Ray AWS Cluster Config Generator #47
Conversation
|
Job PR-47-96d5217 is done. |
| def _update_config( | ||
| self, | ||
| instance_type: Optional[str] = None, | ||
| instance_count: Optional[str] = None, |
There was a problem hiding this comment.
Why not set it to 1 as a default and change type to int in the API?
instance_count: int = 1 ?
There was a problem hiding this comment.
It should be int, that's a mistake. Though if we specify a default value for instance_count, we will always update the config to have instance_count = 1 even when user left it blank. I'll stick with default = None, meaning user doesn't want to update it
src/autogluon/cloud/cluster/ray_aws_cluster_config_generator.py
Outdated
Show resolved
Hide resolved
src/autogluon/cloud/cluster/ray_aws_cluster_config_generator.py
Outdated
Show resolved
Hide resolved
| def _set_provider(self): | ||
| """Set provider to be default ones if user didn't provide any""" | ||
| default_config = self._default_config | ||
| provider: Dict[str, Any] = self.config.get(PROVIDER, None) | ||
| if provider is None: | ||
| provider = default_config[PROVIDER] | ||
| self.config.update({PROVIDER: provider}) |
There was a problem hiding this comment.
Can we templatize repeating methods: I see lots of repeating code which can be reuse between the _update_xxx methods.
There was a problem hiding this comment.
It is hard to templatize these methods because they are essentially trying to update keys and values lying in different layers of the yaml file. The method looks similar, but because the structure of the yaml file differs for each key, it's actually specific
src/autogluon/cloud/cluster/ray_aws_cluster_config_generator.py
Outdated
Show resolved
Hide resolved
|
Job PR-47-a74b280 is done. |
Issue #, if available:
Description of changes:
RayAWSClusterConfigGenerator. It has a default config and accepts user config override. Also allow users to update the config with frequent used parameters that are available infitcall, i.e.instance_typeBy submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.