Many users reacted positively to Paige Bailey's call for AI kernel maintenance benchmarks beyond generation, expressing warm approval and excitement through emojis.
Based on 3 visible X reactions from 6 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@elliotarledge 🥹❤️ timeline cleanse https://x.com/DynamicWebPaige/status/2076443341965918282/photo/1
@a__name__a!! 🤩 https://x.com/DynamicWebPaige/status/2076448172348768486/photo/1
@DynamicWebPaige 😄 https://x.com/a__name__a/status/2076449740259041429/photo/1
kernelbench: can ai write a kernel from scratch literally nothing: can ai keep a kernel alive through a driver update, an arch bump, and three years of neglect 😅 genuine question: is there an eval for ai-driven kernel *maintenance*? not generation, but migration. like, cuda→hip, gfx942→gfx950, rocm n→n+1, stale autotuning configs migration has a built-in verifier (the old kernel is the spec), so it should be way more benchmarkable than greenfield kernel generation -- right? who's working on this
👀 Sharing a couple of @HuggingFace links for CASS-Bench and AgentKernelArena below - but it looks like kernel upgrades from one hardware target to the next version are ripe for an eval! https://x.com/DynamicWebPaige/status/2076470992277541026/photo/1 https://twitter.com/DynamicWebPaige/status/2076378032529715561
@elliotarledge 🥹❤️ timeline cleanse https://x.com/DynamicWebPaige/status/2076443341965918282/photo/1
@DynamicWebPaige 😄 https://x.com/a__name__a/status/2076449740259041429/photo/1
kernelbench: can ai write a kernel from scratch literally nothing: can ai keep a kernel alive through a driver update, an arch bump, and three years of neglect 😅 genuine question: is there an eval for ai-driven kernel *maintenance*? not generation, but migration. like, cuda→hip, gfx942→gfx950, rocm n→n+1, stale autotuning configs migration has a built-in verifier (the old kernel is the spec), so it should be way more benchmarkable than greenfield kernel generation -- right? who's working on this
👀 Sharing a couple of @HuggingFace links for CASS-Bench and AgentKernelArena below - but it looks like kernel upgrades from one hardware target to the next version are ripe for an eval! https://x.com/DynamicWebPaige/status/2076470992277541026/photo/1 https://twitter.com/DynamicWebPaige/status/2076378032529715561
@huggingface 🔗 CASS-Bench: https://huggingface.co/datasets/MBZUAI/cass 🔗 AgentKernelArena: https://github.com/AMD-AGI/AgentKernelArena
Many users reacted positively to Paige Bailey's call for AI kernel maintenance benchmarks beyond generation, expressing warm approval and excitement through emojis.
Based on 3 visible X reactions from 6 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@huggingface 🔗 CASS-Bench: https://huggingface.co/datasets/MBZUAI/cass 🔗 AgentKernelArena: https://github.com/AMD-AGI/AgentKernelArena