c6d6d68803a4055b8b9e4415dde29a9e

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

• Loss: 1.8844
• Data Size: 1.0
• Epoch Runtime: 65.2756
• Bleu: 6.5533

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	27.1197	0	5.9016	0.0027
No log	1	434	25.5910	0.0078	6.3391	0.0029
No log	2	868	21.1422	0.0156	6.9063	0.0034
No log	3	1302	16.4115	0.0312	8.0913	0.0123
No log	4	1736	11.6783	0.0625	9.9254	0.0063
0.7326	5	2170	6.4040	0.125	13.8378	0.0398
6.5943	6	2604	3.9997	0.25	21.2220	0.0502
4.3744	7	3038	3.1096	0.5	35.4124	1.0596
3.7822	8.0	3472	2.7642	1.0	65.5813	1.9793
3.4727	9.0	3906	2.6030	1.0	63.9496	2.5000
3.2909	10.0	4340	2.5019	1.0	65.7724	2.9023
3.1842	11.0	4774	2.4291	1.0	64.0860	3.1515
3.0285	12.0	5208	2.3774	1.0	66.4972	3.3280
2.9569	13.0	5642	2.3256	1.0	64.9085	3.5354
2.9002	14.0	6076	2.2886	1.0	65.4169	3.6909
2.841	15.0	6510	2.2529	1.0	63.9199	3.8546
2.7804	16.0	6944	2.2245	1.0	64.7418	3.9758
2.7425	17.0	7378	2.1928	1.0	63.5011	4.1727
2.687	18.0	7812	2.1735	1.0	64.6357	4.4216
2.6143	19.0	8246	2.1470	1.0	63.3229	4.5601
2.5837	20.0	8680	2.1307	1.0	64.0122	4.7101
2.551	21.0	9114	2.1148	1.0	64.3033	4.9674
2.4965	22.0	9548	2.0982	1.0	64.1991	5.0325
2.4792	23.0	9982	2.0796	1.0	63.8724	5.2364
2.4523	24.0	10416	2.0684	1.0	65.1237	5.2624
2.4142	25.0	10850	2.0532	1.0	68.0813	5.3982
2.3855	26.0	11284	2.0431	1.0	66.2577	5.4425
2.3694	27.0	11718	2.0285	1.0	65.2970	5.5802
2.2867	28.0	12152	2.0215	1.0	63.9735	5.6591
2.2935	29.0	12586	2.0119	1.0	65.4410	5.7569
2.2867	30.0	13020	2.0011	1.0	65.2921	5.7507
2.2363	31.0	13454	1.9920	1.0	67.4263	5.8234
2.2141	32.0	13888	1.9852	1.0	64.8216	5.8478
2.2254	33.0	14322	1.9747	1.0	65.3474	5.8864
2.1917	34.0	14756	1.9680	1.0	65.3715	5.9874
2.1711	35.0	15190	1.9606	1.0	63.8977	6.1146
2.1443	36.0	15624	1.9548	1.0	64.5038	6.1186
2.1275	37.0	16058	1.9464	1.0	64.8805	6.1374
2.1238	38.0	16492	1.9373	1.0	64.7629	6.1837
2.0652	39.0	16926	1.9360	1.0	64.2240	6.3315
2.0453	40.0	17360	1.9288	1.0	64.3513	6.3001
2.0422	41.0	17794	1.9267	1.0	64.4348	6.3532
2.0139	42.0	18228	1.9140	1.0	65.2384	6.3975
2.02	43.0	18662	1.9144	1.0	64.0676	6.4165
2.0035	44.0	19096	1.9110	1.0	66.5184	6.4202
1.9896	45.0	19530	1.9062	1.0	63.4230	6.4573
1.9573	46.0	19964	1.9038	1.0	64.7754	6.4374
1.9302	47.0	20398	1.9011	1.0	63.1561	6.5263
1.9437	48.0	20832	1.8903	1.0	65.2075	6.5562
1.9421	49.0	21266	1.8852	1.0	63.2385	6.6214
1.9439	50.0	21700	1.8844	1.0	65.2756	6.5533

Framework versions

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