Closed PatriceVignola closed 1 year ago
TensorFlow supports int16 and uint16 as the output type of ArgMin and ArgMax, but DirectML only supports int32/int64/uint32/uint64. Therefore, we need to call DirectML with an int32 type and then cast back to int16 when the operation is done.
TensorFlow supports int16 and uint16 as the output type of ArgMin and ArgMax, but DirectML only supports int32/int64/uint32/uint64. Therefore, we need to call DirectML with an int32 type and then cast back to int16 when the operation is done.