Re: Publish MXNet images to DockerHub
Yes, correct cu90 is indeed there, thanks for pointing it. So the question, should we be publishing to Docker Hub as part of the release process so that bindings other than python are also published and there is a policy on what cuda versions we publish? Thanks ANirudh On Sat, Jul 21, 2018 at 9:56 PM Mu Li wrote: > cu90 and cu90mkl are also available, see > https://hub.docker.com/r/mxnet/python/tags/ > > On Sat, Jul 21, 2018 at 9:51 PM, Anirudh Acharya > wrote: > > > The python binding that is actively maintained is > > > > mxnet-mkl 1.2.1 > > > > > > Other versions that use CUDA like mxnet-cu and mxnet-cumkl are > not > > actively maintained. > > > > > > > > - > > > > Anirudh > > > > > > > > > > > > On Sat, Jul 21, 2018 at 9:09 PM Mu Li wrote: > > > > > Surprisingly only the python binding is actively maintained. I remember > > we > > > can easily push all bindings into docker hub through the script in > > > https://github.com/apache/incubator-mxnet/tree/master/docker. > > > > > > On Sat, Jul 21, 2018 at 5:03 PM, Anirudh Acharya < > anirudhk...@gmail.com> > > > wrote: > > > > > > > Hi, > > > > > > > > Docker Hub( https://hub.docker.com/u/mxnet/ ) currently hosts images > > of > > > > MXNet and its various bindings but it is not actively maintained. > > Should > > > we > > > > publish MXNet images to Docker Hub as part of the release process and > > > > actively maintain it? > > > > > > > > The pros of publishing docker images would be ease of use and access > to > > > our > > > > users. Is this something that should be included as part of the > release > > > > process? What does the community think? > > > > > > > > Thanks > > > > Anirudh Acharya > > > > > > > > > >
Re: Publish MXNet images to DockerHub
cu90 and cu90mkl are also available, see https://hub.docker.com/r/mxnet/python/tags/ On Sat, Jul 21, 2018 at 9:51 PM, Anirudh Acharya wrote: > The python binding that is actively maintained is > > mxnet-mkl 1.2.1 > > > Other versions that use CUDA like mxnet-cu and mxnet-cumkl are not > actively maintained. > > > > - > > Anirudh > > > > > > On Sat, Jul 21, 2018 at 9:09 PM Mu Li wrote: > > > Surprisingly only the python binding is actively maintained. I remember > we > > can easily push all bindings into docker hub through the script in > > https://github.com/apache/incubator-mxnet/tree/master/docker. > > > > On Sat, Jul 21, 2018 at 5:03 PM, Anirudh Acharya > > wrote: > > > > > Hi, > > > > > > Docker Hub( https://hub.docker.com/u/mxnet/ ) currently hosts images > of > > > MXNet and its various bindings but it is not actively maintained. > Should > > we > > > publish MXNet images to Docker Hub as part of the release process and > > > actively maintain it? > > > > > > The pros of publishing docker images would be ease of use and access to > > our > > > users. Is this something that should be included as part of the release > > > process? What does the community think? > > > > > > Thanks > > > Anirudh Acharya > > > > > >
Re: Publish MXNet images to DockerHub
The python binding that is actively maintained is mxnet-mkl 1.2.1 Other versions that use CUDA like mxnet-cu and mxnet-cumkl are not actively maintained. - Anirudh On Sat, Jul 21, 2018 at 9:09 PM Mu Li wrote: > Surprisingly only the python binding is actively maintained. I remember we > can easily push all bindings into docker hub through the script in > https://github.com/apache/incubator-mxnet/tree/master/docker. > > On Sat, Jul 21, 2018 at 5:03 PM, Anirudh Acharya > wrote: > > > Hi, > > > > Docker Hub( https://hub.docker.com/u/mxnet/ ) currently hosts images of > > MXNet and its various bindings but it is not actively maintained. Should > we > > publish MXNet images to Docker Hub as part of the release process and > > actively maintain it? > > > > The pros of publishing docker images would be ease of use and access to > our > > users. Is this something that should be included as part of the release > > process? What does the community think? > > > > Thanks > > Anirudh Acharya > > >
Re: Publish MXNet images to DockerHub
Surprisingly only the python binding is actively maintained. I remember we can easily push all bindings into docker hub through the script in https://github.com/apache/incubator-mxnet/tree/master/docker. On Sat, Jul 21, 2018 at 5:03 PM, Anirudh Acharya wrote: > Hi, > > Docker Hub( https://hub.docker.com/u/mxnet/ ) currently hosts images of > MXNet and its various bindings but it is not actively maintained. Should we > publish MXNet images to Docker Hub as part of the release process and > actively maintain it? > > The pros of publishing docker images would be ease of use and access to our > users. Is this something that should be included as part of the release > process? What does the community think? > > Thanks > Anirudh Acharya >
Publish MXNet images to DockerHub
Hi, Docker Hub( https://hub.docker.com/u/mxnet/ ) currently hosts images of MXNet and its various bindings but it is not actively maintained. Should we publish MXNet images to Docker Hub as part of the release process and actively maintain it? The pros of publishing docker images would be ease of use and access to our users. Is this something that should be included as part of the release process? What does the community think? Thanks Anirudh Acharya
Re: How should MXNet treat nan values?
+1 Performance should not be affected by default. Generally, I would treat nan as undefined behaviour and allow to explicitly enable checks for nan which will throw exceptions. Asmus Am Samstag, 21. Juli 2018, 21:31:05 MESZ hat Junru Shao Folgendes geschrieben: However, I am not 100% sure how much performance will be sacrificed if we stick to NumPy's approach which seems to check numeral exceptions on each step. I believe it will be great if we could make the default setting to be "no checking", and leave users an option to turn on these numeral exception checkings. On 2018/07/20 22:19:46, Leonard Lausen wrote: > Hello MXNet community, > > It seems that there is currently no agreed upon principle to handle > `nan` values in operators. This has led to inconsistencies between > operators and also to inconsistency over releases. Some operators ignore > nan values (eg. argmax), others treated it as maximum (e.g. topk up to > mxnet v1.2) or just return “undefined” output (e.g. topk starting with > mxnet v1.3). > > Initially the change in topk was reported as a bug > (https://github.com/apache/incubator-mxnet/issues/8510) as some users > relied on the behavior. However (and rightfully) @asmushetzel, who > contributed the improved topk operator for mxnet v1.3 pointed out that > the change did not break any documented behavior. > > To go forward, please share your opinion how MXNet should handle `nan` > values. Should we continue to treat the behavior as undefined and > possibly silently changing between releases? Should we define a > reasonable standard (e.g. follow numpy) and treat operators that deviate > as buggy? Should we just document how operators behave currently and > warn if the behavior changes? Something else? > > Please make your opinion known so above issue can be resolved/closed and > general guidelines can be defined for future contributions, following > whatever consensus emerges. > > Thanks! > Leonard >
Re: How should MXNet treat nan values?
If you behave like numpy for sparse, then things like dividing any sparse matrix by another sparse matrix will produce a dense matrix with a lot of NaNs in it wherever it encountered a “missing” value in both the source and destination positions of the sparse matrices (ie 0 divided by 0). If i remember correctly, this was the process at first (creating a dense matrix with lots of NaNs, which was equivalent to the same calculation with dense matrices in numpy) but then was replaced by just assuming that you didn’t want those NaNs and what you really wanted was a sparse result. On Sat, Jul 21, 2018 at 12:31 PM Junru Shao wrote: > However, I am not 100% sure how much performance will be sacrificed if we > stick to NumPy's approach which seems to check numeral exceptions on each > step. > > I believe it will be great if we could make the default setting to be "no > checking", and leave users an option to turn on these numeral exception > checkings. > > On 2018/07/20 22:19:46, Leonard Lausen wrote: > > Hello MXNet community, > > > > It seems that there is currently no agreed upon principle to handle > > `nan` values in operators. This has led to inconsistencies between > > operators and also to inconsistency over releases. Some operators ignore > > nan values (eg. argmax), others treated it as maximum (e.g. topk up to > > mxnet v1.2) or just return “undefined” output (e.g. topk starting with > > mxnet v1.3). > > > > Initially the change in topk was reported as a bug > > (https://github.com/apache/incubator-mxnet/issues/8510) as some users > > relied on the behavior. However (and rightfully) @asmushetzel, who > > contributed the improved topk operator for mxnet v1.3 pointed out that > > the change did not break any documented behavior. > > > > To go forward, please share your opinion how MXNet should handle `nan` > > values. Should we continue to treat the behavior as undefined and > > possibly silently changing between releases? Should we define a > > reasonable standard (e.g. follow numpy) and treat operators that deviate > > as buggy? Should we just document how operators behave currently and > > warn if the behavior changes? Something else? > > > > Please make your opinion known so above issue can be resolved/closed and > > general guidelines can be defined for future contributions, following > > whatever consensus emerges. > > > > Thanks! > > Leonard > > >
Re: How should MXNet treat nan values?
However, I am not 100% sure how much performance will be sacrificed if we stick to NumPy's approach which seems to check numeral exceptions on each step. I believe it will be great if we could make the default setting to be "no checking", and leave users an option to turn on these numeral exception checkings. On 2018/07/20 22:19:46, Leonard Lausen wrote: > Hello MXNet community, > > It seems that there is currently no agreed upon principle to handle > `nan` values in operators. This has led to inconsistencies between > operators and also to inconsistency over releases. Some operators ignore > nan values (eg. argmax), others treated it as maximum (e.g. topk up to > mxnet v1.2) or just return “undefined” output (e.g. topk starting with > mxnet v1.3). > > Initially the change in topk was reported as a bug > (https://github.com/apache/incubator-mxnet/issues/8510) as some users > relied on the behavior. However (and rightfully) @asmushetzel, who > contributed the improved topk operator for mxnet v1.3 pointed out that > the change did not break any documented behavior. > > To go forward, please share your opinion how MXNet should handle `nan` > values. Should we continue to treat the behavior as undefined and > possibly silently changing between releases? Should we define a > reasonable standard (e.g. follow numpy) and treat operators that deviate > as buggy? Should we just document how operators behave currently and > warn if the behavior changes? Something else? > > Please make your opinion known so above issue can be resolved/closed and > general guidelines can be defined for future contributions, following > whatever consensus emerges. > > Thanks! > Leonard >
Re: How should MXNet treat nan values?
I think it is worth discussing. NunPy has defined its own rules to handle with numeral exceptions, which makes a lot of sense to me. [Link](https://docs.scipy.org/doc/numpy/user/misc.html#how-numpy-handles-numerical-exceptions) On 2018/07/20 22:19:46, Leonard Lausen wrote: > Hello MXNet community, > > It seems that there is currently no agreed upon principle to handle > `nan` values in operators. This has led to inconsistencies between > operators and also to inconsistency over releases. Some operators ignore > nan values (eg. argmax), others treated it as maximum (e.g. topk up to > mxnet v1.2) or just return “undefined” output (e.g. topk starting with > mxnet v1.3). > > Initially the change in topk was reported as a bug > (https://github.com/apache/incubator-mxnet/issues/8510) as some users > relied on the behavior. However (and rightfully) @asmushetzel, who > contributed the improved topk operator for mxnet v1.3 pointed out that > the change did not break any documented behavior. > > To go forward, please share your opinion how MXNet should handle `nan` > values. Should we continue to treat the behavior as undefined and > possibly silently changing between releases? Should we define a > reasonable standard (e.g. follow numpy) and treat operators that deviate > as buggy? Should we just document how operators behave currently and > warn if the behavior changes? Something else? > > Please make your opinion known so above issue can be resolved/closed and > general guidelines can be defined for future contributions, following > whatever consensus emerges. > > Thanks! > Leonard >