Articles on sayandip18's Blog

Wrapping up!

sayandiph4@gmail.com (sayandip18) — Sun, 22 Aug 2021 23:37:54 +0000

This is my final week. Hameer helped me fix another bug and I updated the hypothesis PR. The tensor class is incomplete as of now. I hope to complete it after GSoC ends.

Weekly check-in #10

sayandiph4@gmail.com (sayandip18) — Sun, 15 Aug 2021 20:28:45 +0000

What did I do this week?

I fixed the bugs mentioned in the previous blog post.

What is coming up?

More bug fixes.

Did I get stuck?

No.

Weekly check-in #9

sayandiph4@gmail.com (sayandip18) — Wed, 04 Aug 2021 11:34:45 +0000

What did I do this week?

John managed to fix the bugs in the cmake code. TACO can now be used by downstream users easily.

Where did I get stuck?

I was unable to contribute and my activity during this time was limited to reading the book.

What is coming up?

The cmake work is now over. I will fix the bugs I detected while transferring the test suite to hypothesis.

Weekly check-in #8

sayandiph4@gmail.com (sayandip18) — Wed, 04 Aug 2021 11:32:35 +0000

What did I do this week?

With John's help, I managed to write some code that attempts to make TACO available to downstream users leveraging on the cmake config functionality.

Where did I get stuck?

The changes are not working as expected. TACO is still not available to downstream users.

What is coming up?

Pin-pointing and fixing the issues.

Weekly check-in #7

sayandiph4@gmail.com (sayandip18) — Wed, 28 Jul 2021 14:29:12 +0000

What did I do this week?

I managed to modify only a few bits and pieces of the code. However, the changes were not really aligned with our immediate goal and have been stashed to be committed later on.

Where did I get stuck?

There have been a lot of hiccups. Most of my time was spent reading the book. I was not particularly productive in coding.

What is coming up?

TACO headers should be detectable for downstream users using the command find_package().

GSoC blog post

sayandiph4@gmail.com (sayandip18) — Thu, 15 Jul 2021 04:48:46 +0000

Finally, the first evaluations have arrived! The coding period is now halfway through. The 6 weeks so far have been a great experience for me. Here is a brief overview of what I did throughout the time.

PR 472 aims at adding Hypothesis to the test suite. During the course of this PR, we have come across 5 bugs in Sparse that were caught from failing Hypothesis tests. The bugs are
- Issue 481
- Issue 500
- Issue 503
- Issue 505
- Issue 506
PR 492 aims to add the tensor class and relevant tests. The PR is stalled for the time being and will be resumed after the upcoming cmake PR is merged.
Cmake: So far, I have attempted twice, albeit unsuccessfully, to get cmake install and build TACO in Sparse. The upcoming cmake PRs will now take care of this.

Weekly check-in #6

sayandiph4@gmail.com (sayandip18) — Thu, 15 Jul 2021 04:12:04 +0000

What did I do this week?

I have started to go through the book "Professional Cmake" which John suggested to read. John has forked the TACO repository where we will be pushing our changes.

The implementation of the tensor class is kept aside for now. It will be resumed after the cmake code has been written for Sparse.

What is coming up?

Studying the cmake code in TACO before making it better.
Finishing up the cmake PR in Sparse after the TACO PR is merged.

Did I get stuck anywhere?

Yes. A lot of concepts in the book are not super clear to me now. But, I will be meeting with John to discuss all of those.

Weekly check-in #5

sayandiph4@gmail.com (sayandip18) — Tue, 06 Jul 2021 08:35:22 +0000

What did I do this week?

I met with John and we came up with a plan to improve the cmake code of TACO and then add cmake to Sparse.

What is coming up?

Studying and improving the cmake code. Finishing up the Tensor wrapper PR.

Did I get stuck anywhere?

No.

Weekly check-in #4

sayandiph4@gmail.com (sayandip18) — Thu, 01 Jul 2021 12:44:23 +0000

What did I do this week?

I designed the __array_ufunc__ protocol for the tensor class. I had a meeting with Hameer where he suggested some changes.

What is coming up?

Completion of the protocol as per Hameer's suggestions.

Did I get stuck anywhere?

No.

Weekly check-in #3

sayandiph4@gmail.com (sayandip18) — Mon, 21 Jun 2021 16:18:50 +0000

What did I do this week?

Added methods to the tensor class and started writing tests. The work is in this PR .

What is coming up?

Completing the work in the PR mentioned.

Did I get stuck anywhere?

No.

Weekly check-in #2

sayandiph4@gmail.com (sayandip18) — Wed, 16 Jun 2021 20:11:40 +0000

What did I do this week

The first three days were spent figuring out how to correctly configure the scikit-build installation. I had a Gmeet meeting with Hameer Abbasi where we tried to solve the issue. Hameer managed to solve most of the issues barring an issue with the setup.py file. The PR has been stalled for the time being.

I have made a PR to add TACO as a submodule.

I am currently working on the implementation of the tensor class. I have started a PR on that too.

What is coming up?

Implementation of the tensor functions, which I've compiled a list of.

Did I get stuck anywhere?

Yes. As of now the setup.py issue is bugging me. I am still trying to get the rest of the tests passed in the test suit migration PR. As of now only 15 tests are failing.

I spent quite some time trying to figure out how the tensor class will be designed. Getting stuck several times has led to fewer work being done in the previous two weeks than I had hoped. However, I'm getting the hang of it now and hoping to speed up the work in the next week.

Weekly check-in #1

sayandiph4@gmail.com (sayandip18) — Thu, 10 Jun 2021 01:10:16 +0000

Hello, world! My name is Sayandip Halder. I am a final year undergrad at Jadavpur University. I am working to implement ndarray-compatible API for Pytaco in the Sparse project. I will be implementing the APIs for the tensor and the sparse objects, add tests and finally add documentation for the implemented classes.

1. What did I do this week?

During the community bonding period I worked on transferring the current pytest-based Sparse test suite to Hypothesis in this PR.

I went through the documentation of cmake and scikit-build. I started this pull request that aims to install and compile taco along during sparse installation.

2. What is coming up next?

The next week will probably see me implementing the actual API of the combined pytaco sparse project.

3. Did I get stuck anywhere?

Moving the test suite seemed easy initially but there are several cases which I haven't been able to correctly generate.

I was not familiar with cmake and scikit-build so going through the docs for the first time was a challenge. The PR needs some more work.