Reverse-engineering Nvidia's CUDA-checkpoint for faster cold starts

20 pointsposted a day ago
by ilreb

3 Comments

zoobab

a day ago

Would it be able to swap models on demand with this?

fergus00

a day ago

https://blog.doubleword.ai/fast-sglang-starts yeah, this is part of the idea. If you get cold starts down to seconds or subseconds, then you can run many models multiplexed on the same GPUs

zoobab

a day ago

I have read a Linkedin post in french recently with a university lab (one big machine serving hundreds of users) mentioning that they could swap models on demand. Will try to find back the link.