Discovering faster matrix multiplication algorithms with reinforcement learning
1–10 of 116 posts
Re: Discovering faster matrix multiplication algorithms with reinforcement learning
#2this is cool. i suppose it's only a matter of time until we see optimizing compilers that use transformer-rl style searches for subsets of their optimization and codegen.
Re: Discovering faster matrix multiplication algorithms with reinforcement learning
#3this is cool. i suppose it's only a matter of time until we see optimizing compilers that use transformer-rl style searches for subsets of their optimization and codegen.
Will also generate some fun bugs!
Re: Discovering faster matrix multiplication algorithms with reinforcement learning
#4This is completely besides the matter, but reading "provably" in the abstract is a frank reminder of how terrible English spelling/pronounciation is. I can't imagine I'm the only well-read native English speaker who read this as "prov-ably" on first take. I don't know about most languages, but you just don't get nonsense like this in French, at least.
Re: Discovering faster matrix multiplication algorithms with reinforcement learning
#5Re: Discovering faster matrix multiplication algorithms with reinforcement learning
#6This is completely besides the matter, but reading "provably" in the abstract is a frank reminder of how terrible English spelling/pronounciation is. I can't imagine I'm the only well-read native English speaker who read this as "prov-ably" on first take. I don't know about most languages, but you just don't get nonsense like this in French, at least.
https://www.merriam-webster.com/dictionary/provably
Edit: Sorry, missed your point on first reading, which perhaps reinforces your point. I read papers like this for a living so the language is unambiguous but I see how it might not be for others, even if English is their first language.
Re: Discovering faster matrix multiplication algorithms with reinforcement learning
#7Re: Discovering faster matrix multiplication algorithms with reinforcement learning
#8This is completely besides the matter, but reading "provably" in the abstract is a frank reminder of how terrible English spelling/pronounciation is. I can't imagine I'm the only well-read native English speaker who read this as "prov-ably" on first take. I don't know about most languages, but you just don't get nonsense like this in French, at least.
I agree. This is completely besides the matter.
Re: Discovering faster matrix multiplication algorithms with reinforcement learning
#9Earlier quoted context omitted.
Will also generate some fun bugs!
Unless it provides or simply works within formal verification?
Formal verification does not prove lack of bugs. In best case, can only catch one certain type of bugs.
Re: Discovering faster matrix multiplication algorithms with reinforcement learning
#10this is cool. i suppose it's only a matter of time until we see optimizing compilers that use transformer-rl style searches for subsets of their optimization and codegen.
Generating pgo data with graph ML is already a thing