my PPO implementation for Cartpole, is code review allowed here?
$begingroup$
I implemented the clipped objective PPO-clip as explained here: https://spinningup.openai.com/en/latest/algorithms/ppo.html
Basically I used a dummy actor network to find the new action probability without training the local actor network.
"""use temp_actor to get new prob so we don't update the actual actor until
we do the clip op"""
curr_weights = self.actor.get_weights()
self.temp_actor.set_weights(curr_weights)
self.temp_actor.fit(state, advantages, epochs=1, verbose=0)
new_policy = self.temp_actor.predict(state, batch_size=1).flatten()
new_aprob = new_policy[action]
Then I worked out the ratio of probabilities and implemented the PPO clipping parts of the algorithm:
ratio = new_aprob / old_aprob
# scale = min(ratio * advantages, K.clip(ratio, 1 - self.epsilon, 1 + self.epsilon) * advantages)
no_clip = ratio * advantages
clipped = np.clip(ratio, 1 - self.epsilon, 1 + self.epsilon) * advantages
self.actor.fit(state, np.minimum(no_clip, clipped), epochs=1, verbose=0)
The full code is here (please excuse some coarse language in the comments): https://github.com/nyck33/openai_my_implements/blob/master/cartpole/my_ppo_cartpole.py
It seems to work but for Cartpole, slightly slower than my DQN and VPG implementations here: https://github.com/nyck33/openai_my_implements/tree/master/cartpole
The std deviation of returns seems a bit lower than with VPG or DQN so I'm guessing the clipping is stabilizing the learning somewhat. However, once I change the learning rates to anything else than what I currently have, it stops learning, ie. it's very brittle.
Thus, I'm looking for any advice here to make it more resilient. If inappropriate please let me know and I will delete this question promptly.
deep-learning reinforcement-learning openai-gym
asked 3 mins ago
mLstudent33mLstudent33
457
$endgroup$
add a comment |
$begingroup$
I implemented the clipped objective PPO-clip as explained here: https://spinningup.openai.com/en/latest/algorithms/ppo.html
Basically I used a dummy actor network to find the new action probability without training the local actor network.
"""use temp_actor to get new prob so we don't update the actual actor until
we do the clip op"""
curr_weights = self.actor.get_weights()
self.temp_actor.set_weights(curr_weights)
self.temp_actor.fit(state, advantages, epochs=1, verbose=0)
new_policy = self.temp_actor.predict(state, batch_size=1).flatten()
new_aprob = new_policy[action]
Then I worked out the ratio of probabilities and implemented the PPO clipping parts of the algorithm:
ratio = new_aprob / old_aprob
# scale = min(ratio * advantages, K.clip(ratio, 1 - self.epsilon, 1 + self.epsilon) * advantages)
no_clip = ratio * advantages
clipped = np.clip(ratio, 1 - self.epsilon, 1 + self.epsilon) * advantages
self.actor.fit(state, np.minimum(no_clip, clipped), epochs=1, verbose=0)
The full code is here (please excuse some coarse language in the comments): https://github.com/nyck33/openai_my_implements/blob/master/cartpole/my_ppo_cartpole.py
It seems to work but for Cartpole, slightly slower than my DQN and VPG implementations here: https://github.com/nyck33/openai_my_implements/tree/master/cartpole
The std deviation of returns seems a bit lower than with VPG or DQN so I'm guessing the clipping is stabilizing the learning somewhat. However, once I change the learning rates to anything else than what I currently have, it stops learning, ie. it's very brittle.
Thus, I'm looking for any advice here to make it more resilient. If inappropriate please let me know and I will delete this question promptly.
deep-learning reinforcement-learning openai-gym
asked 3 mins ago
mLstudent33mLstudent33
457
$endgroup$
add a comment |
$begingroup$
I implemented the clipped objective PPO-clip as explained here: https://spinningup.openai.com/en/latest/algorithms/ppo.html
Basically I used a dummy actor network to find the new action probability without training the local actor network.
"""use temp_actor to get new prob so we don't update the actual actor until
we do the clip op"""
curr_weights = self.actor.get_weights()
self.temp_actor.set_weights(curr_weights)
self.temp_actor.fit(state, advantages, epochs=1, verbose=0)
new_policy = self.temp_actor.predict(state, batch_size=1).flatten()
new_aprob = new_policy[action]
Then I worked out the ratio of probabilities and implemented the PPO clipping parts of the algorithm:
ratio = new_aprob / old_aprob
# scale = min(ratio * advantages, K.clip(ratio, 1 - self.epsilon, 1 + self.epsilon) * advantages)
no_clip = ratio * advantages
clipped = np.clip(ratio, 1 - self.epsilon, 1 + self.epsilon) * advantages
self.actor.fit(state, np.minimum(no_clip, clipped), epochs=1, verbose=0)
The full code is here (please excuse some coarse language in the comments): https://github.com/nyck33/openai_my_implements/blob/master/cartpole/my_ppo_cartpole.py
It seems to work but for Cartpole, slightly slower than my DQN and VPG implementations here: https://github.com/nyck33/openai_my_implements/tree/master/cartpole
The std deviation of returns seems a bit lower than with VPG or DQN so I'm guessing the clipping is stabilizing the learning somewhat. However, once I change the learning rates to anything else than what I currently have, it stops learning, ie. it's very brittle.
Thus, I'm looking for any advice here to make it more resilient. If inappropriate please let me know and I will delete this question promptly.
deep-learning reinforcement-learning openai-gym
asked 3 mins ago
mLstudent33mLstudent33
457
$endgroup$
I implemented the clipped objective PPO-clip as explained here: https://spinningup.openai.com/en/latest/algorithms/ppo.html
Basically I used a dummy actor network to find the new action probability without training the local actor network.
"""use temp_actor to get new prob so we don't update the actual actor until
we do the clip op"""
curr_weights = self.actor.get_weights()
self.temp_actor.set_weights(curr_weights)
self.temp_actor.fit(state, advantages, epochs=1, verbose=0)
new_policy = self.temp_actor.predict(state, batch_size=1).flatten()
new_aprob = new_policy[action]
Then I worked out the ratio of probabilities and implemented the PPO clipping parts of the algorithm:
ratio = new_aprob / old_aprob
# scale = min(ratio * advantages, K.clip(ratio, 1 - self.epsilon, 1 + self.epsilon) * advantages)
no_clip = ratio * advantages
clipped = np.clip(ratio, 1 - self.epsilon, 1 + self.epsilon) * advantages
self.actor.fit(state, np.minimum(no_clip, clipped), epochs=1, verbose=0)
The full code is here (please excuse some coarse language in the comments): https://github.com/nyck33/openai_my_implements/blob/master/cartpole/my_ppo_cartpole.py
It seems to work but for Cartpole, slightly slower than my DQN and VPG implementations here: https://github.com/nyck33/openai_my_implements/tree/master/cartpole
The std deviation of returns seems a bit lower than with VPG or DQN so I'm guessing the clipping is stabilizing the learning somewhat. However, once I change the learning rates to anything else than what I currently have, it stops learning, ie. it's very brittle.
Thus, I'm looking for any advice here to make it more resilient. If inappropriate please let me know and I will delete this question promptly.
deep-learning reinforcement-learning openai-gym
deep-learning reinforcement-learning openai-gym
asked 3 mins ago
mLstudent33mLstudent33
457
asked 3 mins ago
mLstudent33mLstudent33
457
asked 3 mins ago
mLstudent33mLstudent33
457
asked 3 mins ago
mLstudent33mLstudent33
457
asked 3 mins ago
mLstudent33mLstudent33
457
457
add a comment |
add a comment |
0
active
oldest
votes
Your Answer
StackExchange.ready(function() {
var channelOptions = {
tags: "".split(" "),
id: "557"
};
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function() {
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled) {
StackExchange.using("snippets", function() {
createEditor();
});
}
else {
createEditor();
}
});
function createEditor() {
StackExchange.prepareEditor({
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: false,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: null,
bindNavPrevention: true,
postfix: "",
imageUploader: {
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
},
onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
});
}
});
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fdatascience.stackexchange.com%2fquestions%2f49625%2fmy-ppo-implementation-for-cartpole-is-code-review-allowed-here%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
0
active
oldest
votes
0
active
oldest
votes
active
oldest
votes
active
oldest
votes
Thanks for contributing an answer to Data Science Stack Exchange!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
Use MathJax to format equations. MathJax reference.
To learn more, see our tips on writing great answers.
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fdatascience.stackexchange.com%2fquestions%2f49625%2fmy-ppo-implementation-for-cartpole-is-code-review-allowed-here%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
