<heisenbuugGopiMT>
Check here `Dealing with Sparse Matrix`
<shrit[m]>
Yeah, it is not working as suppose?
<shrit[m]>
* Yeah, it is not working as supposed to be?
<heisenbuugGopiMT>
No it is working with `txt` and other formats but we never had support for `csv`.
<shrit[m]>
Did you figure out where the issue is coming from?
<shrit[m]>
I never used sparse matrix, but I think rcurtin did, I am just surprised that it is not working at all
<heisenbuugGopiMT>
Yea, in the `Load()` for `sp_mat`...
<heisenbuugGopiMT>
The exceptino is coming from there
<heisenbuugGopiMT>
exception*
<shrit[m]>
with CSV
<shrit[m]>
I remember you told me there were tests related to sparse matrices which are failing?
<shrit[m]>
it is related to this issue?
<heisenbuugGopiMT>
Yea, coz we removed the old load for sp_mat
<heisenbuugGopiMT>
And the new load is only used when we have a csv file
<heisenbuugGopiMT>
But tests are for loading sparse data from `txt` `tsv` and `bin`
<heisenbuugGopiMT>
We don't even have tests for sp_mat where we are trying to load them for a csv file, I think I mentioned this already...
<heisenbuugGopiMT>
Yupp
<heisenbuugGopiMT>
Sorry we had a power cut, there was a huge sound like boom and then power went away...
<shrit[m]>
Okay, no worries, so basically if you bring back the load function for `sp_mat` it will load all of the extensions apart from the `csv` one.
<shrit[m]>
So we need to add support for `csv` in the load `sp_mat` function
<heisenbuugGopiMT>
Yuppp...
aakashi2001 has quit [Changing host]
aakashi2001 has joined #mlpack
<heisenbuugGopiMT>
But even doing that was not working coz we changed the template parameter and were not able to work out those overloads if you remember.
<heisenbuugGopiMT>
We tried that on call
<heisenbuugGopiMT>
But it was not at all calling the `sp_mat` version of load
<shrit[m]>
I see, I want to put everything here in the chat, because we need to check if this is not done intentionally (e.g, specific reason) I think rcurtin will provide us with his insight in a couple of hours on this one
<heisenbuugGopiMT>
Also one thing, arma does have support for loading sparse data from csv file
<heisenbuugGopiMT>
There is an case for `csv_ascii`
<heisenbuugGopiMT>
* There is a case for `csv_ascii`
aakashi2001 has joined #mlpack
<heisenbuugGopiMT>
Yea, this might be intentional that's what I was thinking.
<heisenbuugGopiMT>
We kinda added a new feature, but we don't know if we actually needed it.
aakashi2001 has quit [Client Quit]
aakashi2001 has joined #mlpack
aakashi2001 has quit [Remote host closed the connection]
aakashi2001 has joined #mlpack
aakashi2001 has quit [Quit: Leaving]
aakashi2001 has joined #mlpack
aakashi2001 has quit [Changing host]
aakashi2001 has joined #mlpack
<jonpsy[m]>
<Aakash-kaushikAa> "Hi, @kartikdutt18, @marcusedel:m" <- looks great, there's a spelling mistake "mlpack/mlapack" in one of your subheading. Awesome regardless!
aakashi2001 has quit [Quit: Leaving]
<jonpsy[m]>
would also be nice to add some pics :)
aakashi2001 has joined #mlpack
aakashi2009 has joined #mlpack
aakashi2009 has quit [Changing host]
aakashi2009 has joined #mlpack
aakashi2009 has quit [Remote host closed the connection]
aakashi2001 has quit [Remote host closed the connection]
aakashi2001 has joined #mlpack
aakashi2001 has quit [Remote host closed the connection]
aakashi2001 has joined #mlpack
aakashi2001 has quit [Changing host]
aakashi2001 has joined #mlpack
aakashi2009 has joined #mlpack
aakashi2009 has joined #mlpack
aakashi2009 has quit [Remote host closed the connection]
aakashi2001 has quit [Ping timeout: 248 seconds]
aakashi2001 has joined #mlpack
<ABHINAVANAND[m]>
zoq can you take a look at the colab notebook that I shared in the PR.
<rcurtin[m]>
shrit: heisenbuug (Gopi M Tatiraju) : that link about loading sparse matrices is actually out of date, so I suppose we should update it 😃
<rcurtin[m]>
I think that loading a CSV is possible for sparse matrix types with the `csv_ascii` format
<rcurtin[m]>
so, I guess it should work, but perhaps it didn't before?
<rcurtin[m]>
at least personally I think it might be okay (and maybe better) to just leave it how it is in master for now, and then open an issue to fix the sparse CSV support later (it probably will be an easy fix, I think)
aakashi2001 has quit [Remote host closed the connection]
<kartikdutt18kart>
Looks great! Awesome work!
aakashi2001 has joined #mlpack
aakashi2001 has quit [Changing host]
aakashi2001 has joined #mlpack
aakashi2001 has quit [Quit: Leaving]
<rcurtin[m]>
Aakash-kaushik (Aakash kaushik): nice post! it is very cool to see all the things that you did in one place 😃 I'm excited that we have all this new support now 🚀
aakashi2001 has joined #mlpack
aakashi2001 has quit [Changing host]
aakashi2001 has joined #mlpack
<jonpsy[m]>
zoq: could you review the PR, I've implemented the logic in code which we discussed yesterday.
<zoq[m]>
<ABHINAVANAND[m]> "zoq can you take a look at the..." <- Yes, will take a look in a moment.
<zoq[m]>
<jonpsy[m]> "zoq: could you review the PR, I'" <- Sure thing.
<Aakash-kaushikAa>
Thank you @jonpsy:matrix.org and @kartikdutt18, also corrected the spelling mistake.
<Aakash-kaushikAa>
Thanks @ryan:ratml.org, I will also be working on some of the things in future that are open right now.
<Aakash-kaushikAa>
> @Aakash-kaushik: nice post! it is very cool to see all the things that you did in one place 😃 I'm excited that we have all this new support now 🚀
<heisenbuugGopiMT>
@shrit:matrix.org We also have mlpack weekly meetup at the same time...
aakashi2001 has quit [Remote host closed the connection]
aakashi2001 has joined #mlpack
aakashi2001 has joined #mlpack
aakashi2001 has quit [Changing host]
aakashi2001 has quit [Ping timeout: 268 seconds]
<swaingotnochill[>
ryan You once talked about GPU Integration for Training with mlpack. Can you give some leads and resources to get started with that?
<jonpsy[m]>
i think bandicoot is under development : )
<swaingotnochill[>
I don't think it is active. Can't wait for them.
<rcurtin[m]>
yeah, unfortunately, there is not much time that I have to contribute to it, and it is pretty inactive right now. I really hope to be able to contribute more time, perhaps once we finish with the mlpack 4 refactoring 😃
<jonpsy[m]>
zoq: hey, you free rn?
<jonpsy[m]>
Looks like not, lets do this tommorow. What I wanted to ask was, in ```BackwardEQL```, remember we also passed ```arma::mat weightSpace``` ?
<jonpsy[m]>
Now, the problem is that the loss is sum of two losses. L = lambda * L_a + (1 - lambda ) * L_b, ```L_a = || Q - Target ||^2 ``` ```L_b = | wQ - wTarget | ```, now where exactly do I make my loss function do that??
<jonpsy[m]>
* Looks like not, lets do this tommorow. What I wanted to ask was, in `BackwardEQL`, remember we also passed `arma::mat weightSpace` ?
<jonpsy[m]>
Now, the problem is that the loss is sum of two losses. ```Loss = lambda * L_a + (1 - lambda ) * L_b```, `L_a = || Q - Target ||^2 ` `L_b = | wQ - wTarget | `, now where exactly do I make my loss function do that??