Re: joining the science team to package spaCy & gensim
On Thu, 2020-08-27 at 19:55 +0800, Paul Wise wrote: > My employer is interested in having spaCy and gensim in Debian. > > https://spacy.io/ Due to the hostility of Explosion AI towards spaCy and thinc being available in the Debian archive, my employer has decided to stop working on this. We may revisit this decision at a later time. https://github.com/explosion/cython-blis/issues/32 https://www.debian.org/doc/manuals/developers-reference/developer-duties.en.html#coordination-with-upstream-developers I'll continue to work on getting gensim into Debian though. [Please CC me in reply, I'm not subscribed to the list] -- bye, pabs https://wiki.debian.org/PaulWise signature.asc Description: This is a digitally signed message part
Re: joining the science team to package spaCy & gensim
Hi, On Fri, Aug 28, 2020 at 07:57:22AM +, Mo Zhou wrote: > > I will do that, are there any guidelines for which team to use for > > specific packages or should I ask on the list(s) about it? > > No, there is no such guideline since debian science team reuses the > maintainer mail address (alioth) of debian science team. > > I tend to put things dedicated for machine/deep learning to the deep > learning team. General stuff to science team. > > > One other issue I encountered is outdated embedded code copies > > (specifically cython-blis has a copy of blis), I'd like to solve that > > upstream, do have a GitHub account and would you mind if I CCed you on > > any GitHub issues that I file about this? > > The BLIS API and ABI are relatively stable IIRC, so I guess > disentangling the embedded BLIS should be feasible. Apart from that, > just feel free to CC me https://github.com/cdluminate I'm fine with whatever repository is used (deeplearning[1] or science). In my understanding Debian Science was always a place where new teams could be evolve from (like we have seen in Debian Astro for instance). I could even imagine a Deep Learning Blend with an on set of metapackages. However, I would love some kind of consistency in the sense that those packages in Debian Science that qualify for deep learning topic should be moved to a single place where people expect them to be. The idea why I'm in favour of this idea is that we could attract more deep learning experts if we show that Debian really cares for this topic and a team has formed around this. Back to the topic: I'd be really happy if we could get spaCy in but my time is to limited to spent a lot of effort into it. Kind regards Andreas. [1] https://salsa.debian.org/deeplearning-team -- http://fam-tille.de
Re: joining the science team to package spaCy & gensim
Hi Paul, On Fri, Aug 28, 2020 at 10:35:41AM +0800, Paul Wise wrote: > > I think you can simply work on the existing repositories. New repos > > can be created under the deep learning team if you like. > > I will do that, are there any guidelines for which team to use for > specific packages or should I ask on the list(s) about it? No, there is no such guideline since debian science team reuses the maintainer mail address (alioth) of debian science team. I tend to put things dedicated for machine/deep learning to the deep learning team. General stuff to science team. > One other issue I encountered is outdated embedded code copies > (specifically cython-blis has a copy of blis), I'd like to solve that > upstream, do have a GitHub account and would you mind if I CCed you on > any GitHub issues that I file about this? The BLIS API and ABI are relatively stable IIRC, so I guess disentangling the embedded BLIS should be feasible. Apart from that, just feel free to CC me https://github.com/cdluminate
Re: joining the science team to package spaCy & gensim
On Fri, 2020-08-28 at 02:26 +, Mo Zhou wrote: > Both NLTK and spaCy suffer from a problem -- they cannot be fully > functional without pretrained models. And you know this is exactly > what the ML-Policy is discussing. Yes, I discovered this while doing the internal packaging. The default English models I looked at seemed to be quite dubious to me too, both from a legal perspective and a DFSG/ML-Policy perspective. > I think you can simply work on the existing repositories. New repos > can be created under the deep learning team if you like. I will do that, are there any guidelines for which team to use for specific packages or should I ask on the list(s) about it? One other issue I encountered is outdated embedded code copies (specifically cython-blis has a copy of blis), I'd like to solve that upstream, do have a GitHub account and would you mind if I CCed you on any GitHub issues that I file about this? > Just feel free to go ahead. But you might want to ask Andreas if he has > any unpushed commits. OK, hopefully he will see this mail :) > Debian science team has the maintainer access to Debian Deep Learning > team by default. OK, great. -- bye, pabs https://wiki.debian.org/PaulWise signature.asc Description: This is a digitally signed message part
Re: joining the science team to package spaCy & gensim
Hi Paul, In the past Andreas Tille and I tried to get spaCy into the archive since it is really useful (in my research projects), and is sometimes more convenient than NLTK (I'm the uploader). Both NLTK and spaCy suffer from a problem -- they cannot be fully functional without pretrained models. And you know this is exactly what the ML-Policy is discussing. I think you can simply work on the existing repositories. New repos can be created under the deep learning team if you like. On Thu, Aug 27, 2020 at 07:55:09PM +0800, Paul Wise wrote: > Hi all, > > My employer is interested in having spaCy and gensim in Debian. > > https://spacy.io/ > https://radimrehurek.com/gensim/ > > I noticed that there is a spaCy package in the team's repository > although it is not yet in Debian and gensim is also a natural language > processing tool so the team seems like the right place for it too. > > https://salsa.debian.org/science-team/spacy > > I have used stdeb to create internal packages of spacy, gensim and > their missing dependencies. The packages all build, some tests fail and > the packaging needs cleanup and fixes. I would like to import the > packages into the team and work on completing them. Some of the > dependencies are probably more suitable for the general Python team or > possibly the machine learning team, so I'll import those elsewhere. Just feel free to go ahead. But you might want to ask Andreas if he has any unpushed commits. > I've submitted my request to join the salsa project. Debian science team has the maintainer access to Debian Deep Learning team by default. > [Please CC me in reply, I'm not subscribed to the list] > > -- > bye, > pabs > > https://wiki.debian.org/PaulWise
Re: joining the science team to package spaCy & gensim
Hello Paul! I have approved your request. Welcome on board! Please contact the team if you need an assistance or help. Best regards, Anton Paul Wise schrieb am Do., 27. Aug. 2020, 13:55: > Hi all, > > My employer is interested in having spaCy and gensim in Debian. > > https://spacy.io/ > https://radimrehurek.com/gensim/ > > I noticed that there is a spaCy package in the team's repository > although it is not yet in Debian and gensim is also a natural language > processing tool so the team seems like the right place for it too. > > https://salsa.debian.org/science-team/spacy > > I have used stdeb to create internal packages of spacy, gensim and > their missing dependencies. The packages all build, some tests fail and > the packaging needs cleanup and fixes. I would like to import the > packages into the team and work on completing them. Some of the > dependencies are probably more suitable for the general Python team or > possibly the machine learning team, so I'll import those elsewhere. > > I've submitted my request to join the salsa project. > > [Please CC me in reply, I'm not subscribed to the list] > > -- > bye, > pabs > > https://wiki.debian.org/PaulWise >
joining the science team to package spaCy & gensim
Hi all, My employer is interested in having spaCy and gensim in Debian. https://spacy.io/ https://radimrehurek.com/gensim/ I noticed that there is a spaCy package in the team's repository although it is not yet in Debian and gensim is also a natural language processing tool so the team seems like the right place for it too. https://salsa.debian.org/science-team/spacy I have used stdeb to create internal packages of spacy, gensim and their missing dependencies. The packages all build, some tests fail and the packaging needs cleanup and fixes. I would like to import the packages into the team and work on completing them. Some of the dependencies are probably more suitable for the general Python team or possibly the machine learning team, so I'll import those elsewhere. I've submitted my request to join the salsa project. [Please CC me in reply, I'm not subscribed to the list] -- bye, pabs https://wiki.debian.org/PaulWise signature.asc Description: This is a digitally signed message part
