System information

main

What is the problem that this feature solves?

Currently, Conv-11 only accepts input T : tensor(float16), tensor(float), tensor(double)

Alternatives considered

No response

Describe the feature

Just add bfloat16 to the supported dtype list.
Should we extend bfloat16 to all operators that only support float16 currently?

Will this influence the current api (Y/N)?

Y

Feature Area

operators

Are you willing to contribute it (Y/N)

None

Notes

No response