This repository serves as the editorial venue for ML Reproducibility Challenge 2020 accepted papers.
ReScience editorial process consists of porting the existing latex sources into ReScience template, and then adding necessary information relevant for final publication. Authors of RC2020 accepted papers are requested to submit their camera-ready papers as outlined in the steps below. The project structure of this repository is organized as:
<paper_citekey>
<-- OpenReview citekey of the accepted paper at RC2020. List of accepted papers inaccepted.bib
openreview
<-- location for the source files submitted to openreviewjournal
<-- location of ReScience C Template files
Some of you may have used Overleaf to write and submit your papers. We recommend to setup the Latex development pipeline locally in your system for full control over the editorial process. If you use a Mac, then use the following Homebrew formula:
brew cask install mactex
For Ubuntu/Debian:
apt-get install texlive
For Windows: Download Tex Live from source.
- Fork this repository and clone it locally (Note: clone recursively using
--recurse-submodules
or--recursive
flag) - Add the latex content of your paper in the
openreview
folder of your paper citekey. (You can check your correct folder from accepted.bib)
Now we need to organize your contents a bit so as to make life easier for the editors!
-
Make a new file
packages.tex
in youropenreview
folder, where you add all your import lines in Latex. (Reference). Import this file in your main tex file. -
Convert your main content in a
content.tex
file in theopenreview
folder, and import it in yourmain.tex
file. Feel free to re-organize your latex code in different files, as long as everything can be imported incontent.tex
. Yourmain.tex
file could look like following:\documentclass{article}[12pt] \input{packages.tex} \title{Your title} \author{Your authors} \begin{document} \maketitle \input{content.tex} \end{document}
-
[Optional] Test that the
openreview
compilation is working. Since the contents ofopenreview
is isolated, you can test this in the Overleaf template directly. -
Now, head over to the
journal
folder in your paper folder, and edit thecontent.tex
such that it is importing the correct../openreview/content.tex
file. Yourcontent.tex
file could look like this:\input{../openreview/content.tex}
Note: It is possible to face compilation issues in imports if you have extra packages/commands. In such case, feel free to copy your main tex contents in the
content.tex
appropriately. -
Edit the
article.tex
file to include yourpackages.tex
file before\begin{document}
. (Reference) -
If you have an appendix, import it in
article.tex
after\printbibliography
line. (Reference)
In the journal
folder you will find a metadata.yaml
file. This file is crucial to fill correctly for the journal to publish your paper correctly. Here is a reference metadata.yaml file from our last iteration.
- Add the
title
of your paper. The paper title should start with[Re]
. - Add the author information in
authors
. This is a yaml list, so just copy paste the following block for all your authors:
- name: John Doe
orcid: <your ORCID id>
email: <your email>
affiliations:
- Make sure to get your ORCID id.
- In the
affiliations
section, add the list of affiliations you have in your author list. At the same time, add thecode
of the affiliations in theauthors
section in a comma separated format. Add a*
for contact author. - In the
code
section, provide theurl
to your Github. - Again, in the
code
section, provide the link to Software Heritage Foundation archival link to your code inswh
. To get your SWH identifier, provide the link to your code in Software Heritage Archive, which will queue it for archival. You can follow the progress of your request, and once the status changes to "succeeded", you can click on the URL link to arrive at the archived page of your code. From this page, click on the "Permalinks" side-button, and copy the code beginning withswh
. Place this code in the yaml file underswh
. (Reference]) - In the
replication
section, add the necessary information about the paper you reproduced. - Provide your abstract in the
abstract
field (including double quotes) domain
should contain "ML Reproducibility Challenge 2020"- In
review
section, paste the url to the OpenReview forum of your paper. - Leave the fields
contributors
,dates
,article
andjournal
fields blank, as those will be populated by the Area Chairs / ReScience editors.
In the journal
folder of your paper, first remove the default metadata.tex
file. Then, run make
to compile. You should get the output article.pdf
to be nicely formatted in the ReScience styles.
- Please do not import
xcolor
in yourpackages.tex
, as rescience.cls already imports this package - Please remove the import of
fontenc
from yourpackages.tex
. ReScience uses custom fonts, which breaks when fontenc is loaded on top of it. (Check #4 for more discussion)
Once your compilation is ready, open a Pull Request (PR) to the repository with your files. Our Area Chairs will then review and accept the submission, and proceed towards obtaining the journal metadata.
While compiling you might face latex issues / errors. In such events, first inspect the log carefully for any offending blocks. If you are unsure, submit your PR anyway, and open an issue linking to your PR, where you attach the logs. To capture the log, pipe the make
output to a file and add it to the PR (make > build.log
). Our AC's will gladly help to you fix the issues! Also, do consult the codebase of our last iteration to narrow down any specific latex issues you are facing.
New: CI with Github Actions: We have now setup a continuous integration with Github Actions to run after every push and PR. Once you submit your PR, you can check the build logs from the PR right itself (currently we build all projects together, so search the log with the name of your folder). Thanks to Xavier Bouthillier for helping to set it up!
- To build all folders, run
./build_all.sh
- Koustuv Sinha (@koustuvsinha) Mail