Tutorial bugfixes and CI integration #472

srush · 2021-01-20T14:42:45Z

There were some a couple remaining issues with the tutorial due to recent changes and conflicts in edits. This fixes them and runs on the latest version.

oxinabox · 2021-01-20T14:46:29Z

Can we add the tutorial back to the CI? so we can be sure not to break it in the future?
It should be just adding it here:

dex-lang/makefile

Lines 88 to 92 in d178dc0

    
           example-names = mandelbrot pi sierpinski rejection-sampler \ 
        
                           regression brownian_motion particle-swarm-optimizer \ 
        
                           ode-integrator mcmc ctc raytrace particle-filter \ 
        
                           isomorphisms ode-integrator linear_algebra fluidsim \ 
        
                           sgd chol

dan-zheng

Thanks for further cleaning up the tutorial!

dan-zheng · 2021-01-20T14:48:57Z

examples/tutorial.dx

+interface Add a
+  add : a -> a -> a
+  sub : a -> a -> a
+  zero : a


Could you please briefly explain why the trailing ' can be dropped from the declaration names here?

I tried dropping the trailing ' from names a few days ago and got "multiple definition" errors due to conflicts with Prelude declarations. Maybe that's not an issue anymore?

I believe the Multiple Definitions error was coming from the type of tableMean

def tableMean' [Add v, VSpace v] (x : n => v) : v =

which I changed to

def tableMean' [VSpace v] (x : n => v) : v =

VSpace implies Add. But not sure why that breaks it.

Also I think you add the scaling in VSpace reversed (it's mul not div right?)

I'm pretty sure re-defining a type class should also error. Can you please add this example to the makefile tests so that we can check if it succeeds in CI?

Added. Needed to download the files in the CI. hopefully that is okay.

Hmmm this is a good question... I honestly don't know whether we can do that or not. The worst part is that the MNIST download page doesn't even mention a license, so I'm a bit reluctant to do so 😕 Could we use a different dataset by any chance?

examples/tutorial.dx

srush · 2021-01-20T15:05:10Z

@oxinabox Can you briefly explain how CI works? How do I indicate that lines in the tutorial should be errors on purpose?

dan-zheng · 2021-01-20T15:25:41Z

Can we add the tutorial back to the CI? so we can be sure not to break it in the future?

Does the tutorial require locally downloading MNIST files? Seems like it would break easily on CI machines, unless there's some Dex dataset downloading logic within Dex itself (getOrCloneMNISTData). I would be really happy to see that!

Can you briefly explain how CI works? How do I indicate that lines in the tutorial should be errors on purpose?

As @oxinabox kindly noted, simply add tutorial to example-names in makefile. Then, you can run tests (including testing examples/tutorial.dx) via make tests.

dex-lang/makefile

Lines 88 to 92 in d178dc0

    
           example-names = mandelbrot pi sierpinski rejection-sampler \ 
        
                           regression brownian_motion particle-swarm-optimizer \ 
        
                           ode-integrator mcmc ctc raytrace particle-filter \ 
        
                           isomorphisms ode-integrator linear_algebra fluidsim \ 
        
                           sgd chol

srush · 2021-01-20T15:54:40Z

My assumption that the CI would fail on an error in the code. How do I indicate in Dex that certain errors are expected?

dan-zheng · 2021-01-20T16:40:59Z

My assumption that the CI would fail on an error in the code. How do I indicate in Dex that certain errors are expected?

Ah, good question. I think dex script foo.dx is designed to produce a quine: you can save the output to a file, and re-running dex script on that file should produce itself as output.

I think this "expected output" checking may apply not only to successful code blocks (e.g. :p value) but also errors? You can individually test examples/tutorial.dx by running make run-examples/tutorial - the Makefile has some neat logic for supporting that.

Hope this helps, let us know if you have further specific questions!

srush · 2021-01-20T16:43:01Z

Thanks! I'll do that in another PR. One last question: Is there a command like dex script to build the static evaluated html (as opposed to the server)?

oxinabox · 2021-01-20T19:38:22Z

I have no answered @dan-zheng has not given better.

srush · 2021-01-20T20:10:00Z

Neat, this passes now using the quine thing. That's really cute, I had no idea those lines were tests.

dan-zheng · 2021-01-20T20:18:02Z

One last question: Is there a command like dex script to build the static evaluated html (as opposed to the server)?

Yep! All the support is in the Makefile.

Rendered static HTML pages can be generated via make docs, which runs dex script --outfmt html and outputs pages to a doc/ directory:

$ make docs
stack exec ... dex -- script examples/mandelbrot.dx --outfmt html > doc/examples/mandelbrot.html
stack exec ... dex -- script examples/pi.dx --outfmt html > doc/examples/pi.html

srush · 2021-01-20T20:23:01Z

Nice. @dan-zheng maybe it should just push to gh-pages directly? I just figured out you could do that that the other day: https://github.com/harvardnlp/pytorch-struct/blob/master/.github/workflows/docs.yml

dan-zheng · 2021-01-20T20:29:34Z

@dan-zheng maybe it should just push to gh-pages directly?

Thanks @srush! I filed #477 to track automated GitHub Pages updates, it should be a straightforward improvement!

There's one consideration on the issue that I'd like to discuss with folks before moving forward with automation.

.github/workflows/ci.yaml

…o tutorialfinal

apaszke · 2021-01-21T13:35:58Z

examples/tutorial.dx

@@ -292,9 +292,9 @@ def tableAdd' (x : m => n => Float32) (y : m => n => Float32) : m => n => Float3

 ' To run this section, move the following binary files to examples:

-' `wget https://github.com/srush/learns-dex/raw/main/mnist.bin`
+' `wget http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz; gunzip t10k-images-idx3-ubyte.gz`


Isn't unzip a bit more standard for this? I think it's equivalent to gunzip on most Linuxes, but also works on BSD-based systems (macOS).

Hmm, is that true? unzip doesn't seem to work for me linux for .gz files. Also gunzip seems to install fine in the macOS CI test. But I don't have a mac to test on. Does unzip work there?

I confirm the following works on macOS:

$ wget http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz; gunzip t10k-images-idx3-ubyte.gz # Produces unzipped t10k-images-idx3-ubyte, 7.5M.

Hope this helps!

apaszke · 2021-01-25T10:35:31Z

examples/tutorial.dx

+interface Add a
+  add : a -> a -> a
+  sub : a -> a -> a
+  zero : a


Hmmm this is a good question... I honestly don't know whether we can do that or not. The worst part is that the MNIST download page doesn't even mention a license, so I'm a bit reluctant to do so 😕 Could we use a different dataset by any chance?

examples/tutorial.dx

srush · 2021-01-25T13:58:54Z

Hmmm this is a good question... I honestly don't know whether we can do that or not. The worst part is that the MNIST download page doesn't even mention a license, so I'm a bit reluctant to do so confused Could we use a different dataset by any chance?

Keras claims: License: Yann LeCun and Corinna Cortes hold the copyright of MNIST dataset, which is a derivative work from original NIST datasets. MNIST dataset is made available under the terms of the Creative Commons Attribution-Share Alike 3.0 license.

Now we are downloading though, so is there an issue? For whatever reason this dataset is always the dataset these are written on.

(For some reason I can't reply to this inline. )

srush · 2021-01-27T13:42:11Z

Ping on this one. Think it is ready to go unless I'm missing something.

apaszke · 2021-02-01T16:00:59Z

Ugh, sorry for letting this slip. Yeah the only reason for the delay were the licensing issues. The Keras note might be good enough, but what if we switched to Fashion MNIST? This one is released under MIT, so there would definitely be no issue with downloading it in CI.

…o tutorialfinal

srush · 2021-02-01T18:26:40Z

Updated. Haha Fashion MNist is fun. It requires literally no changes.

apaszke · 2021-02-01T18:28:40Z

Yeah, that's why I suggested it! LGTM!

srush added 2 commits January 20, 2021 08:54

Small change to update syntax and fix indices

bd1acef

Small changes to type classes

fb5a33b

google-cla bot added the cla: yes label Jan 20, 2021

dan-zheng approved these changes Jan 20, 2021

View reviewed changes

srush added 2 commits January 20, 2021 14:26

add to CI

1c95ce7

download binary files

b4bc3c8

dan-zheng mentioned this pull request Jan 20, 2021

Automate GitHub Pages generation #477

Closed

1 task

apaszke reviewed Jan 21, 2021

View reviewed changes

.github/workflows/ci.yaml Outdated Show resolved Hide resolved

update github actions and tutorial download

b8abc1c

srush changed the title ~~Hotfixes : Tutorial update to latest~~ Tutorial bugfixes and CI integration Jan 21, 2021

dan-zheng and others added 3 commits January 21, 2021 08:00

Remove extraneous whitespace.

14554a2

update the tests for new mnist files

6971098

Merge branch 'tutorialfinal' of https://github.com/srush/dex-lang int…

71f3d4b

…o tutorialfinal

srush mentioned this pull request Jan 21, 2021

Add examples/nn to the CI #479

Open

apaszke reviewed Jan 21, 2021

View reviewed changes

apaszke reviewed Jan 25, 2021

View reviewed changes

Fix errors in tutorial

fe862c4

Merge branch 'main' into tutorialfinal

7efdbd2

srush added 3 commits February 1, 2021 13:23

update to fashion mnist

5e9bab4

Merge branch 'tutorialfinal' of https://github.com/srush/dex-lang int…

7e87f4c

…o tutorialfinal

Merge branch 'main' into tutorialfinal

438038b

apaszke approved these changes Feb 1, 2021

View reviewed changes

apaszke merged commit 540f1a1 into google-research:main Feb 1, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Tutorial bugfixes and CI integration #472

Tutorial bugfixes and CI integration #472

srush commented Jan 20, 2021

oxinabox commented Jan 20, 2021 •

edited

Loading

dan-zheng left a comment

dan-zheng Jan 20, 2021 •

edited

Loading

srush Jan 20, 2021

apaszke Jan 20, 2021

srush Jan 20, 2021

apaszke Jan 25, 2021

srush commented Jan 20, 2021

dan-zheng commented Jan 20, 2021 •

edited

Loading

srush commented Jan 20, 2021

dan-zheng commented Jan 20, 2021 •

edited

Loading

srush commented Jan 20, 2021 •

edited

Loading

oxinabox commented Jan 20, 2021

srush commented Jan 20, 2021

dan-zheng commented Jan 20, 2021 •

edited

Loading

srush commented Jan 20, 2021

dan-zheng commented Jan 20, 2021

apaszke Jan 21, 2021

srush Jan 21, 2021

dan-zheng Jan 21, 2021

apaszke Jan 25, 2021

srush commented Jan 25, 2021

srush commented Jan 27, 2021

apaszke commented Feb 1, 2021

srush commented Feb 1, 2021

apaszke commented Feb 1, 2021

Tutorial bugfixes and CI integration #472

Tutorial bugfixes and CI integration #472

Conversation

srush commented Jan 20, 2021

oxinabox commented Jan 20, 2021 • edited Loading

dan-zheng left a comment

Choose a reason for hiding this comment

dan-zheng Jan 20, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

srush commented Jan 20, 2021

dan-zheng commented Jan 20, 2021 • edited Loading

srush commented Jan 20, 2021

dan-zheng commented Jan 20, 2021 • edited Loading

srush commented Jan 20, 2021 • edited Loading

oxinabox commented Jan 20, 2021

srush commented Jan 20, 2021

dan-zheng commented Jan 20, 2021 • edited Loading

srush commented Jan 20, 2021

dan-zheng commented Jan 20, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

srush commented Jan 25, 2021

srush commented Jan 27, 2021

apaszke commented Feb 1, 2021

srush commented Feb 1, 2021

apaszke commented Feb 1, 2021

oxinabox commented Jan 20, 2021 •

edited

Loading

dan-zheng Jan 20, 2021 •

edited

Loading

dan-zheng commented Jan 20, 2021 •

edited

Loading

dan-zheng commented Jan 20, 2021 •

edited

Loading

srush commented Jan 20, 2021 •

edited

Loading

dan-zheng commented Jan 20, 2021 •

edited

Loading