ilo-toki-rut5-base

This model is a fine-tuned version of ai-forever/ruT5-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7742
• Bleu: 31.1318

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
0.8262	1.0	31041	1.1476	20.3554
1.2097	2.0	62082	1.0090	24.1511
0.3034	3.0	93123	0.9302	26.2398
0.7045	4.0	124164	0.8848	27.3941
0.2932	5.0	155205	0.8512	28.3013
0.2401	6.0	186246	0.8291	28.9026
0.7421	7.0	217287	0.8129	29.5286
0.3411	8.0	248328	0.7978	29.9663
0.516	9.0	279369	0.7910	30.2774
0.6968	10.0	310410	0.7847	30.6053
0.2133	11.0	341451	0.7792	30.8103
0.4	12.0	372492	0.7770	30.9236
0.5006	13.0	403533	0.7755	31.0467
0.3673	14.0	434574	0.7742	31.1312
0.4823	15.0	465615	0.7752	31.1158

Framework versions

• Transformers 4.56.2
• Pytorch 2.8.0+cu129
• Datasets 3.6.0
• Tokenizers 0.22.0