Zhanghao Wu

[Managed Spot] Features required in the managed spot

9

### Future TODOs: - [x] #799 - [ ] #800 - [x] #865 - [ ] Better loggings for the spot controller outputs - [x] Examples: Test real resnet training...

[Storage] Upload multiple files to the same bucket with different directories

2

It would be great if sky storage can support uploading multiple files into the same bucket but different folder in the bucket. Currently, a workaround is to write another program...

Authentication fails after deleting an existing sky-key

1

``` > sky launch -y -c test-huggingface-9ce1ce58-61 examples/huggingface_glue_imdb_app.yaml [?25hTraceback (most recent call last): File "/Users/zhwu/miniconda3/envs/sky-dev/bin/sky", line 33, in sys.exit(load_entry_point('skypilot', 'console_scripts', 'sky')()) File "/Users/zhwu/miniconda3/envs/sky-dev/lib/python3.8/site-packages/click/core.py", line 1128, in __call__ return self.main(*args, **kwargs)...

Autocompletion for the CLI using tab

1

It would be great if we can using `tab` to auto-complete the cluster name when typing the CLI commands. I tried to implement that in [cluster-auto-complete](https://github.com/sky-proj/sky/tree/cluster-auto-complete), but the auto-completion is...

Our `sky down` supports blobbing, like `sky down test-*`, but `sky storage delete` does not have that, making the bucket deleting in batch difficult.

In #978, we created programmatic APIs for basic sky commands, but the on-prem and benchmark-related functions are still in `cli.py`. We may want to move them into `core.py` as well.

This PR adds document for the programmatic API, waiting for #978 TODO: - [ ] Document for the programmatic API - [ ] Example for the API

Unknown TPUVM issue

2

An unknown TPUVM error occurs, during the smoke test: ``` > sky launch -y -c test-tpu-vm-9ce1ce58-69 examples/tpu/tpuvm_mnist.yaml [?25hTraceback (most recent call last): File "/Users/zhwu/miniconda3/envs/sky-dev/bin/sky", line 33, in sys.exit(load_entry_point('skypilot', 'console_scripts', 'sky')())...

Currently, our usage collection does not collect detailed resource information for `sky admin deploy`. We should design the process for the information collected during the on-prem setup by admin.

`sky launch` twice the following yaml will download the data again. This will cause unnecessary costs for downloading. Maybe we should consider using `aws s3 sync` instead of `aws s3...