GCM auto assignment now returns a summary object

[bloebp]

The summary object contains information about the evaluated models and model choices. This object is printable to provide quick summary.

[bloebp]

My recommendation would be to use the global logger directly in functions where you need verbosity, and control the verbosity of the logger instead from the argument. Example that leverages a decorator to modify log level in a function where you need verbosity:

First, I will set the default log level to DEBUG.

@set_log_verbosity
def my_func(arg1, kwarg1=v1, verbosity=False):
     log.info('my verbose message')

Here the set_log_verbosity decorator function will modify the log level to INFO if verbosity is set to True, return the wrapped function, and reset the verbosity to DEBUG again.

The problem is, I want that these messages appear by default and not if a user sets a logger (which 99% of users probably won't). On the other hand, I also don't want to give it out as a print message, i.e., I set the loggign level to info and only add the messages if verbose is true.

Do you have another suggestions where a user does not need to put in additional work to get these insights?

[kailashbuki]

My recommendation would be to use the global logger directly in functions where you need verbosity, and control the verbosity of the logger instead from the argument. Example that leverages a decorator to modify log level in a function where you need verbosity: First, I will set the default log level to DEBUG.

@set_log_verbosity
def my_func(arg1, kwarg1=v1, verbosity=False):
     log.info('my verbose message')

Here the set_log_verbosity decorator function will modify the log level to INFO if verbosity is set to True, return the wrapped function, and reset the verbosity to DEBUG again.

The problem is, I want that these messages appear by default and not if a user sets a logger (which 99% of users probably won't). On the other hand, I also don't want to give it out as a print message, i.e., I set the loggign level to info and only add the messages if verbose is true.

Do you have another suggestions where a user does not need to put in additional work to get these insights?

This is what I mean roughly (there is of course a wiggle room for how to want to parse the verbosity arg):

def set_log_verbosity(func):
    def wrapper(*args, **kwargs):
        if kwargs.get('verbosity', None):
           # change log level to desired level
        res = func(*args, **kwargs)
        if kwargs.get('verbosity', None):
            # reset log level to default level
        return res
    return wrapper

[bloebp]

I revised the logic and the auto assignment now returns a summary object instead which the user can print. This is also helpful to have a more programmatic access to the evaluated models and their performances.