Closed sanjayankur31 closed 2 years ago
We have proposed a "proper" tutorial on GeNN on site. Not sure whether that is a reason to or not to run a satellite.
Thanks, that sounds good. I don't see an issue with tutorials at both myself---given that the "proper" tutorials would be limited to people that can make it to Melbourne only. We can leave this open to decide at a later date---depending on what you all (the devs) and what others in the WG think about it.
Agreed, as we are preparing the material anyway, holding it online as a satellite would only involve the time spent to actually hold it. So could be good value for money for us @neworderofjamie
@tnowotny : Do you think we should run it past the OCNS board to confirm there is no potential conflict in having the same sessions being offered via the free online software tutorials and the paid on-site tutorials? We seem to have a few such cases. (There is naturally a difference between online vs on-site, but wondering if it is better to inform and take approval beforehand).
I would think it would be a nice gesture to ask them - though I don't think there is a problem.
Hi @tnowotny @neworderofjamie
In general, we think we will have three categories of sessions:
We leave the exact duration up to the different organisers. With the sessions being online, we are extremely flexible :)
We are reaching out to all the organisers now for initial bits of information to help us get started on the schedule. Could you please tell us by the 20th (end of next week):
If you can please set up a web-page for the session and give us the link, we will put that on the schedule too. This last one is not necessary just yet. We can also put a post on the Software WG website if you prefer.
We are requesting organisers to record these sessions if possible, and we will work with the INCF folks to get the recordings put up on the INCF training space.
Ok, sorry for the delay. Here is my/our take:
what is the expected duration of the session? 3 hours
what is the preferred start time of the session (in UTC)? no strong preference, but best between 10:00 and 17:00.
are there days that you cannot hold the session in the June 27--July 1 interval? 27 June preferred (then both @tnowotny and @neworderofjamie can lead), otherwise any day with @tnowotny as lead.
would you like us to set up zoom meeting rooms for you (for organisers that may not have access to institutional zoom etc. subscriptions)? We have our own institutional zoom subscription and will set up ourselves.
Thanks very much. We'll have a draft schedule up as soon as we have everyone's responses.
Hi @tnowotny @neworderofjamie , I've published the schedule on our website here, and we'll publish the Zoom links in a different document whose link we'll send out in the registration confirmation e-mail.
Could you please set up a zoom meeting and a web page (optional) and send me the links via e-mail at ankursinha AT fedoraproject DOT org
when you have some time to spare please?
I'll post regular updates on number of registrations and who intends to attend what so all tutors can prepare accordingly on this ticket here---please keep an eye on it: #106
@sanjayankur31 , the link for further information for this tutorial will be https://genn-team.github.io/tutorials.html
@tnowotny @neworderofjamie Great to see that there are GeNN tutorials which can run in notebooks on Collab.
@sanjayankur31 and I have been looking at installing a number of the other simulators for the satellite tutorials on https://www.v2.opensourcebrain.org, so that presenters & attendees could run the notebooks, scripts etc. in its Jupyterlab instance. Any modifications/files generated in the session would be saved in workspaces for reopening later on OSBv2.
I've not tried GeNN yet on here, as the depencence on CUDA would probably complicate things, not least since we're not paying for any GPU nodes. How well does GeNN work in a pure CPU environment (even if it's slower than usual)? Some solution might be possible for installing/running on OSBv2, but probably it's best of you still plan for Google Collab for the CNS22 session.
GeNN works fine on CPU and, for small models, it's pretty fast but I suspect, without GPU, the mnist mushroom body tutorial we're going to use for the second half of the session will be too slow.
I think the tutorial was alright. Ankur, what do we do with the recording? Also, should I record an Intro separately or should we let that lie?
If you've received the recording link already, please note it here and I'll add it to both the public schedule and our private doc.
About the intro---it's up to you. If you think the session feels incomplete without it, it'll be worth recording---and we'll add links to both recordings everywhere?
Ok, here is the video: https://sussex.box.com/s/9kopk3brmepniumshe73524jk2rjxspc And here is the ppt of the introduction: https://sussex.box.com/s/2r7rp7nhrjria2b6vag9jevldnkae1n9 I have not recorded the intro again - I think it is fine as is.
Thanks very much. I've added these links to the public and private docs now.
Recordings have been added to page, closing. I'll reach out to tutors over e-mail for feedback etc.
@tnowotny @neworderofjamie: it'll be great to have a satellite tutorial on GeNN---what do you think?