gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2025-12-17 11:10:20 +01:00
Code Issues Projects Releases Wiki Activity

Merge branch 'master' into imports

Browse Source
This commit is contained in:
Roman Dobosz
2018-04-24 07:43:04 +02:00
parent 1b095aeeca f31159aad6
commit 42a9ec132d
124 changed files with 10828 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
agents/dfp_agent.py
View File

@@ -34,7 +34,7 @@ class DFPAgent(agent.Agent):
# create the inputs for the network
input = current_states
input.append(np.repeat(np.expand_dims(self.current_goal, 0), self.tp.batch_size, 0))
input['goal'] = np.repeat(np.expand_dims(self.current_goal, 0), self.tp.batch_size, 0)
# get the current outputs of the network
targets = self.main_network.online_network.predict(input)
@@ -43,7 +43,7 @@ class DFPAgent(agent.Agent):
for i in range(self.tp.batch_size):
targets[i, actions[i]] = batch[i].info['future_measurements'].flatten()
result = self.main_network.train_and_sync_networks(current_states, targets)
result = self.main_network.train_and_sync_networks(input, targets)
total_loss = result[0]
return total_loss
@@ -55,7 +55,10 @@ class DFPAgent(agent.Agent):
goal = np.expand_dims(self.current_goal, 0)
# predict the future measurements
measurements_future_prediction = self.main_network.online_network.predict([observation, measurements, goal])[0]
measurements_future_prediction = self.main_network.online_network.predict({
"observation": observation,
"measurements": measurements,
"goal": goal})[0]
action_values = np.zeros((self.action_space_size,))
num_steps_used_for_objective = len(self.tp.agent.future_measurements_weights)