clean up

Author: maryamhonari

ml-agents/mlagents/trainers/torch/action_model.py

@@ -9,9 +9,6 @@
 from mlagents.trainers.torch.agent_action import AgentAction
 from mlagents.trainers.torch.action_log_probs import ActionLogProbs
 from mlagents_envs.base_env import ActionSpec
-from mlagents_envs import logging_util
-
-logger = logging_util.get_logger(__name__)
 
 
 EPSILON = 1e-7  # Small value to avoid divide by zero
@@ -175,20 +172,15 @@ def get_action_out(self, inputs: torch.Tensor, masks: torch.Tensor) -> torch.Ten
                 action_out_deprecated = continuous_out
                 deter_continuous_out = torch.clamp(deter_continuous_out, -3, 3) / 3
         if self.action_spec.discrete_size > 0 and dists.discrete is not None:
-            logger.info(
-                f"dist: {[discrete_dist.probs for discrete_dist in dists.discrete]}"
-            )  # TODO: remove
             discrete_out_list = [
                 discrete_dist.exported_model_output()
                 for discrete_dist in dists.discrete
             ]
-            logger.info(f"discretelist {discrete_out_list}")  # TODO: remove
             discrete_out = torch.cat(discrete_out_list, dim=1)
             action_out_deprecated = torch.cat(discrete_out_list, dim=1)
             deter_discrete_out_list = [
                 discrete_dist.deterministic_sample() for discrete_dist in dists.discrete
             ]
-            logger.info(f"deterlist {deter_discrete_out_list}")  # TODO: remove
             deter_discrete_out = torch.cat(deter_discrete_out_list, dim=1)
 
         # deprecated action field does not support hybrid action

ml-agents/mlagents/trainers/torch/model_serialization.py

@@ -169,6 +169,5 @@ def export_policy_model(self, output_filepath: str) -> None:
                 input_names=self.input_names,
                 output_names=self.output_names,
                 dynamic_axes=self.dynamic_axes,
-                verbose=True,  # TODO: remove
             )
         logger.info(f"Exported {onnx_output_path}")

ml-agents/mlagents/trainers/torch/networks.py

@@ -571,9 +571,6 @@ def forward(
 
 class SimpleActor(nn.Module, Actor):
     MODEL_EXPORT_VERSION = 3  # Corresponds to ModelApiVersion.MLAgents2_0
-    is_stochastic_action_sampling = (
-        True
-    )  # TODO: this should be a user input both for training and inference
 
     def __init__(
         self,
@@ -585,7 +582,6 @@ def __init__(
     ):
         super().__init__()
         self.action_spec = action_spec
-        # self.is_continuous_int_deprecated = is_stochastic_action_sampling # TODO:
         self.version_number = torch.nn.Parameter(
             torch.Tensor([self.MODEL_EXPORT_VERSION]), requires_grad=False
         )