mlpack IRC logs, 2021-08-04

Logs for the day 2021-08-04 (starts at 0:00 UTC) are shown below.

>
August 2021
Sun
Mon
Tue
Wed
Thu
Fri
Sat
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
--- Log opened Wed Aug 04 00:00:45 2021
02:51 < jonpsy[m]> <ShahAnwaarKhalid> "c2078893-7d08-4a64-a3d4-89a26a6c..." <- My day just got 1000000x better
02:51 < jonpsy[m]> but if it's really looking for heat, try compiling mlpack with ```-j4``` ;)
12:43 < RishabhGarg108Ri> @rcurtin I need your suggestion in one thing.
12:44 < RishabhGarg108Ri> in `DecisionTreeRegressor::Train()`, we have size checks like [this](https://github.com/mlpack/mlpack/blob/e6c8f7dec0a741deba6ba2a6c4558630b8930d11/src/mlpack/methods/decision_tree/decision_tree_regressor_impl.hpp#L439)
12:46 < RishabhGarg108Ri> This function compares `dataset.n_cols == labels.n_elem`. Since we are passing labels as a matrix it is causing an error because our labels is a matrix having two rows.
12:47 < RishabhGarg108Ri> So, would it make sense to change `dataset.n_cols == labels.n_elem` to `dataset.n_cols == labels.n_cols` or shall I try to change these checks in `DecisionTreeRegressor::Train()` ?
12:50 < RishabhGarg108Ri> Is there any case in mlpack other than xgboost where labels are matrix?
15:02 < rcurtin[m]> Shah Anwaar Khalid: haha, I worked hard to train my cats to stay off the keyboard 😃
15:02 < rcurtin[m]> jjb: able to join the meeting with Nippun this morning?
15:03 < rcurtin[m]> RishabhGarg108 (RishabhGarg108): I think it's just fine to use `labels.n_cols` for the check; responses should at least be a row vector anyway, meaning that the number of columns should be the same as the data
15:03 < rcurtin[m]> I don't think there is currently any other case than XGBoost where the labels/responses are a matrix 👍️
15:03 < RishabhGarg108Ri> Ok. Thanks!
16:23 < RishabhGarg108Ri> @ryan:ratml.org A couple of days back, we were discussing about the optimal value for `MultipleRandomDimensionSelect` for regression case. Since we are implementing `XGBoostTreeRegressor` should we worry about it now or is it fine?
16:33 < heisenbuugGopiMT> @rcurtin what is it comparing exactly?
16:33 < heisenbuugGopiMT> [Failed Test](https://dev.azure.com/mlpack/mlpack/_build/results?buildId=7075&view=logs&j=24d3abe3-ef0b-5deb-3aab-64d839de2c3c&t=b3d23cc5-b695-5043-0f03-2084bf2ff0b5&l=33)
16:34 < heisenbuugGopiMT> When I am printing the values from matrix in I am getting the right values, but somehow this test is failing.
16:34 < heisenbuugGopiMT> Any idea why?
16:35 < heisenbuugGopiMT> Also about this [comment](https://github.com/mlpack/mlpack/pull/2942#issuecomment-892262144)
16:40 < shrit[m]> heisenbuug (Gopi M Tatiraju): I will give it a look soon, if you did not hear from me, do not hesitate in pinging me 👍️
16:40 < shrit[m]> * heisenbuug (Gopi M Tatiraju): I will give it a look soon, if you do not hear from me, do not hesitate in pinging me 👍️
16:49 < rcurtin[m]> heisenbuug (Gopi M Tatiraju): that's just checking that the first dimension of the data has the values `{1, 2, 3, 4, 5}`
17:28 < heisenbuugGopiMT> Okay, I am looking for `canParse` now, I will update on that soon.
17:30 < shrit[m]> Perfect,
17:55 < shrit[m]> I have exported ensmallen to nuget using vcpkg from Linux
17:55 < shrit[m]> I have now a .nupkg file on my machine, does anyone have an idea how to verify if the file is correctly exported before submitting it?
18:08 < zoq[m]> There is not really a submitting process, it's basically create a nuget accoutn and upload the file, you can always remove the package afterwards.
18:08 < zoq[m]> So I would just do that and trigger the CI.
18:50 < swaingotnochill[> zoq if I load a gan model, how can i access the generator and discriminator? I can't seem to find any method for that.
18:51 < swaingotnochill[> ```data::Save("./saved_csv_files/ouput_mnist.csv", generatedData, false, false); ```
18:52 < swaingotnochill[> ```data::Load("./saved_models/ganMnist.bin", "ganMnist", ganModel)```
18:55 < say4n[m]> Line #295 of this: https://www.mlpack.org/doc/mlpack-3.2.1/doxygen/gan_8hpp_source.html ?
18:55 < say4n[m]> swaingotnochill: ^
18:57 < swaingotnochill[> say4n[m]: I can use the generator in the training file itself where I created my GAN. But, I am not able to access it after loading a trained GAN model.
18:58 < say4n[m]> <swaingotnochill[> "```data::Load("./saved_models/ga" <- Taking a wild shot but does ganModel.Generator() not work?
18:59 < swaingotnochill[> say4n[m]: sadly no...
18:59 < say4n[m]> Ah :/
19:00 < heisenbuugGopiMT> @shrit:matrix.org I found this on boost's documentation [page](https://theboostcpplibraries.com/boost.spirit-api#:~:text=boost%3A%3Aspirit%3A%3Aqi%3A%3Aparse()%20does,of%20boost%3A%3Aspirit%3A%3Aqi%3A%3Aparse().)
19:00 < heisenbuugGopiMT> If I understand correctly if we are anyways trimming the token, we won't enter this case, right?
19:00 < heisenbuugGopiMT> * @shrit:matrix.org I found this on boost's documentation [page](https://theboostcpplibraries.com/boost.spirit-api#:~:text=boost%3A%3Aspirit%3A%3Aqi%3A%3Aparse()%20does,of%20boost%3A%3Aspirit%3A%3Aqi%3A%3Aparse())
19:00 < heisenbuugGopiMT> If I understand correctly if we are anyways trimming the token, we won't enter this case, right?
19:01 < swaingotnochill[> swaingotnochill[: I am not sure how it is serialized. It might be one of the reason I can't access it...Or I am just wrong from the start 😞
19:01 < heisenbuugGopiMT> * @shrit:matrix.org I found this on boost's documentation page
19:01 < heisenbuugGopiMT> https://theboostcpplibraries.com/boost.spiritapi#:~:text=boost%3A%3Aspirit%3A%3Aqi%3A%3Aparse()%20does,of%20boost%3A%3Aspirit%3A%3Aqi%3A%3Aparse()
19:01 < heisenbuugGopiMT> If I understand correctly if we are anyways trimming the token, we won't enter this case, right?
19:06 < heisenbuugGopiMT> * @shrit:matrix.org I found this on boost's documentation page
19:06 < heisenbuugGopiMT> https://theboostcpplibraries.com/boost.spirit-api#:~:text=boost%3A%3Aspirit%3A%3Aqi%3A%3Aparse()%20does,of%20boost%3A%3Aspirit%3A%3Aqi%3A%3Aparse().
19:06 < heisenbuugGopiMT> If I understand correctly if we are anyways trimming the token, we won't enter this case, right?
19:08 < shrit[m]> heisenbuug (Gopi M Tatiraju): which case you are referring too?
19:09 < heisenbuugGopiMT> Regarding `canParse`
19:09 < heisenbuugGopiMT> `canParse = qi::parse(...)`
19:10 < swaingotnochill[> (edited) zoq if ... => zoq Kartik K. Khullar if ...
19:10 < swaingotnochill[> (edited) ... Khullar if ... => ... Khullar if ...
19:10 < swaingotnochill[> (edited) zoq Kartik K. Khullar if ... => zoq if ...