ChanServ changed the topic of #mlpack to: "mlpack: a fast, flexible machine learning library :: We don't always respond instantly, but we will respond; please be patient :: Logs at http://www.mlpack.org/irc/
xiaohong has joined #mlpack
xiaohong has quit [Remote host closed the connection]
xiaohong has joined #mlpack
xiaohong has quit [Remote host closed the connection]
xiaohong_ has joined #mlpack
xiaohong_ has quit [Remote host closed the connection]
xiaohong has joined #mlpack
< xiaohong> Hi, I have a question about how to create an NN model. Why the output was 'matrix size: 0x1' when I print the NN model parameters?
< xiaohong> Did I missing something to create the model?
< xiaohong> In the PPO unit test, I create model using the same the way with QlearningTest.
< jenkins-mlpack2> Project docker mlpack nightly build build #399: STILL UNSTABLE in 3 hr 53 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/399/
< xiaohong> I find that after I call the ResetParameters() function, then `Parameters()` still return the matrix size with [0X1].
< xiaohong> How can I create the model correctly?
xiaohong has quit [Remote host closed the connection]
xiaohong has joined #mlpack
xiaohong has quit [Remote host closed the connection]
xiaohong has joined #mlpack
ImQ009 has joined #mlpack
xiaohong2 has joined #mlpack
KimSangYeon-DGU has joined #mlpack
xiaohong2 has quit [Ping timeout: 276 seconds]
< zoq> xiaohong: After you copied the model? Printed Paramters() from https://github.com/mlpack/mlpack/blob/86c9d452b9a21c837d25e7d0f791ccbd5a7cd30c/src/mlpack/tests/q_learning_test.cpp#L50 after ResetParameters(), and the output looks good.
xiaohong has quit [Remote host closed the connection]
xiaohong has joined #mlpack
< xiaohong> zoq: Yes, I print the model in the q_learning_test, the output looks good.
< xiaohong> In the QLearningTest, before calling the ResetParameters(), the output of model.Parameters() is [matrix size: 0x0]. after ResetParameters(), and the output was not the same with PPOTest.
< zoq> xiaohong: If you call ResetParameters() right after you passed the model to the constructor, does that solve the problem?
< xiaohong> Do you mean just call ResetParameters() without any condition check?
< zoq> xiaohong: Just to see if that works, if it does, we could check if the size of the parameters is >= 1.
favre49 has joined #mlpack
KimSangYeon-DGU has quit [Remote host closed the connection]
< xiaohong> Okay.
< xiaohong> I recompile the project, so maybe it need a little time to complete the compile process.
< zoq> sure, no worries
< favre49> zoq: I'll be going back to college tomorrow, so I'm not sure how much work I'll be able to complete over the next couple days
< zoq> favre49: Okay, I will continue with the review of the NEAT PR, so whenever you have time, let's discuss/update the comments.
< favre49> Sounds great :) Hopefully my train doesn't get cancelled, i hear it's flooding in the area
< zoq> favre49: Ohh, fingers crossed.
favre49 has quit [Remote host closed the connection]
< xiaohong> zoq: During the debugging time, I output the Parameters() function before call the ResetParameters(), the result was [matrix size: 0x0].
< zoq> xiaohong: Right, ResetParameters() is called once Training begins or if you call the Forward or Predict function
< xiaohong> So, I think in the PPO constructor, the actorNetwork.ResetParameters() was called.
< zoq> xiaohong: I think in your case it's not empty since it's [0,1]?
< xiaohong> zoq: Oh, you are right. Before I call the constructor, I output the model's parameters, the result was correct.
< zoq> xiaohong: It should be empty, not sure yet why it isn't; but for now I think we could check for >= 1?
< xiaohong> So in the PPO constructor, the behavior was not correct.
< xiaohong> The code was look like this, https://paste.ubuntu.com/p/H2ZCYrVSmY/
< xiaohong> The output of the code, https://paste.ubuntu.com/p/wzQRVPt7JH/
< zoq> xiaohong: Right, ideally we could do this in the constructor, which we do if it is empty, I guess for some reason it isn't empty
< xiaohong> After create the model, the model is empty, so the output was [matrix size: 0x0], right?
< zoq> right
< xiaohong> Yes, before call the constructor, the behavior is correct. I don't know what happened in the constructor.
< zoq> xiaohong: can you check the output of actorNetwork.Parameters().n_elem instead of is_empty()?
< xiaohong> The output is 0.
< zoq> xiaohong: Does it work as well if we switch is_empty with n_elem?
< xiaohong> It's really wired. lol
< xiaohong> The output of is_empty is 1.
< xiaohong> zoq: I find the problem. lol
< zoq> xiaohong: Oh nice
< xiaohong> It seems that the parameter and argument can not be the same name. = =
< zoq> if you remove the second move it works I guess?
< xiaohong> The second?
< xiaohong> Do you mean
< xiaohong> when call the constructor, I call the std::move.
< zoq> right
< xiaohong> In the constructor implementation, I call the std::move again?
< zoq> yeah once you call the constructor with move and inside the constructor you use move again
< xiaohong> Okay, I see. Because I didn't realize that the name was not the same in the Qlearning constructor.
< xiaohong> I think is not good to use the std::move statement twice, right?
< zoq> Shouldn't be an issue.
< xiaohong> Even though the argument and parameter use different name.
< zoq> Right, I thought you meant the move.
< xiaohong> zoq: Thanks, I think I can move on now.
< zoq> xiaohong: You figured this out, so thanks to you.
< xiaohong> After I remove the second move, it still have warning.
< xiaohong> ppo_impl.hpp:46:16: warning: field 'actorNetwork' is uninitialized when used here
< zoq> okay, let#s keep the move.
< xiaohong> zoq: Great, have a good weekend.
< zoq> xiaohong: Thanks, you too.
xiaohong has quit [Remote host closed the connection]
xiaohong has joined #mlpack
xiaohong has quit [Remote host closed the connection]
xiaohong has joined #mlpack
xiaohong has quit [Remote host closed the connection]
travis-ci has joined #mlpack
< travis-ci> robertohueso/mlpack#51 (mc_kde_error_bounds - f35b22d : Roberto Hueso Gomez): The build was broken.
travis-ci has left #mlpack []
< sakshamB> ShikharJ: sorry for the late response. Actually there is no wifi at my hostel right now and I am not getting even 3G connectivity.
< sakshamB> ShikharJ: I will try to update all the PRs soon and open a PR for CGAN.
xiaohong has joined #mlpack
xiaohong has quit [Remote host closed the connection]
xiaohong has joined #mlpack
xiaohong has quit [Remote host closed the connection]
vivekp has joined #mlpack
KimSangYeon-DGU has joined #mlpack
< ShikharJ> sakshamB: Okay no worries. Have a good weekend.
< zoq> Switched the jenkins style check plugin, to resolve the display issues we had with the cppcheck plugin.
< zoq> Will do the same for the Static Code Analysis task later.
vivekp has quit [Ping timeout: 245 seconds]
ImQ009 has quit [Quit: Leaving]
KimSangYeon-DGU has quit [Remote host closed the connection]