szha commented on issue #20898: URL: https://github.com/apache/incubator-mxnet/issues/20898#issuecomment-1046412771
In general we want to conform to the array API standard. While fp16 is not in the array standard yet, I tried the same with ints and the same discrepancy exists. ``` In [17]: onp.array(-2, dtype=onp.int64) + onp.array([2, 0], dtype=onp.int16) Out[17]: array([ 0, -2], dtype=int16) In [18]: onp.array([-2], dtype=onp.int64) + onp.array([2, 0], dtype=onp.int16) Out[18]: array([ 0, -2]) In [19]: np.array(-2, dtype=onp.int64) + np.array([2, 0], dtype=onp.int16) Out[19]: array([ 0, -2], dtype=int64) In [20]: np.array([-2], dtype=onp.int64) + np.array([2, 0], dtype=onp.int16) Out[20]: array([ 0, -2], dtype=int64) ``` According to the [type promotion rules](https://data-apis.org/array-api/latest/API_specification/type_promotion.html) in array API standard, mxnet's current behavior is desired. > Type promotion rules must apply when determining the common result type for two array operands during an arithmetic operation, regardless of array dimension. Accordingly, zero-dimensional arrays must be subject to the same type promotion rules as dimensional arrays. @rgommers do you expect this to become a problem for API standard adoption on the numpy side? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
