How to set convolution math type in cuDNN 9.2.0?


Using the latest version of cuDNN (9.2.0), with graph API, and cudnn-frontend, how can one set the convolution math type to use CUDNN_TENSOR_OP_MATHor CUDNN_TENSOR_OP_MATH_ALLOW_CONVERSION?

Additionally, how can one also set the convolution math mode to CUDNN_CONVOLUTION or CUDNN_CROSS_CORRELATION, through cudnn-frontend?

If you can’t answer questions regarding cudnn-frontend, I would still appreciate it if you could tell me how to set the math type directly in the graph API. Thank you!