gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2025-12-17 19:20:19 +01:00
Code Issues Projects Releases Wiki Activity

Multiple improvements and bug fixes (#66)

Browse Source 
* Multiple improvements and bug fixes:

    * Using lazy stacking to save on memory when using a replay buffer
    * Remove step counting for evaluation episodes
    * Reset game between heatup and training
    * Major bug fixes in NEC (is reproducing the paper results for pong now)
    * Image input rescaling to 0-1 is now optional
    * Change the terminal title to be the experiment name
    * Observation cropping for atari is now optional
    * Added random number of noop actions for gym to match the dqn paper
    * Fixed a bug where the evaluation episodes won't start with the max possible ale lives
    * Added a script for plotting the results of an experiment over all the atari games
This commit is contained in:
Itai Caspi
2018-02-26 12:29:07 +02:00
committed by GitHub
parent 4fe9cba445
commit a7206ed702
20 changed files with 465 additions and 158 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
architectures/tensorflow_components/general_network.py
View File

@@ -56,7 +56,8 @@ class GeneralTensorFlowNetwork(TensorFlowArchitecture):
# the observation can be either an image or a vector
def get_observation_embedding(with_timestep=False):
if self.input_height > 1:
return ImageEmbedder((self.input_height, self.input_width, self.input_depth), name="observation")
return ImageEmbedder((self.input_height, self.input_width, self.input_depth), name="observation",
input_rescaler=self.tp.agent.input_rescaler)
else:
return VectorEmbedder((self.input_width + int(with_timestep), self.input_depth), name="observation")
@@ -191,7 +192,7 @@ class GeneralTensorFlowNetwork(TensorFlowArchitecture):
if tuning_parameters.agent.optimizer_type == 'Adam':
self.optimizer = tf.train.AdamOptimizer(learning_rate=tuning_parameters.learning_rate)
elif tuning_parameters.agent.optimizer_type == 'RMSProp':
self.optimizer = tf.train.RMSPropOptimizer(self.tp.learning_rate, decay=0.9, epsilon=0.01)
self.optimizer = tf.train.RMSPropOptimizer(tuning_parameters.learning_rate, decay=0.9, epsilon=0.01)
elif tuning_parameters.agent.optimizer_type == 'LBFGS':
self.optimizer = tf.contrib.opt.ScipyOptimizerInterface(self.total_loss, method='L-BFGS-B',
options={'maxiter': 25})