94fdaedfc9b48f0ed0b5bf29cb185c4e

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

• Loss: 2.9088
• Data Size: 1.0
• Epoch Runtime: 16.3464
• Bleu: 2.6969

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	17.5427	0	1.9032	0.0327
No log	1	88	17.6538	0.0078	2.3328	0.0366
No log	2	176	17.6828	0.0156	2.6445	0.0330
No log	3	264	17.1684	0.0312	3.2106	0.0415
No log	4	352	16.5233	0.0625	4.1401	0.0369
No log	5	440	15.7354	0.125	4.9029	0.0416
1.4026	6	528	13.0467	0.25	6.2121	0.0460
5.2705	7	616	9.4958	0.5	9.2868	0.0785
9.3402	8.0	704	6.3986	1.0	15.7509	0.1740
7.883	9.0	792	4.8116	1.0	15.0764	0.2808
6.3317	10.0	880	4.4173	1.0	15.4257	0.7106
5.6835	11.0	968	4.2185	1.0	15.6292	0.6120
5.2724	12.0	1056	4.0331	1.0	16.0407	0.6033
5.1159	13.0	1144	3.8686	1.0	15.7916	0.6826
4.8563	14.0	1232	3.6990	1.0	16.0089	0.8737
4.6141	15.0	1320	3.5557	1.0	16.7219	1.0237
4.4255	16.0	1408	3.4427	1.0	15.0475	1.2239
4.3427	17.0	1496	3.3700	1.0	16.3657	1.2493
4.213	18.0	1584	3.3003	1.0	16.0512	1.3956
4.1149	19.0	1672	3.2423	1.0	15.8517	1.5251
4.0094	20.0	1760	3.1971	1.0	15.9138	1.6057
3.9619	21.0	1848	3.1592	1.0	16.2662	1.6730
3.9209	22.0	1936	3.1399	1.0	16.9278	1.7254
3.8336	23.0	2024	3.1136	1.0	15.0629	1.8268
3.79	24.0	2112	3.0924	1.0	15.4468	1.9491
3.7128	25.0	2200	3.0796	1.0	16.4523	1.9698
3.6885	26.0	2288	3.0701	1.0	15.4163	2.0019
3.612	27.0	2376	3.0550	1.0	16.1792	2.0334
3.5716	28.0	2464	3.0425	1.0	15.9046	2.0828
3.5489	29.0	2552	3.0267	1.0	15.9448	2.1437
3.508	30.0	2640	3.0195	1.0	14.9253	2.1404
3.45	31.0	2728	3.0112	1.0	15.3964	2.1229
3.4475	32.0	2816	3.0026	1.0	15.4170	2.1722
3.4265	33.0	2904	2.9929	1.0	15.7752	2.2442
3.3774	34.0	2992	2.9869	1.0	15.7856	2.2355
3.3686	35.0	3080	2.9780	1.0	16.1228	2.2419
3.3166	36.0	3168	2.9766	1.0	16.1321	2.2914
3.3171	37.0	3256	2.9657	1.0	14.8927	2.3810
3.2881	38.0	3344	2.9582	1.0	15.4943	2.3324
3.2738	39.0	3432	2.9541	1.0	16.0475	2.4149
3.2355	40.0	3520	2.9469	1.0	15.4458	2.5267
3.1962	41.0	3608	2.9466	1.0	15.3254	2.5223
3.1724	42.0	3696	2.9407	1.0	15.8035	2.5439
3.1306	43.0	3784	2.9319	1.0	15.7676	2.5258
3.1571	44.0	3872	2.9286	1.0	15.8733	2.5348
3.1056	45.0	3960	2.9272	1.0	15.0118	2.5468
3.0972	46.0	4048	2.9254	1.0	15.4880	2.6190
3.0718	47.0	4136	2.9178	1.0	15.5234	2.6083
3.0491	48.0	4224	2.9140	1.0	15.7977	2.6580
3.041	49.0	4312	2.9099	1.0	16.3834	2.6792
3.0405	50.0	4400	2.9088	1.0	16.3464	2.6969

Framework versions

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