fe0067602ffc58054d328baa11512e53

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

• Loss: 2.5716
• Data Size: 1.0
• Epoch Runtime: 68.2664
• Bleu: 4.7904

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	16.9363	0	6.2257	0.2295
No log	1	419	16.4046	0.0078	6.8040	0.1692
No log	2	838	15.3530	0.0156	7.8658	0.2415
0.5259	3	1257	13.3344	0.0312	8.6694	0.2233
0.5259	4	1676	10.1193	0.0625	10.4582	0.2867
0.8973	5	2095	7.0832	0.125	14.0381	0.4247
0.9941	6	2514	4.8725	0.25	21.9718	1.0292
5.543	7	2933	4.0364	0.5	36.4432	2.6491
4.5345	8.0	3352	3.4394	1.0	67.3780	1.8519
4.1989	9.0	3771	3.2016	1.0	65.9117	2.3605
4.0201	10.0	4190	3.1116	1.0	66.6560	2.6499
3.8094	11.0	4609	3.0445	1.0	66.6173	2.8497
3.762	12.0	5028	2.9907	1.0	68.2071	3.0438
3.607	13.0	5447	2.9527	1.0	66.4570	3.1419
3.6053	14.0	5866	2.9145	1.0	66.8698	3.2869
3.4769	15.0	6285	2.8889	1.0	66.6928	3.3860
3.4706	16.0	6704	2.8592	1.0	66.1817	3.4789
3.398	17.0	7123	2.8366	1.0	67.2897	3.5737
3.337	18.0	7542	2.8161	1.0	66.3748	3.6925
3.3071	19.0	7961	2.7961	1.0	68.0592	3.7006
3.2478	20.0	8380	2.7807	1.0	66.8403	3.7873
3.2372	21.0	8799	2.7652	1.0	66.6487	3.8471
3.1859	22.0	9218	2.7470	1.0	66.6997	3.9231
3.1804	23.0	9637	2.7382	1.0	67.0227	3.9345
3.0999	24.0	10056	2.7254	1.0	66.9358	4.0121
3.0769	25.0	10475	2.7176	1.0	67.4841	4.0503
3.0325	26.0	10894	2.7020	1.0	66.8775	4.1205
3.0314	27.0	11313	2.6945	1.0	67.6380	4.1286
3.0271	28.0	11732	2.6808	1.0	67.6813	4.1524
3.0078	29.0	12151	2.6764	1.0	67.2010	4.2473
2.9698	30.0	12570	2.6673	1.0	67.1944	4.2897
2.9458	31.0	12989	2.6592	1.0	67.6115	4.3278
2.9098	32.0	13408	2.6556	1.0	68.5527	4.3470
2.8914	33.0	13827	2.6473	1.0	67.4845	4.3764
2.8482	34.0	14246	2.6433	1.0	66.6381	4.4274
2.8809	35.0	14665	2.6297	1.0	67.1876	4.4493
2.8277	36.0	15084	2.6298	1.0	67.0466	4.4916
2.7705	37.0	15503	2.6269	1.0	66.8672	4.5097
2.7754	38.0	15922	2.6216	1.0	66.6425	4.5756
2.7631	39.0	16341	2.6131	1.0	68.8301	4.5331
2.7462	40.0	16760	2.6092	1.0	66.9419	4.5978
2.7292	41.0	17179	2.6047	1.0	68.3140	4.6401
2.711	42.0	17598	2.5998	1.0	66.9488	4.6628
2.7046	43.0	18017	2.5940	1.0	68.0415	4.6829
2.6644	44.0	18436	2.5930	1.0	68.1630	4.6488
2.6406	45.0	18855	2.5867	1.0	66.8644	4.7508
2.6352	46.0	19274	2.5872	1.0	68.0415	4.6944
2.6503	47.0	19693	2.5811	1.0	67.5611	4.7259
2.5851	48.0	20112	2.5854	1.0	68.1429	4.7679
2.6197	49.0	20531	2.5776	1.0	66.6917	4.7975
2.5818	50.0	20950	2.5716	1.0	68.2664	4.7904

Framework versions

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