123 points by mlhacker 5 months ago flag hide 16 comments
john_doe 5 months ago next
Just came across this. WebAssembly for ML inference? Mind blown!
ml_enthusiast 5 months ago next
Heard about this approach before. It's an interesting technique. Waiting for more real-world applications.
greene_field 5 months ago next
Could you share some resources on this technique? I'd love to read up on it.
ric_k 5 months ago prev next
This can potentially democratize ML by enabling inference on low-end devices. Exciting!
ml_enthusiast 5 months ago next
@greene_field I think this paper covers it well: https://example.com/paper
greene_field 5 months ago next
Thanks! Will check it out.
coder007 5 months ago prev next
Has anyone tried to implement this on a real-world project yet? Would love to hear some insights.
ai_xpert 5 months ago next
@coder007 We are currently working on an implementation for a mobile application. So far, so good.
ps_fans 5 months ago next
That's interesting. Inference on mobile devices is a game changer. Keep us posted!
cutting_edge 5 months ago prev next
Definitely holds great potential. But wouldn't browser support be a challenge?
ml_enthusiast 5 months ago next
@cutting_edge The research is still in progress, but they seem to be managing it pretty well by optimizing performance.
coder007 5 months ago prev next
@cutting_edge they are currently experimenting with WebAssembly advancements, which could make this problem obsolete
robot_lover 5 months ago prev next
Curious if anyone knows whether WebGL is also considered for this technique? Or is WebAssembly a clear winner?
john_doe 5 months ago next
@robot_lover From what I've read, WebAssembly has lower latency and smaller binary size than WebGL in this context.
ml_enthusiast 5 months ago next
@robot_lover However, WebGL can be helpful in rendering visualizations when working with ML models. It could be useful to have both in the ecosystem.
learn_more 5 months ago prev next
I think this is a promising area. Should look into contributing to the research. Thanks for sharing!