Zhanghao Wu

icon indicating an email [email protected]

icon company Sky Computing Lab, UC Berkeley icon location Berkeley, CA icon location Ph.D. student @ UC Berkeley Sky Computing Lab; Previously, RA @ MIT HAN Lab; Undergrad @ SJTU ACM Honors Class

Results 315 comments of Zhanghao Wu

Add ovhcloud

> Is there still a plan to add OVH or is this put on hold? Thanks for asking @kishb87! We are still very excited to support OVHCloud. The implementation of...

How to pack Skypilot jobs and clusters onto GPU nodes with Kubernetes?

An important case for this, when a k8s cluster has both CPU and GPU nodes. The jobs controller should be prioritized on CPU node first. Otherwise, the GPU node may...

[k8s] Unexpected error when relaunching an INIT cluster on k8s which failed due to capacity error

This should have been resolved by @DanielZhangQD's recent PRs: https://github.com/skypilot-org/skypilot/pull/7553, https://github.com/skypilot-org/skypilot/pull/7491

Depending on dev releases makes conda-forge release impossible

This is being fixed by #5375

[SERVE] Add the ability to update a service in place

Thanks for the suggestion @JGSweets! This is a optimization we are thinking for the serve update as well! We should definitely get to it soon. Any contribution would be really...

[SERVE] Add the ability to update a service in place

> Thanks @Michaelvll! > > Follow-up Q and likely a separate issue / request, but if we specify order can we suggest the min / max count for each resource...

[SERVE] Add the ability to update a service in place

Hi @JGSweets, we added the support for reusing existing replicas #2956, when only the `services` field is changed and no `workdir`, `file_mounts` are specified. Please check the hint in the...

Implement automated weekly smoke tests with GitHub Actions

Thanks for adding this @andylizf! There are several things to investigate for this, including, how we reduce the cost of the smoke test, how to reduce the flakiness of the...

Implement automated weekly smoke tests with GitHub Actions

Also, I just canceled all the smoke test actions, as that can take a long time and we don't have the credentials setup (should we?). I guess a better way...

Implement automated weekly smoke tests with GitHub Actions

Sounds good! Please be careful with the leaked resources in cloud and cost for running in CI. : )