skypilot

skypilot copied to clipboard

Bug report from Daniel: `nvidia-smi` doesn't work on `sky gpunode --cloud gcp`. ``` NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA...

infwinston

This PR enables TPU Pod usage. To change from single TPU and TPU pod, user only needs to modify `accelerators: tpu-v2-8` to `accelerators: tpu-v2-32`. `sky launch` and `sky exec` will...

infwinston

This PR adds the `vCPUs` column in the optimizer message (which was requested by Justin a while ago). ``` $ sky gpunode --gpus V100 I 08-14 03:51:35 optimizer.py:605] == Optimizer...

WoosukKwon

 The problem may caused by clock misalign between the local server and the spot controller.

Michaelvll

Too many `ray job` relatted commands block newly arrived `ray job`

2

Although not much CPU and memory are used, the `sky-spot-controller` can still fail to take new `ray job` commands, due to a lot of `ray job` commands running and the...

Michaelvll

Closes #977. `GetPublicAccessBlock` is no longer a reliable method to check if a s3 bucket is public. Buckets can be publicly readable yet not allow access to `GetPublicAccessBlock`. I'm not...

romilbhardwaj

Fixes #1073. This PR makes sure that a normal task YAML can be run with `sky spot launch` without any modification. It is currently blocked by #1069. Tested: - [...

Michaelvll

Add SkyPilot Benchmark doc

1

WoosukKwon

TPU: investigate whether on-demand TPUs get killed occasionally

1

A TPU user mentioned that for TPUs, even with on-demand TPU, it will be killed at any time within every 2 days, and no logs can be found for the...

concretevitamin

[Onprem] Align Python Versions

1

After some discussion, the conclusion from last week to align user and admin's python (for submitting Ray jobs, since the Ray cluster is launched via admin's python) is to add...

michaelzhiluo