vLLM Support

#1
by benchislett - opened

I am reviewing vLLM support for EAGLE3 for Kimi K2, and I was not able to get this running with Kimi K2.5 + https://github.com/vllm-project/vllm/pull/36361. I got very low (< 1.5 AL) acceptance rates.

https://github.com/torchspec-project/TorchSpec/issues/38
That may be the reason, but I am still in the process of deploying the vLLM environment. Waiting for confirmation.
Thank you for finding this bug.

I've uploaded a fixed version. You can give it a try.

lightseek changed discussion status to closed

Sign up or log in to comment