New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Improve notebook controller API #114

Merged

corentinlger merged 8 commits into main from corentin/improve_notebook_controller

Nov 26, 2024

Collaborator

corentinlger commented Nov 25, 2024 •

edited

Loading

Description

Improve notebook controller API on various points:

Only execute routines / behaviors of existing entities in the controller run
Add helper attributes to access existing / non-existing agents
Add a catch_error flag to either automatically catch errors for routines or not (could be renamed debug_mode or something similar). You can launch it like that with the run function of the controller:

controller.run(catch_errors=False) #default is set to True

Add locking mechanism to routine handler to prevent modifying the routines while they are called (e.g removing a routine during a routine step, resulting in an error)

While updating the new eating function, I got this (old is the old eating routine function, new the one I did, and Clem the one you did):

import timeit

from functools import partial

num_iterations = 1000

new = timeit.timeit(new, number=num_iterations)
old = timeit.timeit(old, number=num_iterations)
clem = timeit.timeit(clem, number=num_iterations)

print(f"Old took {old:.6f} seconds")
print(f"New took {new:.6f} seconds")
print(f"Clem took {clem:.6f} seconds")

Session 5:

Old took 0.806134 seconds
New took 1.754245 seconds
Clem took 1.760777 seconds

Session neuroevo_ABC:

Old took 0.043899 seconds
New took 0.124735 seconds
Clem took 0.116994 seconds

Seems like the old function works better, can you also try this on your computer ?

corentinlger added 8 commits

November 25, 2024 15:05


          Add helper attributes for accessing agents, clean comments

3079add


          Add error catching option to routines

c312595


          Add locking mechanism to routines

a03e0a0


          Update eating mechanism

1c798f9


          Separate eating and entities spawning mechanisms

3351e31


          Fix notebook controller details

a92079f


          Fix controller pause and agents sensed_entities problems

33ddcf2


          Update notebook sessions

491c98f

corentinlger merged commit 9741618 into main

2 checks passed

clement-moulin-frier reviewed

View reviewed changes

Collaborator

clement-moulin-frier left a comment

Thanks! See my comments in the code. Also:

could be renamed debug_mode or something similar: Yes
Seems like the old function works better, can you also try this on your computer ? : Can you send me the code to run?

vivarium/controllers/notebook_controller.py

@@ @@ -127,7 +126,9 @@ def __init__(self, config): @@
                       self.active_behaviors = {}
                       self.eating_range = 10
                       self.diet = []
+                      # TODO : see if we keep both

Collaborator

clement-moulin-frier Nov 27, 2024

No need It I think. Is has_eaten_since(time=1) equivalent to the previous has_eaten()?
Maybe also avoid the work "meal", I'd like to keep a low-level biological flavor in this library. Maybe time_since_last_consumption?

vivarium/controllers/notebook_controller.py

                                   # execute behaviors of agents
                                   if entity.etype == EntityType.AGENT:
                                       entity.behave(self.time)
+                                      # TODO: see if we do this in a dedicated function or not
+                                      # increment time since last meal for all alive agents
+                                      entity.time_since_last_meal += 1

Collaborator

clement-moulin-frier Nov 27, 2024

Shouldn't this be inside the routine instead?

Collaborator

clement-moulin-frier Nov 27, 2024

In general we should avoid specific mechanisms within the generic run loop, they should all be in routines

vivarium/controllers/notebook_controller.py

+                      :return: True if the agent has eaten since the given time, False otherwise
+                      """
+                      return self.time_since_last_meal < time

Collaborator

clement-moulin-frier Nov 27, 2024

Double check this will work as expected (init of the variable, comparison operator..)

vivarium/controllers/notebook_controller.py

                       # finally stop the simulation
-                      self.stop()
+                      self.pause()

Collaborator

clement-moulin-frier Nov 27, 2024

To discuss

vivarium/controllers/notebook_controller.py

                   def stop(self):
-                      """Stop the simulation
+                      """Stop the simulation and detach all routines

Collaborator

clement-moulin-frier Nov 27, 2024

Should also detach all behaviors?

vivarium/controllers/notebook_controller.py

-              def eating_routine_fn(controller):
-                  """make agents of the simulation eating the entities in their diet
+              def eating_routine_classic(controller):

Collaborator

clement-moulin-frier Nov 27, 2024

Could rename it eating_range_routine

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet