72f8ead6bd34da5b20f31b4f45782f04

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

• Loss: 2.6907
• Data Size: 1.0
• Epoch Runtime: 13.4135
• Bleu: 6.9056

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	19.1617	0	1.8577	0.0371
No log	1	77	18.7040	0.0078	2.3486	0.0386
No log	2	154	18.0672	0.0156	2.4706	0.0432
No log	3	231	17.3457	0.0312	3.0730	0.0426
No log	4	308	16.3310	0.0625	3.4843	0.0446
No log	5	385	14.0487	0.125	4.4177	0.0639
1.6356	6	462	11.9825	0.25	5.7809	0.0664
5.9738	7	539	8.7642	0.5	8.9595	0.1177
9.755	8.0	616	6.5462	1.0	14.8261	0.2852
8.1596	9.0	693	4.9882	1.0	14.6642	1.1719
6.5962	10.0	770	4.3808	1.0	13.6869	2.8509
6.123	11.0	847	4.0902	1.0	13.5663	4.5473
5.5678	12.0	924	3.8831	1.0	13.8026	6.0170
5.1927	13.0	1001	3.7217	1.0	13.7691	7.0997
5.0328	14.0	1078	3.5994	1.0	13.7621	8.0758
4.7556	15.0	1155	3.4841	1.0	14.3207	4.0440
4.6266	16.0	1232	3.4026	1.0	16.0493	3.9682
4.4619	17.0	1309	3.3142	1.0	13.2891	4.1916
4.3612	18.0	1386	3.2396	1.0	13.3192	4.4027
4.2083	19.0	1463	3.1748	1.0	14.3403	4.5299
4.1459	20.0	1540	3.1182	1.0	14.1322	4.6811
4.0325	21.0	1617	3.0735	1.0	13.9144	4.8197
3.961	22.0	1694	3.0337	1.0	13.8626	4.8900
3.8951	23.0	1771	2.9961	1.0	14.7550	5.0725
3.8334	24.0	1848	2.9643	1.0	15.1979	5.2900
3.7544	25.0	1925	2.9408	1.0	13.3086	5.5003
3.6625	26.0	2002	2.9122	1.0	13.1137	5.6379
3.6362	27.0	2079	2.9014	1.0	13.7472	5.6542
3.5742	28.0	2156	2.8781	1.0	14.1743	5.8338
3.525	29.0	2233	2.8576	1.0	13.6097	5.8931
3.4736	30.0	2310	2.8528	1.0	13.6878	5.9780
3.4634	31.0	2387	2.8342	1.0	14.2088	6.0422
3.3948	32.0	2464	2.8207	1.0	14.3360	6.0153
3.3701	33.0	2541	2.8055	1.0	13.1667	6.0926
3.345	34.0	2618	2.7958	1.0	13.3323	6.2214
3.3188	35.0	2695	2.7860	1.0	13.9009	6.2773
3.2653	36.0	2772	2.7809	1.0	14.0012	6.2834
3.267	37.0	2849	2.7752	1.0	14.2416	6.3490
3.2001	38.0	2926	2.7736	1.0	14.3047	6.3966
3.2026	39.0	3003	2.7522	1.0	14.7744	6.3813
3.1474	40.0	3080	2.7474	1.0	15.0182	6.4648
3.1017	41.0	3157	2.7368	1.0	13.8057	6.4602
3.0898	42.0	3234	2.7306	1.0	13.8111	6.5112
3.0812	43.0	3311	2.7275	1.0	14.4567	6.5096
3.0772	44.0	3388	2.7285	1.0	14.3655	6.5219
3.0363	45.0	3465	2.7189	1.0	14.4521	6.6068
3.0036	46.0	3542	2.7014	1.0	14.5037	6.7089
2.9677	47.0	3619	2.7051	1.0	14.9735	6.7117
2.97	48.0	3696	2.7042	1.0	14.8996	6.7611
2.9287	49.0	3773	2.6895	1.0	15.0100	6.8334
2.9143	50.0	3850	2.6907	1.0	13.4135	6.9056

Framework versions

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