[Release]: v0.7.0 Release Tracker

[simon-mo]

We plan a release after V1 is ready.

Planned release Dec 16-17.

We do not block release on feature request, minor issues. We do block release on major feature development, important user impacting issues, and compatibility fixes.

Add pending PRs directly here ⬇️

• [core] overhaul memory profiling and fix backward compatibility #10511
• [V1][VLM] Proper memory profiling for image language models #11210

[youkaichao]

• [torch.compile] fast inductor #11108 cc @tlrmchlsmth

[robertgshaw2-redhat]

• [Kernel]: Cutlass 2:4 Sparsity + FP8/Int8 Quant Support #10995

[Jeffwan]

• Support Lora prometheus metrics #11117

I will try to polish this change by end of tomorrow. If I can not make it on time, feel free to skip

[simon-mo]

@khluu and I will make a v0.6.5 release for now given we haven't released for 30 days. We will release v0.7.0 when V1 is ready.

[xfalcox]

@khluu and I will make a v0.6.5 release for now given we haven't released for 30 days. We will release v0.7.0 when V1 is ready.

That would be lovely. The bug fixed at #10511 is heavily impacting, and I'd love a new docker image to fix it.

[youkaichao]

@xfalcox the docker image is updated to 0.6.5 in https://hub.docker.com/r/vllm/vllm-openai/tags