tdpeuter/2026SEL3-project-Brittle_Star_Controller_Modularity
tdpeuter/2026SEL3-project-Brittle_Star_Controller_Modularity
Archived
1
Fork 0
Code Issues Releases Activity

removed unnecessary comments

Browse source
This commit is contained in:
cedric 2026-03-27 04:40:13 +00:00
parent 31c2d188b7
commit a452912b36
1 changed files with 0 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

11
src/ppo.py
View file

@ -8,14 +8,6 @@ import jax.numpy as jnp
# Chose to use a class as it seemed the easiest way to integrate the CleanRL code style # Chose to use a class as it seemed the easiest way to integrate the CleanRL code style
# with our need to seperate concerns # with our need to seperate concerns
class PPO: class PPO:
"""
args: params for training
input_network: network that takes input state and returns hidden state
action_network: network that takes hidden state and returns action distribution
critic: value network used for ppo
message_passer: function that executes message passing and state agregation X times
"""
def __init__( def __init__(
self, args, input_network, action_network, critic, message_passer=None self, args, input_network, action_network, critic, message_passer=None
): ):
@ -96,9 +88,6 @@ Should be ok to use partial here, since the references to network,
actor and critic should not change at runtime actor and critic should not change at runtime
The cost of seperating concerns is to somehow pass these values The cost of seperating concerns is to somehow pass these values
that are now not in the same scope that are now not in the same scope
Chose to pass the actual apply functions, since they should never change
Other option was to pass the networks, but how does it influence compilation when their
class values would ever change? Better safe than sorry.
""" """