Probably should make this into a backend of torch.compile | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		qihqi 5 months ago \| parent \| context \| favorite \| on: Compiling LLMs into a MegaKernel: A path to low-la... Probably should make this into a backend of torch.compile

zhihaojia 5 months ago [–]

Yes, it would be a lot of fun if MPK can enable torch.compile to generate megakernels. Torch-generated kernels are currently too slow for latency-sensitive workloads.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact