e3babb8015398f22a63fe6901713d68e

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

• Loss: 3.4856
• Data Size: 1.0
• Epoch Runtime: 13.2890
• Bleu: 2.4818

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	18.1002	0	1.7283	0.0562
No log	1	75	18.1645	0.0078	1.9587	0.0345
No log	2	150	17.2923	0.0156	2.3066	0.0580
No log	3	225	16.4469	0.0312	2.9083	0.0569
No log	4	300	15.5651	0.0625	3.3489	0.0556
No log	5	375	14.3630	0.125	4.2416	0.0595
No log	6	450	12.1033	0.25	5.4460	0.0447
No log	7	525	9.0282	0.5	8.2194	0.0658
8.9277	8.0	600	6.5578	1.0	13.9101	0.1856
7.7238	9.0	675	5.2655	1.0	14.2897	0.4651
6.5062	10.0	750	4.9267	1.0	13.0144	0.5096
6.2022	11.0	825	4.7138	1.0	13.0152	0.7079
5.8041	12.0	900	4.5339	1.0	12.9225	1.2388
5.5943	13.0	975	4.4047	1.0	12.8617	1.5846
5.3682	14.0	1050	4.2823	1.0	12.8135	1.9420
5.2353	15.0	1125	4.1989	1.0	12.8609	1.0069
5.0499	16.0	1200	4.0935	1.0	13.3976	1.1048
4.9827	17.0	1275	4.0182	1.0	12.6508	1.2535
4.8603	18.0	1350	3.9544	1.0	12.9864	1.3214
4.787	19.0	1425	3.9019	1.0	13.4680	1.4019
4.6777	20.0	1500	3.8460	1.0	13.8981	1.4917
4.6164	21.0	1575	3.8066	1.0	14.8219	1.5288
4.5516	22.0	1650	3.7706	1.0	14.7882	1.6005
4.4776	23.0	1725	3.7467	1.0	14.7274	1.6067
4.4162	24.0	1800	3.7196	1.0	14.5843	1.7865
4.4035	25.0	1875	3.6955	1.0	12.8680	1.9358
4.3159	26.0	1950	3.6742	1.0	12.9960	1.8525
4.3001	27.0	2025	3.6600	1.0	13.4016	1.9151
4.2257	28.0	2100	3.6387	1.0	14.1796	2.0088
4.1948	29.0	2175	3.6233	1.0	15.2012	2.0336
4.1688	30.0	2250	3.6159	1.0	14.7225	2.0623
4.138	31.0	2325	3.6047	1.0	14.7262	2.1031
4.1102	32.0	2400	3.5928	1.0	15.0099	2.1592
4.0425	33.0	2475	3.5763	1.0	13.2776	2.1841
4.0315	34.0	2550	3.5702	1.0	13.7277	2.2052
3.9818	35.0	2625	3.5615	1.0	14.4340	2.2627
3.9634	36.0	2700	3.5535	1.0	14.3716	2.2744
3.9299	37.0	2775	3.5521	1.0	14.8856	2.2306
3.9164	38.0	2850	3.5422	1.0	14.8994	2.3079
3.8966	39.0	2925	3.5384	1.0	14.8891	2.3486
3.8606	40.0	3000	3.5302	1.0	15.0663	2.3331
3.822	41.0	3075	3.5221	1.0	15.6057	2.4048
3.7902	42.0	3150	3.5173	1.0	13.2500	2.3580
3.7706	43.0	3225	3.5143	1.0	13.1484	2.4666
3.7576	44.0	3300	3.5095	1.0	13.2237	2.3792
3.7302	45.0	3375	3.5057	1.0	13.7542	2.4260
3.6948	46.0	3450	3.4992	1.0	13.5008	2.4733
3.6967	47.0	3525	3.4913	1.0	13.9558	2.4751
3.6627	48.0	3600	3.4882	1.0	14.6121	2.5331
3.6272	49.0	3675	3.4887	1.0	16.1375	2.5500
3.6257	50.0	3750	3.4856	1.0	13.2890	2.4818

Framework versions

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