45cde632e48083150b2baa2c26179d42

This model is a fine-tuned version of google/mt5-small on the Helsinki-NLP/opus_books [de-pt] dataset. It achieves the following results on the evaluation set:

• Loss: 2.5995
• Data Size: 1.0
• Epoch Runtime: 6.7684
• Bleu: 3.2270

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	31.3452	0	1.1404	0.0044
No log	1	27	30.9914	0.0078	1.5898	0.0061
No log	2	54	30.3526	0.0156	1.6757	0.0056
No log	3	81	29.6529	0.0312	2.0404	0.0071
No log	4	108	28.9018	0.0625	2.4687	0.0045
No log	5	135	25.1327	0.125	2.9368	0.0076
No log	6	162	19.7554	0.25	3.2487	0.0060
No log	7	189	15.7126	0.5	4.3466	0.0094
4.4185	8.0	216	12.0272	1.0	6.4605	0.0154
4.4185	9.0	243	9.8194	1.0	5.9703	0.0141
13.998	10.0	270	8.0210	1.0	5.9067	0.0107
13.998	11.0	297	6.6757	1.0	5.9823	0.0119
9.7665	12.0	324	5.7102	1.0	6.4256	0.0096
7.2327	13.0	351	5.1367	1.0	6.5897	0.0204
7.2327	14.0	378	4.4280	1.0	6.8430	0.0239
5.7897	15.0	405	3.7653	1.0	5.3993	0.0450
5.7897	16.0	432	3.5210	1.0	5.7351	0.4392
4.8733	17.0	459	3.3821	1.0	5.5197	0.7493
4.8733	18.0	486	3.2930	1.0	5.6848	0.9425
4.4002	19.0	513	3.2074	1.0	6.0612	1.0981
4.4002	20.0	540	3.1383	1.0	6.3865	1.1585
4.1301	21.0	567	3.0881	1.0	7.0542	1.1100
4.1301	22.0	594	3.0515	1.0	6.7974	1.2151
3.9187	23.0	621	3.0005	1.0	6.9739	1.2858
3.9187	24.0	648	2.9695	1.0	7.2571	1.5159
3.7751	25.0	675	2.9338	1.0	7.1805	1.5693
3.6794	26.0	702	2.9014	1.0	7.6884	1.6659
3.6794	27.0	729	2.8833	1.0	7.4668	1.7657
3.5659	28.0	756	2.8624	1.0	8.1261	1.8383
3.5659	29.0	783	2.8419	1.0	8.3726	1.9546
3.486	30.0	810	2.8249	1.0	8.2575	2.0172
3.486	31.0	837	2.8029	1.0	5.4925	2.0143
3.4077	32.0	864	2.7929	1.0	5.7737	2.2607
3.4077	33.0	891	2.7721	1.0	5.7768	2.3054
3.3307	34.0	918	2.7613	1.0	6.1863	2.3481
3.3307	35.0	945	2.7427	1.0	6.3883	2.3710
3.2612	36.0	972	2.7318	1.0	6.9401	2.3938
3.2612	37.0	999	2.7155	1.0	6.7833	2.4118
3.217	38.0	1026	2.7033	1.0	6.6972	2.4964
3.1534	39.0	1053	2.6954	1.0	6.9671	2.4831
3.1534	40.0	1080	2.6767	1.0	7.3655	2.6130
3.1121	41.0	1107	2.6673	1.0	7.4479	2.7751
3.1121	42.0	1134	2.6608	1.0	8.0463	2.8065
3.0758	43.0	1161	2.6494	1.0	8.4453	2.8540
3.0758	44.0	1188	2.6424	1.0	7.9917	2.8893
3.0036	45.0	1215	2.6317	1.0	8.0109	2.9841
3.0036	46.0	1242	2.6262	1.0	5.5405	3.0465
2.9671	47.0	1269	2.6213	1.0	5.4395	3.1307
2.9671	48.0	1296	2.6121	1.0	5.8908	3.1183
2.9548	49.0	1323	2.6074	1.0	6.8068	3.1747
2.897	50.0	1350	2.5995	1.0	6.7684	3.2270

Framework versions

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