Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
1–10 of 33 posts
Re: Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
#2Obligatory: Either written by AI or by a human who has spent so much time with AI that they adopted its writing style. Anyways.
> Over 35 hours it performed 432 kernel evaluations. Each cycle meant writing code, compiling it, running it, reading the profiling output, deciding what to change, and trying again. The model diagnosed compilation failures it hadn’t seen before, identified performance bottlenecks through runtime feedback rather than prior knowledge, and redesigned the kernel architecture multiple times when incremental improvements stopped working.
Anyone remember genetic algorithms? This might be an improvement, but it still feels a little like deja vu.
Re: Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
#3At this point the models should just start improving themselves.
Re: Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
#4Obligatory: Either written by AI or by a human who has spent so much time with AI that they adopted its writing style. Anyways. > Over 35 hours it performed 432 kernel evaluations. Each cycle meant writing code, compiling it, running it, reading the profiling output, deciding what to change, and trying again. The model diagnosed compilation failures it hadn’t seen before, identified performance bottlenecks through ru…
Key word is non-differentiable optimization. That's what generic algorithms were traditionally good at.
Re: Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
#5Obligatory: Either written by AI or by a human who has spent so much time with AI that they adopted its writing style. Anyways. > Over 35 hours it performed 432 kernel evaluations. Each cycle meant writing code, compiling it, running it, reading the profiling output, deciding what to change, and trying again. The model diagnosed compilation failures it hadn’t seen before, identified performance bottlenecks through ru…
So a LLM wrote 432 kernel variations and it found what was the faster...
Re: Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
#6LLM written.
See the authors twitter, he speaks english at a rather basic level and certainly did not write this https://x.com/mohitgeryani/with_replies
Re: Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
#7I don't doubt that it did it but I wouldn't want to maintain whatever it ended up spewing after 35 hrs.
In my experience, AI fixes problems by mostly adding more code.
It's a short term gain for a long term hurt.
Re: Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
#8Obligatory: Either written by AI or by a human who has spent so much time with AI that they adopted its writing style. Anyways. > Over 35 hours it performed 432 kernel evaluations. Each cycle meant writing code, compiling it, running it, reading the profiling output, deciding what to change, and trying again. The model diagnosed compilation failures it hadn’t seen before, identified performance bottlenecks through ru…
Yeah, I remember. I still have Usenet postings about the genetic algorims conference back in the '90s and some magazine clippings about researcher from the University of Sussex where I first learned about genetic algorithms back in high school.
Re: Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
#9At this point the models should just start improving themselves.
Rumor is that anthropic writes all their code with Claude. So it kind of is.
Re: Qwen3.7-Max Ran for 35 Hours on Unknown Hardware and Achieved a 10× Speedup
#10LLM written. See the authors twitter, he speaks english at a rather basic level and certainly did not write this https://x.com/mohitgeryani/with_replies