Training Planning Meeting
Meeting 2020-03-09
Present: Kilian Lieret, Sam Meehan, Clemens Lange, David Lange, Mark Neubauer, Adam Parker, Mason Louis Proffitt, Eduardo Rodrigues, Henry Schreiner, Dan Katz
Apologies: Graeme (meeting clash), Sudhir (ill)
pyHEP and HSF Training
-
Eduardo’s talk was recorded
-
pyHEP uses gitter channels for a lot of communication
-
Very low overlap between SWC and pyHEP conference audience (very different levels, pyHEP becomes relevant a year later)
-
pyHEP training focuses on specific python packages: awkward-array, uproot, etc. -- it does not focus on training the basics of python and vcs/git
-
Important to at least make SWC attendees aware of pyHEP material (e.g. slides with links in the last session)
-
In the past (last SWC), the last day covered material from pyHEP
-
-
Style of pyHEP trainings: Jupyter notebook (hosted de-centrally)
-
Henry: Jupyterbook as a new training format: Collections of notebooks together with table of contents rendered into webpage https://jupyterbook.org
-
somewhere in between plain juptyer notebooks (where participants only shift-enter everything) and very verbose carpentry-style websites
-
Examples:
-
https://cranmer.github.io/madminer-tutorial/intro
-
https://henryiii.github.io/compclass-book/week1/0_IntroductionAndLogin
-
-
Jupyterbook might be bridge between SWC-style HSF/training modules and HSP/pyhep training
-
Sam: Might do a small pyhep/training bootcamp, where we create a new jupyter-book style module
-
Carpentry-style trainings are well-tested now, this would be a first, so need to test this
-
-
-
HSF-pyHEP is curious about our experiences with software training: How can training sessions be made more didactic and a better experience for students.
-
It would be nice to provide a similar look and feeling for both pyHEP and HSF material: Students profit from perceived consistency and homogeneous feeling
-
Might do a combined advanced-SWC + pyHEP training workshop?
-
https://hsf-training.github.io/analysis-essentials/git/03-create.html : LHCb Starterkit material, ALICE etc.
ATLAS/CMS Analysis preservation post-mortem
-
Slides: https://docs.google.com/presentation/d/1saqe1J0VHPd0x6dEjk5LBLf53camnW2U-QA02y0rjPQ/edit#slide=id.p
-
Use “awesome H(tautau) analysis” as playground also for merge conflicts etc.?
-
Organizing discussion with theorists was difficult
-
Dinner was a nice addition (IRISH-HEP covered food, drinks: 8 franks per person, ): Makes it an "official" event, such that people who sign out actually show up. Good for networking and fostering real connections. Makes it feel like a real conference (not just another training session)
-
Criticism: Rooms (too cramped), too small screens?
-
Length of workshop:
-
Sam: 3 full days are the maximum number of days for training that participants can handle
-
CMS has 1 week long analysis starterkits (--> Clemens)
-
Most students can’t handle 8h of pure information
-
Important to have enough breaks
-
Depends on how much material builds on itself/cascades; if workshop consists of lots of orthogonal modules, it can be longer
-
-
Might need to advertise HSF and training material etc. more (lot’s of people aren’t aware of it). Eduardo: Important to have contacts in the different collaborations and advertise the material on the mailing lists. `
-
Future instructors: Lot’s of participants want to teach themselves; collected their emails
-
VIP workshop for seniors (most training workshops aimed at young people)
-
Have small numbers of seniors also in “normal” trainings, but could increase that by specifically targeting seniors
-
Might be uncomfortable for seniors to ask “stupid questions” openly? → survey
-
Might do survey:
-
E.g.: How comfortable are you with X?
-
Eduardo made a large LHCb survey. Specifically stress that also the opinion of the seniors is requested here
-
-
-
Clemens: Would ideally need two large screens (one for vidyo webcasting for remote participants and for the slides)
Software Carpentry @CERN
-
For the Software Carpentry event we have said “postponed”, which I think means that when we find another date we keep the registrations as they are, but we confirm with people their continued interest in attending.
-
I think there’s no point at all in trying to reschedule in advance of June; I had a quick look at the Indico calendar and Kjell Johnsen is fairly busy all of June; virtually unused in July (could tentatively book 7-10 July…?)
-
Can’t schedule anything, before CERN revises Covid-19 rules.
-
Do we want to announce support for self-study in the old slot?
-
Status of the lessons:
-
Software Carpentry core lessons are complete;
-
material on ML and numpy it’s less clear that this exists in a shape where it can be used;
-
Mason’s uproot+awkward is in decent shape AFAICR.
-
-
Life support (chat etc.):
-
If yes, how do we do it? Mattermost channel? (Then people could break out into a permanent Vidyo room for the event if needed?)
-
-
ESCAPE Summer School
-
(22 June - 1 July), https://indico.in2p3.fr/event/20306/
-
Thomas Vuillaume is organising: thomas.vuillaume@lapp.in2p3.fr
-
Mason agreed to re-teach his uproot+awkward tutorial as a HEP domain specific part (thanks!)
-
There is 12 hours scheduled for Machine Learning topics in the school; we are trying to identify suitable material and tutors for this
-
Kevin Nelson (Michigan) did volunteer to help (thanks!)
-
Discussed with Martino the material (as he was going to do a scikit-learn session for us); carpentries are developing something (https://github.com/machinelearningcarpentry/machine-learning-novice) but it’s little more than a skeleton
-
Do people have knowledge of other material we could use?
-
Would people like to get involved in developing material like this in the next few months?
-
-
Are other people interested in helping out at the event itself? Let Graeme know.
-
MISC
-
Adam has time to work on modules for uproot, etc.