<ShubhamAgrawal[4>
`return arma::accu(arma::square(input - target)) / target.n_cols;` This code is actually returns double
<turska79turska79>
this target.n_cols is some templated type, but i didnt figure it out that where i need to declare its type
<ShubhamAgrawal[4>
target.n_cols is arma matrix type
<turska79turska79>
my type declaration is
<turska79turska79>
using FeedForwardNetwork = mlpack::ann::FFN<mlpack::ann::MeanSquaredError<>, mlpack::ann::RandomInitialization>;
<turska79turska79>
using ReplayMethod = mlpack::rl::RandomReplay<ContinuousActionEnvironment>;
<turska79turska79>
using SoftActionCritic = mlpack::rl::SAC<ml::ContinuousActionEnvironment, FeedForwardNetwork, FeedForwardNetwork, ens::AdamUpdate, ReplayMethod>;
<turska79turska79>
Which class should have that member?
<ShubhamAgrawal[4>
<turska79turska79> "Which class should have that..." <- Target Class
<NabanitaDash[m]>
I made changes in `simple_dqn.hpp`. I want to run the test file `reward_clipping_test` to check if the made changes are correct. How to run that test file?
<ShubhamAgrawal[4>
NabanitaDash[m]: You need to recompile whole code, and then run the test
<ShubhamAgrawal[4>
NabanitaDash[m]: What change did you make btw?
<NabanitaDash[m]>
changed `boost::get` to `std::get` to check if it works
<swaingotnochill[>
Refer this..You can test specific tests quite easily.
<NabanitaDash[m]>
yup I am trying to find how to make only `simple_dqn` as `mlpack_knn`. Should I use `mlpack_simple_dqn`? and what about the test file ?
<rcurtin[m]>
Nabanita Dash: there's no standalone program for `simple_dqn`. if you want to build and run the tests, you'll need to make `mlpack_test` and run the corresponding test suite
<NabanitaDash[m]>
ah, okay
<rcurtin[m]>
(I know that can take a while to compile, but that should only be the first time. if you make subsequent modifications to the code, CMake should only recompile the files necessary)
<NabanitaDash[m]>
I see. That's better.
<rcurtin[m]>
if you are working to remove boost, that should hopefully help improve compilation times too and make it less painful 😄
<NabanitaDash[m]>
Yes, well. Why don't we use pybind rather than boost? It makes compilations easier
<rcurtin[m]>
I'm not sure I understand how pybind and boost are related, or how either makes compilation easier... can you clarify?
<NabanitaDash[m]>
Boost has got an extensive library than pybind. Pybind is header only and has STL. Using it might reduce execution time.
<NabanitaDash[m]>
header only implementation of boost*
<rcurtin[m]>
unless I am mistaken, pybind is a library for use in Python, but mlpack is C++, so I'm not following
<rcurtin[m]>
in any case, our goal is to remove boost and replace it with STL or other implementations anyway, so I am not sure that pybind would be helpful here
<NabanitaDash[m]>
I am just talking for the python bindings. It totally unrelated to building tests that I had mentioned above.
<rcurtin[m]>
oh---our Python bindings are done using Cython, not boost 👍️
<rcurtin[m]>
https://github.com/mlpack/mlpack/issues/2982 is open to replace our use of Cython with pybind11, which would be great, but I haven't heard anything from that person in a while
<NabanitaDash[m]>
Ah, I see. I will look into this.
<rcurtin[m]>
yeah, I need to check in on the status of that because it would be a really nice improvement
<rcurtin[m]>
but I am still catching up after spending most of December underwater doing other things
<rcurtin[m]>
I have lots of issues and PRs flagged to check in on... could take a while 😄
<NabanitaDash[m]>
Ah well.. I will let you know if I find something there.
<jonpsy[m]>
we have meet tmrw?
<rcurtin[m]>
yeah, first week of the month 👍️
<jonpsy[m]>
have you ever tried workation?
<jonpsy[m]>
whats your opinion on that??
<rcurtin[m]>
in some sense that's kind of what conference travel often is (or was before covid?) for me... I would go to some other city and work from there for a while and attend conference sessions
<rcurtin[m]>
it is cool because you get to see other places, but ultimately unsatisfying if you only work while you are there---time needs to be set aside to actually see the new place and do things there
<rcurtin[m]>
the other problem (at least for me, I am pretty introverted) is that it takes a while to make friends in a new place, etc., so there is generally no community and it can kinda be a lonely experience
<rcurtin[m]>
but overall, with "conference workations" for me, I have had a good time and seen a lot of really cool places (although I did not always make time to explore the cool parts of those places and spent a lot of time in conference centers 😄)
<jonpsy[m]>
i see thanks! Im an introvert myself! But I was planning workation with frens
<shrit[m]>
jonpsy: nope, I have not seen anyone who managed to make it with workation.
<shrit[m]>
It simply does not work, because we will be tired of the travel, and you have to work at the same time.
<shrit[m]>
Also, you need to keep the hours 9-5, since you are not working alone in most cases
<jonpsy[m]>
I see. If you're only going there for a 1-3 days that'd be a problem. But if you stay for a month or two there, the turnover actually comes out positive
<jonpsy[m]>
* out positive ( or so I've heard). So, you rent an apartment, and actually stay there, wander around in weekends.
<ShahAnwaarKhalid>
<jonpsy[m]> "i see thanks! Im an introvert..." <- Count me in 🖤😺
<jonpsy[m]>
Hahah sure! Wait, you're in kashmir right?
<jonpsy[m]>
that's literally paradise on earth
<jonpsy[m]>
I'm counting on you for free tourism there ;)
<turska79turska79>
I used to do some travel work in the past. Unfornately a bit long hours 6-7 days a week for few weeks at the time. Less touristing but got see some places at least
<jonpsy[m]>
I see, thanks for sharing :)
<ShahAnwaarKhalid>
<jonpsy[m]> "I'm counting on you for free..." <- Haha sure! Hit me up if you ever decide to come 🌼