[GitHub] [incubator-mxnet] leezu commented on pull request #18319: [numpy] symbolic advanced indexing

2020-07-15 Thread GitBox
leezu commented on pull request #18319: URL: https://github.com/apache/incubator-mxnet/pull/18319#issuecomment-659020475 @mxnet-bot run ci [unix-cpu] This is an automated message from the Apache Git Service. To respond to the

[GitHub] [incubator-mxnet] leezu commented on pull request #18319: [numpy] symbolic advanced indexing

2020-06-29 Thread GitBox
leezu commented on pull request #18319: URL: https://github.com/apache/incubator-mxnet/pull/18319#issuecomment-651328832 @sxjscience @AntiZpvoh thank you for investigating and clarifying the shortcomings. To ensure consistency between the imperative and symbolic interface, how about adding

[GitHub] [incubator-mxnet] leezu commented on pull request #18319: [numpy] symbolic advanced indexing

2020-06-10 Thread GitBox
leezu commented on pull request #18319: URL: https://github.com/apache/incubator-mxnet/pull/18319#issuecomment-642188409 @sxjscience We only need to add a special case for "all inputs are NDArrays of supported type" which calls the advanced indexing operator introduced by @AntiZpvoh. Featu

[GitHub] [incubator-mxnet] leezu commented on pull request #18319: [numpy] symbolic advanced indexing

2020-05-19 Thread GitBox
leezu commented on pull request #18319: URL: https://github.com/apache/incubator-mxnet/pull/18319#issuecomment-630909479 @AntiZpvoh on the operator side in the backend there is no distinction of "imperative" or "symbolic". All operators just take some array as input and compute some output