Hi there, I am following the example on this webpage to use Hydra with metaflow https://outerbounds.com/blog/metaflow-in-notebooks-and-scripts/#running-metaflow-in-scripts-with-custom-configurations . I was able to get the example to work, but it seems I have to define every config parameter with a

Parameter()

statement in the Flow file? Is there a different way to pass configs to the flows? Also, how to use more complicated configs than shown in this example? For example, how to use nested configs (e.g.,

parameter.arg1

,

parameter.arg2

), or those spread over multiple files?

Very timely - we have that feature in the works. Meanwhile, you can also use

JSONType

as a parameter type.

Not quite yet — it’s forthcoming

@dry-beach-38304 thanks! for now, i have to define every param in a single hydra config, then declare them all with

Parameter()

calls in the flow file, right?

what does your flow look like now?

i am following the example on the webpage: a single flat hydra config, and a corresponding

Parameter()

for every config parameter.

you should be able to just havea single paramete called

my_config

for example and pass the entire hydra dict as the value for that. The Parameter should be of type JSONType and that should work.

So, something like:

Parameter('my_config', default=None)

in the flow file?

(apologies in the middle of something so haven’t tried but something like:

Parameter("my_config", type=JSONType, default={})

and then in the runner thing, you can pass it using

run(my_config=**cfg)

or something like that. Basically my_config will contain all the values in a dictionary form.

thanks, will try it!

And you have a parameter in that flow? That’s weird, it should be injected as a valid name. Could you send me the snippet you are using and I can try to debug.

Thanks! I'll send you a DM with the MWE