Vikhrmodels
/

VikhrT5-3b

Text2Text Generation Transformers Safetensors

Russian t5 Inference Endpoints text-generation-inference

Model card Files Files and versions Community

Edit model card

VikhrT5-3b: Улучшенная модель на базе FLAN T5 3b для русского языка

Cкорее всего она лучше чем FRED T5XL

Dataset	VikhrT5-3b	FRED-T5-XL(1.7b)	FLAN-t5-xl(3b)
ru_mmlu	0.32	0.252	0.28 (лол2)
xwinograd_ru	0.71 (lol)	0.57	0.52
xnli_ru	0.4280	0.34	0.33

Downloads last month: 1

Safetensors

Model size

2.96B params

Tensor type

F32

·

Dataset used to train Vikhrmodels/VikhrT5-3b

Space using Vikhrmodels/VikhrT5-3b 1