64d8f948cacf221429e77c320da2532d

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

• Loss: 2.4869
• Data Size: 1.0
• Epoch Runtime: 129.4884
• Bleu: 5.4676

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	17.5537	0	11.5478	0.2652
No log	1	808	16.9223	0.0078	13.6298	0.2909
No log	2	1616	16.1671	0.0156	13.5581	0.2858
No log	3	2424	12.2129	0.0312	15.7445	0.3541
0.4768	4	3232	8.5979	0.0625	19.4036	0.4252
9.1281	5	4040	5.8144	0.125	26.2148	0.4692
6.2506	6	4848	4.4376	0.25	40.5381	2.0356
4.9771	7	5656	3.8054	0.5	70.3143	1.4880
4.3807	8.0	6464	3.3574	1.0	128.5006	2.2071
4.0574	9.0	7272	3.2059	1.0	127.6823	2.5751
3.8897	10.0	8080	3.1111	1.0	129.5306	2.9060
3.7886	11.0	8888	3.0435	1.0	126.2402	3.1349
3.6421	12.0	9696	2.9903	1.0	126.9960	3.3167
3.5954	13.0	10504	2.9475	1.0	125.7064	3.4359
3.5397	14.0	11312	2.9082	1.0	130.3183	3.5929
3.4525	15.0	12120	2.8768	1.0	129.0098	3.6854
3.4442	16.0	12928	2.8474	1.0	127.6041	3.8392
3.3699	17.0	13736	2.8168	1.0	127.4958	3.9735
3.2891	18.0	14544	2.8023	1.0	126.1674	4.0733
3.2682	19.0	15352	2.7864	1.0	126.2859	4.1759
3.2188	20.0	16160	2.7522	1.0	127.2454	4.2918
3.1926	21.0	16968	2.7417	1.0	125.8983	4.3479
3.1308	22.0	17776	2.7235	1.0	127.5630	4.3960
3.1241	23.0	18584	2.7035	1.0	127.4127	4.4906
3.0632	24.0	19392	2.6946	1.0	129.8589	4.4850
3.0322	25.0	20200	2.6779	1.0	127.4462	4.5792
3.0082	26.0	21008	2.6611	1.0	127.6021	4.6459
2.9876	27.0	21816	2.6531	1.0	127.3374	4.7064
2.9531	28.0	22624	2.6383	1.0	126.5137	4.7342
2.9486	29.0	23432	2.6295	1.0	131.0503	4.8236
2.8676	30.0	24240	2.6218	1.0	129.3260	4.8436
2.892	31.0	25048	2.6124	1.0	131.0708	4.8687
2.8892	32.0	25856	2.6039	1.0	129.8470	4.9018
2.8521	33.0	26664	2.5910	1.0	130.3516	4.9640
2.8069	34.0	27472	2.5810	1.0	129.2665	4.9932
2.7613	35.0	28280	2.5755	1.0	127.8814	5.0271
2.8178	36.0	29088	2.5714	1.0	132.1693	5.0610
2.7811	37.0	29896	2.5606	1.0	128.5839	5.1224
2.7436	38.0	30704	2.5472	1.0	127.9664	5.1168
2.673	39.0	31512	2.5474	1.0	127.7580	5.1887
2.6948	40.0	32320	2.5331	1.0	127.9504	5.2169
2.6752	41.0	33128	2.5311	1.0	128.3656	5.2262
2.6663	42.0	33936	2.5276	1.0	132.9179	5.2630
2.6074	43.0	34744	2.5201	1.0	129.5735	5.2929
2.637	44.0	35552	2.5098	1.0	130.0661	5.3471
2.615	45.0	36360	2.5142	1.0	127.6709	5.3372
2.6158	46.0	37168	2.5082	1.0	128.2857	5.3916
2.5884	47.0	37976	2.4974	1.0	128.4521	5.4065
2.5419	48.0	38784	2.4940	1.0	128.1002	5.4234
2.5401	49.0	39592	2.4886	1.0	127.8107	5.4552
2.5403	50.0	40400	2.4869	1.0	129.4884	5.4676

Framework versions

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