How to use taichi to implement sparse 2d convolutional network calculations to achieve inference acceleration?

[Caiyq2019]

In the PyTorch framework, it is challenging to achieve an improvement in inference speed with pruned 2D convolutional networks. However, can Taichi implement sparse 2D convolution operations to achieve inference acceleration?

[xixioba]

Is there any progress?