Zhanghao Wu

Here is a proposal reverted in the PR #1400, for further discussion. Add `SKYPILOT_JOB_ID` to the `sky spot queue -a` to make it easier for the user to find. ```...

Did you turn on the `gradient_checkpointing`? We need to turn on that for the `train_mem.py` as that will turn off the `use_cache` implicitly.

> job Managed Job CLI (jobs with auto-recovery). > spot Alias for Managed Job CLI (default to spot instances). > Leaving this for brainstorming. Kind of feel like we should...

Thanks for the comments @concretevitamin! Fixed the optimizer table: ``` sky job launch -n pipeline tests/test_yamls/pipeline.yaml I 04-30 19:24:21 optimizer.py:816] Best plan: I 04-30 19:24:21 optimizer.py:821] ------------------------------------------------------------------------------------------------------------------ I 04-30 19:24:21...

> accurate, now that we've added more back compat handling? Good catch! Fixed the description. > This problem still exists? Can we use local storage handling just as what we...

> Another problem: > > * upgraded to [c78077a](https://github.com/skypilot-org/skypilot/commit/c78077a8c5d21d29cc8a2d03bcfc2e67820470af) > * ran some `job launch` > * after a while the controller autostopped > * now ran `status -r` >...

Tested: - [x] `pytest tests/test_smoke.py --managed-jobs` (47c1a83) with a controller started on master - [x] `pytest tests/test_smoke.py --managed-jobs` with a controller started on this PR - [x] `bash -i tests/backend_compatibility_tests.sh...

> Fixed a bug introduced in #3484. Merging now @cblmemo could you elaborate the bug a bit for future reference?

A minor point: for `sky show-gpus -a`, it would be nice to have the hint to be shown at the top instead of in the middle, since the latter is...

Seems we have missed one place below: https://github.com/JGSweets/skypilot/blob/22221f7ec7a4270c6308c256b10d2068c06a14be/sky/execution.py#L58-L59 It should be the following instead? ``` resources_utils.ClusterName(clone_disk_from, handle.cluster_name_on_cloud) ```