<jonpsy[m]>
Can someone explain the difference between ```DiscreteActionEnvType``` and ```ContinuousActionEnvType```, they look pretty much the same thing to me
<RishabhGarg108Ri>
I used `scp benchmark.cpp -P 2222 rishabh@orange.ratml.org:~` to copy a file, but it is giving me this error - `ssh: connect to host orange.ratml.org port 22: Operation timed out`
<RishabhGarg108Ri>
Can somebody help me out with what am I doing wrong here?
<say4n[m]>
<RishabhGarg108Ri> "I used `scp benchmark.cpp -P 222" <- You need to pass the port argument before the filename for scp to actually use the port. Usage would be something like `scp -P port path/to/local_file remote_host:path/to/remote_file`.
<RishabhGarg108Ri>
Oops silly me, jumbled the order of arguments😅. Thanks @say4n:matrix.org :)
<say4n[m]>
Np! :)
<RishabhGarg108Ri>
@rcurtin How do I compile mlpack programs on your machine?
<RishabhGarg108Ri>
When I tried linking with armadillo, it couldn't find it, but while building from source, cmake was able to find it
<shrit[m]>
What is the message error you are getting?
<RishabhGarg108Ri>
Hey @shrit:matrix.org I was trying `mosh`. I installed iit and when I run `mosh -p 2222 rishabh@orange.ratml.org` then nothing happens, while connecting via ssh is working just fine. Is there some initial setup which I have to do after installing `mosh`?
<heisenbuugGopiMT>
Can you also paste the command you are using to compile `benchmark.cpp`?
<heisenbuugGopiMT>
I think you are using something like `g++ -o benchmark benchmark.cpp`?
<RishabhGarg108Ri>
`g++ benchmark.cpp -I $MLPACK_INCLUDE -lmlpack -larmadillo` Here `MLPACK_INCLUDE` is the path to the include folder of mlpack
<rcurtin[m]>
hmm, and what is the output this gives? (sorry for the slow response, I slept in today 😃)
<rcurtin[m]>
oh! you pasted it already
<rcurtin[m]>
here the situation is that there is a version of mlpack installed to `/usr/local/lib/` which is out of date
<shrit[m]>
Maybe there are different version of armadillo
<shrit[m]>
it is looking for armadillo9
<rcurtin[m]>
so just like you specified `-I`, you'll also need to specify `-L /path/to/mlpack/build/lib/`, and I think the `-L` needs to come before the linker commands `-l`
<RishabhGarg108Ri>
I also tried compiling with `g++ benchmark.cpp -I $MLPACK_INCLUDE -L $MLPACK_LIB -larmadillo. But that too gave some undefined reference errors
<rcurtin[m]>
and, when you run, you'll need to set `LD_LIBRARY_PATH` to that same directory you used for `-L`
<RishabhGarg108Ri>
Yeah. I did set the `LD_LIBRARY_PATH = $MLPACK_LIB` too.
<rcurtin[m]>
you'll still need `-lmlpack` even if you specify `-L` (and `-L` should be a directory, not a library name)
<RishabhGarg108Ri>
Adding `-L $MLPACK_LIB` and `-lmlpack` did the trick :)
<rcurtin[m]>
awesome! sorry that I have an old mlpack installation floating around on that system 😃
<RishabhGarg108Ri>
That's fine. I am gradually getting used to these. Now its not that hard to debug these kind of issues as it used to be when I started on mlpack :)
<rcurtin[m]>
:) I think that is how it always goes. the first time I encountered an error where there was another version of libmlpack.so that I was unintentionally linking to, it took me quite a while to figure it out
<heisenbuugGopiMT>
Just a thought, maybe we can have a section on mlpack's website where we can address these common issues?
<heisenbuugGopiMT>
Even I faced the same issue initially.
<rcurtin[m]>
yeah, maybe a developer FAQ or something? that could be really nice
<RishabhGarg108Ri>
@ryan:ratml.org If I install any python package, then will it create any problem for you?
<RishabhGarg108Ri>
I am asking this because I want to install the xgboost python library. There are not many good tutorials on the usage of xgboost CLI
<rcurtin[m]>
I can do it from the system package manager I think, hang on
<RishabhGarg108Ri>
I did it myself :)
<RishabhGarg108Ri>
I was just asking if it will interfere with your existing environment or not. If that's the case, I will work in a virtual environment
<rcurtin[m]>
oh, ok, well in any case I installed python3-xgboost too
<rcurtin[m]>
it won't interfere with the existing environment---everything you are doing is local to your user 👍
<rcurtin[m]>
if the python3-xgboost package causes an issue just let me know and I will remove it :)
<RishabhGarg108Ri>
Sure. I will tell you if it causes issues :+1:
<RishabhGarg108Ri>
@ryan:ratml.org I have started the benchmark script. I used the same datasets that I used last time for `DecisionTreeRegressor` and using just one core.
<rcurtin[m]>
👍️ sounds good
<RishabhGarg108Ri>
Lets see how long it takes.
<RishabhGarg108Ri>
Right now I used only the default parameters and single core for both
<RishabhGarg108Ri>
Tomorrow, I am thinking of doing it for a variety of hyperparameters and multiple cores.