rcurtin_irc changed the topic of #mlpack to: mlpack: a scalable machine learning library (https://www.mlpack.org/) -- channel logs: https://libera.irclog.whitequark.org/mlpack -- NOTE: messages sent here might not be seen by bridged users on matrix, gitter, or slack
SlackIntegration has quit [Ping timeout: 268 seconds]
SlackIntegration has joined #mlpack
krushia has quit [*.net *.split]
krushia has joined #mlpack
<jonpsy[m]> Eshaan Agarwal, zoq when is the submission date?
rcurtin[m] has quit [Ping timeout: 268 seconds]
rcurtin[m] has joined #mlpack
<EshaanAgarwal[m]> <jonpsy[m]> "Eshaan Agarwal, zoq when is the..." <- Oct 31 to Nov 7.
<EshaanAgarwal[m]> jonpsy: zoq: you guys get a chance to look at it ?
<EshaanAgarwal[m]> Also if it could be possible can we get on PPO code pushed and merged ?
<zoq[m]> I did some debugging, and I think the reward is not correct.
<EshaanAgarwal[m]> <zoq[m]> "I did some debugging, and I..." <- Reward as in the reward given by the environment? 👀
<zoq[m]> Yes
<EshaanAgarwal[m]> zoq[m]: Can you please elaborate more on this ?
<zoq[m]> Leave a comment later on the PR.
<EshaanAgarwal[m]> zoq[m]: Sure ! Can I help in cleaning and getting read PPO code. If it is possible for you to push it ?
<zoq[m]> But please take another look at the policy implementation, ideally you can step through the code with gdb.
<zoq[m]> Looks like the reward is fine.
<zoq[m]> So what we have to do is to step through each line, to find out what’s wrong.
<EshaanAgarwal[m]> zoq[m]: I will take a look again.
<EshaanAgarwal[m]> zoq[m]: If I am not wrong you are talking about this in regard to HER right ?
<EshaanAgarwal[m]> * Sorry but If I am not wrong you are talking about this in regard to HER right ?
<zoq[m]> EshaanAgarwal[m]: Correct
<EshaanAgarwal[m]> jonpsy: zoq fieryblade https://meet.google.com/pnp-rtjw-unz
<EshaanAgarwal[m]> <EshaanAgarwal[m]> "jonpsy: zoq fieryblade https://..." <- can we reschedule this to tomorrow at same time 10:00 PM IST ? Meanwhile, i will take another look at HER with gdb. I also intend to update the progress repository so that we also get started with wrapping up the documentation and report.
<jonpsy[m]> not sure i can make it monday
<jonpsy[m]> Eshaan Agarwal: have you written this code entirely from scratch or have taken help from a resource?
<EshaanAgarwal[m]> <jonpsy[m]> "not sure i can make it monday" <- When can we do it then ?
<EshaanAgarwal[m]> <jonpsy[m]> "Eshaan Agarwal: have you..." <- I have used the psuedocode in paper and implementation done in Intel Coach Labs as reference. I have mentioned that in PR. But since the mlpack library had a different structure, I implemented in accordance it.
<jonpsy[m]> Okay I had one question
<jonpsy[m]> `modifiedState(action.action) = 1 - modifiedState(action.action)` ?
<jonpsy[m]> What exactly are we trying to achieve here.
<EshaanAgarwal[m]> > <@jonpsy:matrix.org> `modifiedState(action.action) = 1 - modifiedState(action.action)` ?
<EshaanAgarwal[m]> >
<EshaanAgarwal[m]> flipped the bit in n length binary vector ! here agent will provide index as the action ! that index will be filpped and then produced vector will be next state
<EshaanAgarwal[m]> > What exactly are we trying to achieve here.
<EshaanAgarwal[m]> EshaanAgarwal[m]: that will give 1 for 0 at that index and 0 when its 1 at that index
<jonpsy[m]> `modifiedState(action.action) = 1 - modifiedState(action.action)`
<jonpsy[m]> so we're calling constructor?
<EshaanAgarwal[m]> jonpsy[m]: Modified state is a col vec right ?
<EshaanAgarwal[m]> In which we stored our present state
<jonpsy[m]> is it?
<jonpsy[m]> <EshaanAgarwal[m]> "> <@jonpsy:matrix.org> `..." <- If `modifiedState(action.action)` is 0, what happens? It becomes -1?
<EshaanAgarwal[m]> jonpsy[m]: it will become 1 right ? 1-0 is 1 which will get stored at that place
<jonpsy[m]> assuming the above code is correct, we could've just done `nextState.Data() = 1 - state.Data()` ? Why create unnecessary copies
<jonpsy[m]> * - state.Data()(action.action)` ?
<EshaanAgarwal[m]> jonpsy[m]: we have to flip that index right ? nextState is empty as of now
<akhunti1[m]> Hi Team , any suggestion what should i use to deploy Mlpack C++ based model ? like flask and Django use for python based Machine learning model .
<EshaanAgarwal[m]> EshaanAgarwal[m]: i think what you mean to say is this `nextState.Data()= state.Data(); nextState.Data()(action.action) = 1 - state.Data()(action.action);`
<EshaanAgarwal[m]> `
<jonpsy[m]> or you could just `std::move` the last part
<jonpsy[m]> but yes, avoid copies
<EshaanAgarwal[m]> jonpsy[m]: sure i will make that change.