TensorFlow Contributor or Community forum

First of all, thank you for putting effort into the unification of information streams around TensorFlow! The website works fast and can be quite a benefit in the long-term for the community.

Currently, the header of the forum says: TensorFlow Contributor Forum. One can have a different definition of the word contributor. For me, it sounds like intended topics around how to contribute to TensorFlow or Addons; how to improve documentation, and how to support our highly valued SIGs!

I know that it’s only very early days, but currently, it feels like just Tensorflow Community forum. Both are pretty valuable, but I wonder if the forum should have a more clearly defined goal. Obviously, the community audience is orders of magnitude bigger than the contributor audience.

Please do not take it as a critique. I am just trying to see this from the eyes of the general public joining a bit later.

7 Likes

Awesome feedback - @Joana and @thea - thoughts? Any insight we can share on current name and willingness to just go with “TF Forum” ?

1 Like

We are specifically trying to focus on contributors for the moment due to some scope limitations on our end. And, somewhat relatedly, due to the fact that we need to limit sign-ups to active TF contributors who have signed the CLA and are already participating on our existing contributor mailing lists. These should both be resolved when we rebrand to the “TensorFlow Forum” in May. At that point, we will have all contributing-related discussion in one or more categories for that audience.

I think it’s great feedback that there doesn’t feel like there’s enough contributor-specific content in the forum at the moment. We’re hoping for our SIG Leads & MLGDEs to help us shape that content and structure. What do you think would be most helpful here @lc0?

1 Like

I think TF Forum is a better option in the long run and it was probably in the roadmap as well if I understood it correctly in yesterday’s meeting (note, I am in IST).

For contributor-specific content, a separate discussion topic can always be created and even further tags could be made to easily facilitate the related conversations.

4 Likes

PyTorch recently launched a separate forum for its contributors: https://dev-discuss.pytorch.org/ (vs https://discuss.pytorch.org which has been there for years)

Quote: They are used as long-form communication focused on technical discussions, as opposed to reporting issues (on github) or short-form chat (on slack).

A dedicated venue for TF contributors would also be valuable.

3 Likes

We could evaluate this with a traffic estimate of the two macro topics.

Also often it is hard to separate a general feature request/issue from a discussion topic on the forum.

This is a quite common topic and also if unmaintained now there was a Issue to Discourse plugin at Introducing GitHub Issues to Discourse - extras - Discourse Meta.

I think that it is a quite frequent use cause caused by the natural overlapping ambiguity of the tools. Also Github has this use case How to move existing issues into Discussions · Discussion #2952 · github/feedback · GitHub

Instead Slack, Discuss or Matrix.org/Element (after the Gitter fusion) is another topic related to the general sync vs async pro/cons evaluations (knowledge search, timezones etc.). Here wave another potential overlap of the roles here but I think it could be still useful to find a role to a sync communication tool.

1 Like

That is an open question. For now, I think it makes sense to have a central forum for everyone. We can assess what the needs are in the long term, for certain. It would certainly be easier to support one forum for starters, and I think providing a great forum experience is the highest priority.

Another open question is recommendations around tools to use for synchronous communication, as @Bhack pointed out. It might be worth discussing with the SIG Leads together once everyone’s had a chance to play around.

4 Likes

this is a very good example! thanks for bringing it @byronyi

I also agree with @Bhack that it’s hard to draw a line between applied use-cases and actual contributions.

I think the biggest part would be answering the question:

  • are we ready to maintain a general discussion forum, that would require more resources to maintain in a good state? Ex: answering questions, structuring for all use-cases, engaging with beginners, and so on
  • or are we focusing primarily on people contributing to the codebase.

From the explanation above, thanks @thea, it looks like the intention is to serve a broader audience and contributors are just a first cohort. If it’s the case, I would suggest we

  • change the name to TensorFlow Forum or TensorFlow Community Forum
  • ensure we have forum categories around contributions
2 Likes

This is a good point but I hope that we could bootstrap the process in the right way so as we scale it could reach a sort of “critical mass” for self-sustainability.

3 Likes

Yes, it would be a good problem to have! :blush:

1 Like

This community is amazing.
After watching Google I/O session on “What’s new in ML?”, I am just so confused. I thought I knew ML but now I feel like there is so much more to learn.
It’s overwhelming to learn about everything that tensor flow offers. I wish to learn that I hope I will get helped by this forum.

8 Likes