KVarN: Native vLLM KV-cache quantization back end by Huawei

throwa356262 an hour ago | parent | next [-]

Better performance than TQ and better quality than FP16?

Am I reading this right??

	▲	qeternity 8 minutes ago \| parent \| next [-]
		It's not better quality: 59.3% vs 59.4% fp16 on AIME 25
	▲	thefox96 10 minutes ago \| parent \| prev [-]
		Faster than Fp16, not better quality i guess

▲

v3ss0n an hour ago | parent | prev [-]

Why this is not a PR for vLLM ?

▲

esafak an hour ago | parent [-]

It's the output of a research paper; the authors are not trying to build up vLLM, and they probably have no incentive to do so. You can submit a PR, though! It's easier now while the divergence is low, so don't wait. Since there are six authors, I bet you could get help with the inevitable review chores if you just take the step of creating the PR.

edit: It might not be clear that it is based on vLLM 0.22, which is the current version: https://github.com/huawei-csl/KVarN/commit/d6290e99098d7426d.... All you have to do is create a diff off it; it's fairly straightforward.

	▲	jmalicki an hour ago \| parent [-]
		And with the help of AI, pointing at AI at this paper and saying "making a vLLM PR from this paper" tends to work surprisingly well, even if you need to nudge it a little bit along the way.