wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-08-15 12:45:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

[RLlib] Issue 8319 DDPG (MA or num_envs_per_worker > 1) broken. (#8324)

Browse Source
This commit is contained in:
Sven Mika
2020-05-08 08:26:32 +02:00
committed by GitHub
parent 5f278c6411
commit d7eaacb5fe
5 changed files with 100 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rllib/models/tf/tf_action_dist.py
+10
View File
@@ -326,6 +326,11 @@ class SquashedGaussian(TFActionDistribution):
unsquashed = tf.math.atanh(save_normed_values)
return unsquashed
@staticmethod
@override(ActionDistribution)
def required_model_output_shape(action_space, model_config):
return np.prod(action_space.shape) * 2
class Beta(TFActionDistribution):
"""
@@ -371,6 +376,11 @@ class Beta(TFActionDistribution):
def _unsquash(self, values):
return (values - self.low) / (self.high - self.low)
@staticmethod
@override(ActionDistribution)
def required_model_output_shape(action_space, model_config):
return np.prod(action_space.shape) * 2
class Deterministic(TFActionDistribution):
"""Action distribution that returns the input values directly.