dstack tool is at the early preview stage. There are lots of things it doesn't support yet:
- Private repositories: Currently,
dstackonly supports public Git repositories. Private Git repositories cannot be used with
dstackyet. Also, make sure to use HTTPS protocol for authentication.
- The SSH protocol is not supported yet.
- Dashboard interface: Currently, you can work with
dstackonly via the CLI. There is not yet a user interface to sign up, and manage runners, runs, artifacts, and logs.
- Hardware metrics: Currently,
dstackdoesn't report hardware metrics for the runners, e.g. the use and availability of memory, CPU, GPU, etc.
- Hardware requirements: Currently, there is no way to specify hardware requirements per workflow (e.g. GPU, CPU, memory, etc.)
- Own storage: Currently,
dstackuses its own storage for artifacts and logs. There is not yet a way to use own storage.
- Own server: Currently, the
dstackserver is hosted with
dstack.aiand stores the information about the user, runs, and runners on its own server. There is not yet a way to have a
dstackserver up on your own machine.
- Triggers: Currently, there is no way to configure your own events or webhooks to trigger new runs. All runs must
be manually started via the
- Sweeps: Currently, there is no way to run a multiple combinations of variables in parallel (i.e. hyperparameter tuning, e.g. grid search, bayesian search, etc.)
Something critical is not on the list? Something didn't work or was unclear? Please, let me know. I'll look into it ASAP.