Skip to content

Bugfix policy gradinet reinforce tf2 #29

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
asokraju wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Bugfix policy gradinet reinforce tf2 #29

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
1b782d9
bugfix
asokraju Sep 10, 2020
27bcaa7
fixing a typo
asokraju Sep 10, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion policy_gradient_reinforce_tf2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,14 @@ def update_network(network, rewards, states, actions, num_actions):
discounted_rewards -= np.mean(discounted_rewards)
discounted_rewards /= np.std(discounted_rewards)
states = np.vstack(states)
loss = network.train_on_batch(states, discounted_rewards)
# the following commented next line for training is not working
# loss = network.train_on_batch(states, discounted_rewards)
# to fix this we make two changes,
# 1. one_hot_encode the actions
one_hot_encode = np.array([[1 if a==i else 0 for i in range(2)] for a in actions])
# 2. pass the discounted rewards using 'sample_weight' parameter of 'categorical_crossentropy' loss function
loss = network.train_on_batch(states,target_actions, sample_weight=discounted_rewards)

return loss


Expand Down