Sparsity support

[zhang662817]

Branch/Tag/Commit

main

Docker Image Version

pytorh

GPU name

A100

CUDA Driver

main

Reproduced Steps

In support matrix, only bert and encode support Sparsity (after Ampere);
Did you have plan to support other models?

If not, what's the main reason?
no perfencemance improvement or accuracy drop?

Thanks.