Thank you to everyone who attended the first meeting of the Numerical/ML Working Group!
In our initial meeting we performed some brainstorming and established consensus of our initial focus towards expanding Swift's matrix capabilities. Additionally, we established consensus on communication channels (Slack and Swift Forums), project management tooling (GitHub), and a unified commitment towards making tangible progress.
Hope to see you on the next meeting @JaapWijnen. Notes are on the post previous to this and communication and collaborations channels links are open.
For reference, here is the link to tomorrow’s meeting at 9am PT https://meet.google.com/gci-boyn-dma
All further meetings will be coordinated solely through the previously posted slack channel.
Hi,
I am very much interested in collaborating and joining the Numerical/ML WG. I am thinking about working on a DL framework called swiftograd, inspired by tinygrad, after I saw that S4TF was archived. This is intended to be a very small DL library like tinygrad. I would love to see the community opinion on this and whether there should be a larger project for a DL library to replace S4TF.
Also @rex-remind could you please post an updated invitation link of the slack channel, as the one posted here expired.
Hi @KhalidAl-otaibi, unfortunately moment died in September and work stopped. However, the long term success of AI/ML on Swift is still very important to me so maybe we can get things going again.
Here is a permanent invite link Slack for anyone who sees this going forward.
For posterity, if things ever seem quiet in there feel free to tag me in a message. Volunteer work can ebb and flow a times.
To reply to your other question, we were working on a fresh Numerics/ShapedArray lib pulling pieces from S4TF that made sense without pulling in too much of the complexity, and building out from there. This may align with what you're trying to achieve. Let's discuss in slack.