Sometimes questions on how to use the fastai library go unanswered here. Sometimes even for days. How can we get support on a timely basis so we can consider using the fastai library in a professional setting? It’s a great library for learning, but without timely responses to support questions I don’t see how it can be considered for professional settings.
I smell a business opportunity…
Do libraries like TensorFlow offer paid support?
We try our best to answer questions we can as soon as possible. Lately with the new version of the library most focus is on that, but again we try to answer what we can you’ll have better luck getting feedback on the #fastai-users category than any of the Parts as those are actively monitored by Sylvain and Jeremy.
(By we I mean the senior users of this library, if not Jeremy/Sylvain)
@muellerzr thanks for letting me know how it works. For professional use I would need something more than that, so @machinethink does have a point that there might be a business opportunity here. As someone who is currently learning this stuff, I would certainly pay student adjusted pricing for support. And with a true commercial support offering I would have far less hesitation investing my time in learning this library over others. For sure it is frustrating to post a question in the “right” category with lots of details and just have it go completely ignored. If either of you have the time and interest, this is the question I am stuck on Custom ItemBase and ItemList needed or not?
I have no authority on anything related to fast.ai but in general open source projects are run by volunteers, and any support you get is also done by volunteers. You can ask questions, of course, but there is no guarantee anyone will actually answer them.
Paid support is a way to get such guarantees. It exists for many types of technology & tech products (in a way, I offer paid support for Core ML and iOS-related questions, since people hire me to consult when they run into Core ML-related issues, even though I have no official affiliation with the people who make Core ML).
Right now, no one seems to be offering paid support with SLAs etc for fast.ai.
This is why things like Red Hat and Canonical exist. We can’t run businesses on software where we only hope to get support if a volunteer feels like it. For sure it is a bigger challenge for smaller projects like this. But just as an example of what can be done, I think Paul Davis over at Ardour does a good job of raising money for his project, and as a result he and his team always reply very quickly to support requests. In that case the project is his only full time job. A different approach is projects like React and PyTorch which have the backing of a big company so they can hire enough people to provide a good level of free support. Which begs the question. I wonder if Jeremy and team would consider other funding sources and expanding the team. Or even individual sponsors.