It's common to want to convert something Dask already does, and convert it to use high level graph under the hood.
First you need to find the place in the code where the dask task dictionary is created.
Typically this looks like a variable called dsk
or dsk_out
that is a dictionary mapping the key names to individual tasks.
Found it? Great, this is the spot we're going to insert an instance of your (new, not yet created) high level graph class, eg: dsk_out = MyNewLayer(input_args, ...)