i want to define a graph/job with some ops that can run simultaneously (independent) and when they finish - a final op will start. I have something like this in my graph:

fanouts=[]

sublists=['list','with','locations','with','ever_changing','list_length']

for i in range(len(sublists)):

fanouts.append(getCritOp(locs=sublists[i])())

#finally, this op should execute after all the others are successful

combineCritdata(list=fanouts)()

Note the extra () on getCritOp and combineCritdata - these are op factories

Currently in this state , 3 ops will start, and the first two will finish, then the combineCritdata will start prematurely (i have max_concurrent = 3 configured) . I need it to require the other 3 (or however many it will be, i dont know) to finish first.

in order to define dependencies, you'll want to pass in the fanouts to the op invocation, rather than the op factory itself. So this would look something like

combineCritdata()(start_after=fanouts)

, assuming combineCritdata looks something like

def combineCritdata():
    @op(name="combineCritdata", ins={"start_after": In(Nothing)})
    def my_op():
        ...
    return my_op

This worked thank you so much!