Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Apr 12 04:52

    hameerabbasi on master

    Fixed typo in docstring (#457) (compare)

  • Apr 12 04:52
    hameerabbasi closed #457
  • Apr 12 01:18
    KanzaSheikh opened #457
  • Apr 08 09:53
    AmPhIbIaN26 commented #443
  • Apr 08 09:52
    AmPhIbIaN26 commented #443
  • Apr 08 08:46
    hameerabbasi commented #443
  • Apr 07 21:17
    AmPhIbIaN26 commented #443
  • Apr 07 03:48

    hameerabbasi on master

    Attempt to reformat docstring t… (compare)

  • Apr 07 03:48
    hameerabbasi closed #456
  • Apr 07 03:45
    Carreau commented #456
  • Apr 07 01:45
    Carreau opened #456
  • Apr 02 17:18
    hameerabbasi commented #455
  • Apr 02 17:18

    hameerabbasi on master

    Update docs for GCXS and CSR/CS… (compare)

  • Apr 02 17:18
    hameerabbasi closed #455
  • Apr 02 15:16
    daletovar commented #455
  • Apr 02 15:13
    daletovar synchronize #455
  • Apr 01 21:52
    daletovar synchronize #455
  • Apr 01 21:50
    daletovar synchronize #455
  • Apr 01 20:23
    daletovar commented #455
  • Apr 01 20:11
    daletovar synchronize #455
AmPhIbIaN26
@AmPhIbIaN26
@sayandip18 thanks a lot I will look these up
Harshit Gupta
@hgupta1153:matrix.org
[m]
hameerabbasi: Sir I have edited my proposal as you asked. can you check and review it ?
hameerabbasi
@habbasi:matrix.org
[m]
Harshit Gupta: Looks pretty good. Some grammar mistakes and so on to fix up, but that doesn't matter too much, I think. My previous comment stands about your internship -- Working 50-hour weeks isn't sustainable, and may burn you out.
1 reply
It's missing some broadcasting elemwise operations, but the rest looks good.
Harshit Gupta
@hgupta1153:matrix.org
[m]

It's missing some broadcasting elemwise operations, but the rest looks good.

hameerabbasi: Sir I have edited my proposal as you asked. can you check and review it ?

habbasi
@habbasi:matrix.org
[m]
Harshit Gupta: If you're still thinking of doing the internship; I'd put it in, it's best to declare all forseeable commitments, I may be wrong but it's in the rules too.
1 reply
hameerabbasi
@habbasi:matrix.org
[m]
Okay, so you won't do an internship if you do GSoC?
Harshit Gupta
@hgupta1153:matrix.org
[m]
right.
hameerabbasi
@habbasi:matrix.org
[m]
Then that's fine. :) How to distribute your work is up to you, but 2 hours a day on weekends sounds a bit weird to me given you won't have an internship if you do GSoC.
1 reply
Harshit Gupta
@hgupta1153:matrix.org
[m]
hameerabbasi: Sir I have edited it accordingly. You want to add something?
hameerabbasi
@habbasi:matrix.org
[m]
That looks good to me! Thanks for working on it so much. I think it's more-or-less ready for submission.
1 reply
Sayandip Halder
@sayandip18
@hameerabbasi This is my proposal so far. Let me know where it is lacking and how it can be improved, especially the plan.
Hameer Abbasi
@hameerabbasi

A lot of stuff will be made easy by the TACO folks working on the array API similar to NumPy.

One difference will be that we're planning for a new package that wraps pytaco.

hameerabbasi
@habbasi:matrix.org
[m]
@sayandip18: It'll be called ndsparse. Your proposal looks a bit thin, you should put in enough info so that we can determine you understand what you're meant to do. :) Most of it looks good in terms of timeline, etc, just needs beefing up in terms of technical details.
Sayandip Halder
@sayandip18
In that case, there will be a lot of changes in the plan, right? The plan will be finalised during the community bonding period.
And I've also mentioned in the proposal that the road map I have proposed will be subject to a lot of changes. So, this shouldn't be a problem, right?
hameerabbasi
@habbasi:matrix.org
[m]
Yes, I realise that things will be in flux, so that's okay IMO. The main thing there is you should understand what you're doing. Mention overall details of codegen, tech stack (pybind11), etc.
Sayandip Halder
@sayandip18

Mention overall details of codegen, tech stack (pybind11), etc.

Sure. Will do.

Arib Alam
@aribalam
@hameerabbasi I have updated my doc accordingly. I have added a section for handling reductions. Kindly have a look!
Sayandip Halder
@sayandip18
@hameerabbasi here is my updated proposal
Let me know if the details are enough now
Hameer Abbasi
@hameerabbasi

@hameerabbasi I have updated my doc accordingly. I have added a section for handling reductions. Kindly have a look!

That looks ready to submit! Thanks!

@hameerabbasi here is my updated proposal

That looks great too, thanks! :D

Everyone, remember: April 13, 18:00 UTC is the submission deadline. That's in 2 days and a little less than 10 hours. Don't miss it!
Sayandip Halder
@sayandip18
Hi! I've uploaded my proposal to the GSoC dashboard.
Harshit Gupta
@hgupta1153:matrix.org
[m]
I have also uploaded my proposal to dashboard on 9th April
Sayandip Halder
@sayandip18
In pydata/sparse#163, it is mentioned "The function name to test (both sparse and NumPy/SciPy) or callable". What does this mean?
Roberto Rosmaninho
@Robertorosmaninho
Hello everyone, I just wrote a draft of my project proposal! Can someone check and review it? @guilhermeleobas @hameerabbasi
Any feedbacks are welcome :)
Guilherme Leobas
@guilhermeleobas
Hi @Robertorosmaninho, your proposal looks good to me. If possible, you might want to include a couple of links indicating the work you’ve done in the past. Specially the MLIR for the D lang
Roberto Rosmaninho
@Robertorosmaninho
Hi @guilhermeleobas, thanks for the feedback! I just updated the project with some useful links!
Harshit Gupta
@hgupta1153:matrix.org
[m]
hameerabbasi: Sir why are we calling self.func() in __get_fill_value(). I was tinkering with the code and it just showed me recursion limit exceeded.
1 reply
Hameer Abbasi
@hameerabbasi
@Robertorosmaninho I'm sorry I got to this so late, but it'd be awesome if you could include a rough idea of the parts of code that you'd have to touch. I believe you can get them from the old TACO PR.
@hgupta1153:matrix.org Because you're passing in the same types now, whereas before, it was only getting passed ndarrays I believe.
Sayandip Halder
@sayandip18

In pydata/sparse#163, it is mentioned "The function name to test (both sparse and NumPy/SciPy) or callable". What does this mean?

@hameerabbasi

Hameer Abbasi
@hameerabbasi
@sayandip18 Mostly that it'll take two callables: sparse_func, numpy_func and compare the results. We were also thinking of moving to Hypothesis for the tests: https://hypothesis.readthedocs.io/en/latest/
Thanks for tagging me. :)
If it throws an exception, the exception should be the same in both cases as well.
Exception type I should say -- The message/traceback isn't to be compared.
Roberto Rosmaninho
@Robertorosmaninho
Hello @hameerabbasi, don't worry I still have some time! Could you be clearer about what do you expect with this rough idea of the coding part?
Hameer Abbasi
@hameerabbasi
Just a list of files you'd have to touch (incomplete is fine, just need a starting point), and a brief description of the changes.
Harshit Gupta
@hgupta1153:matrix.org
[m]

Because you're passing in the same types now, whereas before, it was only getting passed ndarrays I believe.

Yes Sir I am passing CSR directly.

Hameer Abbasi
@hameerabbasi

Yes Sir I am passing CSR directly.

If you do that, it'll come back into __array_function__ with the exact same arguments, and then call the ufunc again. :) What you actually need to do is add an extra case in __array_function__ if all of them are CSR/CSC and actually implement the algorithm there with Numba. :)

Harshit Gupta
@hgupta1153:matrix.org
[m]
ok Sir I would start learning about numba before community bonding period.
Roberto Rosmaninho
@Robertorosmaninho
@hameerabbasi I just updated the proposal, could you check again please?
Harshit Gupta
@hgupta1153:matrix.org
[m]
@Robertorosmaninho: Sir you have shared editable link.
Roberto Rosmaninho
@Robertorosmaninho
Hi @hgupta1153:matrix.org , thank you for the advice! I just edit the permission. :)
Hameer Abbasi
@hameerabbasi
@Robertorosmaninho Looks good for submission, thanks!
Roberto Rosmaninho
@Robertorosmaninho
@hameerabbasi Thank you for your time and consideration!