ea5503f5eb0f44f08690df71a7b5d2e3

This model is a fine-tuned version of google-t5/t5-base on the Helsinki-NLP/opus_books [de-ru] dataset. It achieves the following results on the evaluation set:

• Loss: 0.9953
• Data Size: 1.0
• Epoch Runtime: 106.0105
• Bleu: 10.2895

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	2.7597	0	8.4820	0.4253
No log	1	434	2.6077	0.0078	10.2618	0.4821
No log	2	868	2.3576	0.0156	10.6719	0.5806
No log	3	1302	2.2020	0.0312	11.8069	0.7310
No log	4	1736	2.0808	0.0625	16.5864	1.0534
0.0918	5	2170	1.9764	0.125	21.5360	0.9362
2.078	6	2604	1.8608	0.25	32.7242	2.0158
1.92	7	3038	1.7406	0.5	52.7636	2.8017
1.7925	8.0	3472	1.5954	1.0	100.6842	3.5782
1.6637	9.0	3906	1.4976	1.0	99.4240	4.2416
1.5808	10.0	4340	1.4214	1.0	105.0762	4.9118
1.5098	11.0	4774	1.3634	1.0	102.8156	5.4506
1.4356	12.0	5208	1.3186	1.0	104.7841	5.8554
1.3905	13.0	5642	1.2750	1.0	102.3238	6.1622
1.3419	14.0	6076	1.2443	1.0	104.0530	6.5960
1.3127	15.0	6510	1.2162	1.0	103.2813	6.7289
1.2715	16.0	6944	1.1914	1.0	101.6225	7.0543
1.2421	17.0	7378	1.1681	1.0	101.3635	7.2670
1.2025	18.0	7812	1.1462	1.0	104.1814	7.4852
1.1593	19.0	8246	1.1349	1.0	102.9349	7.7047
1.1375	20.0	8680	1.1171	1.0	103.0859	7.9432
1.1153	21.0	9114	1.1020	1.0	104.5360	8.0993
1.081	22.0	9548	1.0891	1.0	106.6906	8.2592
1.0659	23.0	9982	1.0768	1.0	103.5095	8.4539
1.0454	24.0	10416	1.0716	1.0	104.2793	8.5494
1.032	25.0	10850	1.0601	1.0	106.2393	8.7721
0.9979	26.0	11284	1.0589	1.0	107.3540	8.8809
0.9874	27.0	11718	1.0484	1.0	107.2710	8.9978
0.9461	28.0	12152	1.0379	1.0	113.0742	9.1393
0.9435	29.0	12586	1.0340	1.0	114.6773	9.2591
0.9268	30.0	13020	1.0300	1.0	103.2993	9.2599
0.8959	31.0	13454	1.0233	1.0	103.4103	9.3891
0.8819	32.0	13888	1.0211	1.0	107.7374	9.4677
0.8915	33.0	14322	1.0094	1.0	104.0952	9.5236
0.8674	34.0	14756	1.0122	1.0	103.2641	9.6102
0.8526	35.0	15190	1.0142	1.0	102.7155	9.6117
0.8341	36.0	15624	1.0082	1.0	108.7346	9.7729
0.8221	37.0	16058	1.0088	1.0	104.8310	9.8156
0.8143	38.0	16492	1.0000	1.0	104.3322	9.7812
0.786	39.0	16926	1.0022	1.0	109.3286	9.9159
0.7788	40.0	17360	0.9976	1.0	101.4251	9.9975
0.7692	41.0	17794	0.9967	1.0	100.4028	9.9916
0.7575	42.0	18228	0.9919	1.0	105.7741	10.0278
0.7472	43.0	18662	0.9993	1.0	101.8167	10.0009
0.7334	44.0	19096	0.9982	1.0	102.0833	10.0496
0.7249	45.0	19530	1.0036	1.0	107.9295	10.2319
0.698	46.0	19964	0.9953	1.0	106.0105	10.2895

Framework versions

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