Add passthrough of logger from policy to logger. Set policy in eval()…

… mode during evaluation phase

diploma_thesis/agents/utils/policy/action_policy.py

@@ -1,7 +1,8 @@
-from enum import StrEnum
+import logging
+import torch
 from typing import Dict
 
-from agents.utils import NeuralNetwork, Phase
+from agents.utils import NeuralNetwork
 from agents.utils.action import ActionSelector, from_cli as action_selector_from_cli
 from .policy import *
 
@@ -32,6 +33,15 @@ def __post_init__(self):
         if self.noise_parameters is not None:
             self.model.to_noisy(self.noise_parameters)
 
+    def with_logger(self, logger: logging.Logger):
+        super().with_logger(logger)
+
+        for module in [self.model, self.action_selector]:
+            if isinstance(module, Loggable):
+                module.with_logger(logger)
+
+        return self
+
     def configure(self, configuration: RunConfiguration):
         self.run_configuration = configuration
 
@@ -44,7 +54,7 @@ def configure(self, configuration: RunConfiguration):
             self.model = self.model.to(configuration.device)
 
     def update(self, phase: Phase):
-        self.phase = phase
+        super().update(phase)
 
         for module in [self.model, self.action_selector]:
             if isinstance(module, PhaseUpdatable):

diploma_thesis/agents/utils/policy/policy.py

@@ -5,16 +5,15 @@
 from enum import StrEnum
 from typing import TypeVar, Generic
 
-import torch
-
 from tensordict import TensorDict
 from tensordict.prototype import tensorclass
 from torch import nn
 
 from agents.base.state import State
-from agents.utils import PhaseUpdatable
+from agents.utils import PhaseUpdatable, Phase, EvaluationPhase
 from agents.utils.nn.layers.linear import Linear
 from agents.utils.run_configuration import RunConfiguration
+from utils import Loggable
 
 Action = TypeVar('Action')
 Rule = TypeVar('Rule')
@@ -37,13 +36,22 @@ class Keys(StrEnum):
     POLICY = 'policy'
 
 
-class Policy(Generic[Input], nn.Module, PhaseUpdatable, metaclass=ABCMeta):
+class Policy(nn.Module, Loggable, Generic[Input], PhaseUpdatable, metaclass=ABCMeta):
 
     def __init__(self):
         super().__init__()
 
         self.noise_parameters = None
 
+    def update(self, phase: Phase):
+        super().update(phase)
+
+        if isinstance(phase, EvaluationPhase):
+            self.eval()
+        else:
+            self.train()
+
+
     @abstractmethod
     def forward(self, state: State):
         """

diploma_thesis/agents/workcenter/model/deep_multi_rule.py

@@ -8,7 +8,7 @@
 from .rule import ALL_ROUTING_RULES, RoutingRule, IdleRoutingRule
 
 
-class DeepMultiRule(DeepPolicyWorkCenterModel, DiscreteAction):
+class DeepMultiRule(DeepPolicyWorkCenterModel):
 
     def __init__(self, rules: List[RoutingRule], policy: Policy[WorkCenterInput]):
         super().__init__(policy)