eea42195f3d59ca4da361c5aaaccc2d5

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

• Loss: 2.0907
• Data Size: 1.0
• Epoch Runtime: 119.0935
• Bleu: 4.9392

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	26.7429	0	10.4804	0.0025
No log	1	808	23.2439	0.0078	11.5344	0.0026
No log	2	1616	16.3456	0.0156	12.8278	0.0035
No log	3	2424	11.1294	0.0312	15.0454	0.0080
0.5325	4	3232	7.4730	0.0625	18.0596	0.0126
8.1758	5	4040	5.1044	0.125	24.7052	0.0089
4.8679	6	4848	3.4269	0.25	38.8175	0.4411
4.0404	7	5656	3.1126	0.5	65.5406	0.9482
3.6995	8.0	6464	2.8786	1.0	122.8830	1.4482
3.4488	9.0	7272	2.7572	1.0	120.1845	1.8188
3.3188	10.0	8080	2.6761	1.0	119.4552	2.0977
3.218	11.0	8888	2.6154	1.0	118.6943	2.3441
3.0844	12.0	9696	2.5652	1.0	117.9412	2.5308
3.055	13.0	10504	2.5266	1.0	118.7930	2.7109
2.9957	14.0	11312	2.4910	1.0	117.7511	2.8970
2.9387	15.0	12120	2.4616	1.0	117.9348	3.0056
2.909	16.0	12928	2.4300	1.0	120.4783	3.1653
2.8459	17.0	13736	2.4048	1.0	119.8377	3.3012
2.7946	18.0	14544	2.3831	1.0	119.1125	3.3432
2.7558	19.0	15352	2.3660	1.0	118.0165	3.4575
2.7101	20.0	16160	2.3451	1.0	118.5927	3.5280
2.6871	21.0	16968	2.3291	1.0	117.7284	3.6822
2.6403	22.0	17776	2.3141	1.0	118.3463	3.7477
2.6353	23.0	18584	2.2937	1.0	118.7333	3.8137
2.5808	24.0	19392	2.2826	1.0	119.5167	3.8677
2.5392	25.0	20200	2.2667	1.0	120.4301	3.9240
2.5302	26.0	21008	2.2532	1.0	120.6806	3.9855
2.5152	27.0	21816	2.2484	1.0	118.2256	4.0550
2.4817	28.0	22624	2.2330	1.0	119.0125	4.0892
2.4721	29.0	23432	2.2280	1.0	118.4721	4.1392
2.4041	30.0	24240	2.2160	1.0	118.0716	4.1792
2.4295	31.0	25048	2.2086	1.0	119.0087	4.2651
2.4075	32.0	25856	2.1984	1.0	120.8782	4.2717
2.4039	33.0	26664	2.1881	1.0	120.7347	4.3501
2.3536	34.0	27472	2.1791	1.0	120.3007	4.3648
2.3221	35.0	28280	2.1699	1.0	119.2118	4.4064
2.3602	36.0	29088	2.1696	1.0	119.3953	4.4432
2.3203	37.0	29896	2.1568	1.0	118.4084	4.4994
2.301	38.0	30704	2.1498	1.0	118.5280	4.5255
2.2393	39.0	31512	2.1470	1.0	118.5955	4.5601
2.265	40.0	32320	2.1366	1.0	118.7773	4.5875
2.2429	41.0	33128	2.1345	1.0	119.2913	4.6291
2.2311	42.0	33936	2.1315	1.0	121.1980	4.6798
2.1934	43.0	34744	2.1214	1.0	119.8991	4.7206
2.1981	44.0	35552	2.1188	1.0	120.9426	4.7160
2.1768	45.0	36360	2.1094	1.0	119.6998	4.7537
2.2005	46.0	37168	2.1056	1.0	119.4823	4.8064
2.1601	47.0	37976	2.0986	1.0	118.5197	4.8277
2.1152	48.0	38784	2.0929	1.0	118.7602	4.8407
2.1164	49.0	39592	2.0906	1.0	119.1980	4.9085
2.117	50.0	40400	2.0907	1.0	119.0935	4.9392

Framework versions

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