Simon Guo on X: "At the CUDA Mode IRL hackathon, @anneouyang and I tried this dream of ours as performance engineers: Can we get LLMs to go CUDA Mode 🧑💻 and write CUDA kernels? Code → ⚙️ LLM → Optimized ⚡ Code, that goes brrrrrr 💨 on GPUs? As much as I enjoyed learning and writing custom" / X
x.com · saved by 1 readers
At the CUDA Mode IRL hackathon, @anneouyang and I tried this dream of ours as performance engineers: Can we get LLMs to go CUDA Mode 🧑💻 and write CUDA kernels? Code → ⚙️ LLM → Optimized ⚡ Code, that goes brrrrrr 💨 on GPUs? As much as I enjoyed learning and writing custom kernels for LLMs @coh…