You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mxnet.apache.org by GitBox <gi...@apache.org> on 2021/06/22 18:27:45 UTC
[GitHub] [incubator-mxnet] fhieber opened a new issue #20373: Extended support for existing `mx.nd` ops in `mx.np(x)`.
fhieber opened a new issue #20373:
URL: https://github.com/apache/incubator-mxnet/issues/20373
## Description
We would like to see operators that currently exist in the mx.nd/mx.sym namespace for MXNet v1.x to also be ported to the new np/npx namespaces (MXNet 2.0). The following operators are required for migrating [Sockeye](https://github.com/awslabs/sockeye) to MXNet 2.0.
## Missing operators:
```
mx.nd.contrib.interleaved_matmul_*
mx.nd.contrib.intgemm_*
mx.nd.add_n
mx.nd.SequenceMask
mx.nd.slice_like
```
--
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org
[GitHub] [incubator-mxnet] fhieber commented on issue #20373: Extended support for existing `mx.nd` ops in `mx.np(x)`.
Posted by GitBox <gi...@apache.org>.
fhieber commented on issue #20373:
URL: https://github.com/apache/incubator-mxnet/issues/20373#issuecomment-867414441
Thanks @barry-jin, looking forward to #20375 be merged.
What about `mx.nd.random.multinomial()`? Is there an equivalent in the np(x) namespace?
--
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org
[GitHub] [incubator-mxnet] barry-jin commented on issue #20373: Extended support for existing `mx.nd` ops in `mx.np(x)`.
Posted by GitBox <gi...@apache.org>.
barry-jin commented on issue #20373:
URL: https://github.com/apache/incubator-mxnet/issues/20373#issuecomment-866444408
Hi @fhieber
For operators `mx.nd.contrib.interleaved_matmul_*`, `mx.nd.add_n` and `mx.nd.slice_like`, they are ported to npx namespace in PR #20375 as well as some test cases.
For `mx.nd.contrib.intgemm_*`, they have been ported to npx name space and you can find related test cases [here](https://github.com/apache/incubator-mxnet/blob/master/tests/python/unittest/test_contrib_intgemm.py).
For `mx.nd.SequenceMask`, it is already ported and you can find some implementations in [rnn_cell.py](https://github.com/apache/incubator-mxnet/blob/da4ff3a4dc0bd6a54af3d75c492021d18ba1867b/python/mxnet/gluon/rnn/rnn_cell.py#L86).
The replacement of `mx.nd.identity` in numpy mode is [`mx.np.copy`](https://github.com/apache/incubator-mxnet/blob/da4ff3a4dc0bd6a54af3d75c492021d18ba1867b/python/mxnet/numpy/multiarray.py#L1594)
--
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org
[GitHub] [incubator-mxnet] fhieber commented on issue #20373: Extended support for existing `mx.nd` ops in `mx.np(x)`.
Posted by GitBox <gi...@apache.org>.
fhieber commented on issue #20373:
URL: https://github.com/apache/incubator-mxnet/issues/20373#issuecomment-869120214
Hi @barry-jin,
I think `mx.np.random.multinomial` behaves differently. I now use `npx.random.categorical`, which does what we need (drawing 1 sample from multiple distributions (by row)).
I saw that #20375 has been merged. When do these changes show up in nightly builds at `https://dist.mxnet.io/python`? `mxnet-2.0.0b20210627` still doesn't seem to contain your changes (addition of slice_like).
--
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-unsubscribe@mxnet.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org
[GitHub] [incubator-mxnet] barry-jin commented on issue #20373: Extended support for existing `mx.nd` ops in `mx.np(x)`.
Posted by GitBox <gi...@apache.org>.
barry-jin commented on issue #20373:
URL: https://github.com/apache/incubator-mxnet/issues/20373#issuecomment-867842718
Yes, the equivalent is `mx.np.random.multinomial` and can be found [here](https://github.com/apache/incubator-mxnet/blob/7b4d61d11d1ee29b6000588f03fc074c13d2043e/python/mxnet/numpy/random.py#L386-L425).
--
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org
[GitHub] [incubator-mxnet] fhieber edited a comment on issue #20373: Extended support for existing `mx.nd` ops in `mx.np(x)`.
Posted by GitBox <gi...@apache.org>.
fhieber edited a comment on issue #20373:
URL: https://github.com/apache/incubator-mxnet/issues/20373#issuecomment-869120214
Hi @barry-jin,
I think `mx.np.random.multinomial` behaves differently. I now use `npx.random.categorical`, which does what we need (drawing 1 sample from multiple distributions (by row)).
--
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-unsubscribe@mxnet.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org