An open source AutoML toolkit for automate machine learning lifecycle, including feature engineering, neural architecture search, model compression and hyper-parameter tuning.
What would you like to be added:
documentation for environment variables used by nni.
Why is this needed:
I was trying to save pytorch model checkpoint while using nnictl experiment create. But in the trail script, I couldn't find a proper way to access path to current trial directory. I did some digging into the source code and found out some environment variables were saved in non-public api deep in the library.
Without this feature, how does current nni work：
User has to dig into nni source code and find where and what environment variable is set under the hood.
Components that may involve changes:
nni documentation and potentially nni api.
Brief description of your proposal if any:
add a documentation so that users know what environment variable is set. perhaps add additional helper function to allow easy access to values of environment variables.
What would you like to be added: documentation for environment variables used by nni.
Why is this needed: I was trying to save pytorch model checkpoint while using
nnictl experiment create. But in the trail script, I couldn't find a proper way to access path to current trial directory. I did some digging into the source code and found out some environment variables were saved in non-public api deep in the library.Without this feature, how does current nni work： User has to dig into nni source code and find where and what environment variable is set under the hood.
Components that may involve changes: nni documentation and potentially nni api.
Brief description of your proposal if any: add a documentation so that users know what environment variable is set. perhaps add additional helper function to allow easy access to values of environment variables.