	19.0	7410	2.9004	1.0	59.8943	5.4838
3.3728	20.0	7800	2.8899	1.0	61.1778	5.5841
3.3449	21.0	8190	2.8693	1.0	61.2387	5.6640
3.2885	22.0	8580	2.8398	1.0	61.3963	5.7439
3.2772	23.0	8970	2.8316	1.0	59.5462	5.8515
3.2107	24.0	9360	2.8136	1.0	59.9638	5.8933
3.1825	25.0	9750	2.8082	1.0	61.0646	5.9557
3.1763	26.0	10140	2.7892	1.0	59.8982	6.0271
3.1125	27.0	10530	2.7771	1.0	60.4019	6.0954
3.1258	28.0	10920	2.7660	1.0	59.3301	6.1228
3.0994	29.0	11310	2.7471	1.0	60.0506	6.1887
3.0527	30.0	11700	2.7310	1.0	60.1198	6.2310
2.9866	31.0	12090	2.7313	1.0	60.4529	6.2855
2.9615	32.0	12480	2.7176	1.0	61.4655	6.3646
2.9527	33.0	12870	2.7145	1.0	60.0622	6.4026
2.9215	34.0	13260	2.7037	1.0	60.4421	6.4369
2.9249	35.0	13650	2.7005	1.0	62.3335	6.4936
2.9285	36.0	14040	2.6947	1.0	62.6724	6.4983
2.8703	37.0	14430	2.6830	1.0	62.9112	6.5631
2.8218	38.0	14820	2.6731	1.0	62.3739	6.6175
2.8165	39.0	15210	2.6701	1.0	63.4032	6.6356
2.7918	40.0	15600	2.6651	1.0	64.0902	6.6879
2.7974	41.0	15990	2.6552	1.0	62.0214	6.7311
2.802	42.0	16380	2.6527	1.0	62.3090	6.7668
2.7434	43.0	16770	2.6454	1.0	62.3300	6.7609
2.7407	44.0	17160	2.6422	1.0	63.2888	6.8010
2.724	45.0	17550	2.6369	1.0	62.5514	6.8637
2.7156	46.0	17940	2.6350	1.0	63.5730	6.8781
2.6456	47.0	18330	2.6227	1.0	63.4639	6.9180
2.7103	48.0	18720	2.6244	1.0	62.5968	6.9265
2.646	49.0	19110	2.6157	1.0	63.1561	6.9574
2.6257	50.0	19500	2.6168	1.0	62.5072	6.9745

Framework versions

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