-
Notifications
You must be signed in to change notification settings - Fork 26
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
*: prerelease fixes #688
*: prerelease fixes #688
Conversation
c4d12ed
to
c86b033
Compare
01313bd
to
c011a11
Compare
c011a11
to
91a2844
Compare
91a2844
to
e86b86d
Compare
847d130
to
28c7a2e
Compare
for the async iter every batch vs every round: does it affect training speed or not much? (thinking of the cases where local iterations might be super fast but communication might be slow, which might often be the case) |
hum, not from what I remember. we didn't actively test with slow communications but it didn't seem like a high time cost in comparaison to the training (sub-seconds vs whole seconds). of course, if the network is very flaky, it might come as an issue. |
28c7a2e
to
9ac364d
Compare
ho good catch, I mixed two tailwindcss classes, fixed in the new commits.
gladly, web UI is not really my strongest suit :) |
It's not mine either but I unfortunately have strong opinions on the matter... If that's fine with you I'll make changes in a follow-up PR rather than this one |
a4cd86f
to
a730240
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Great work!! Mostly minor comments but a few important ones:
- the preprocessing time measurement are now wrong
- the chart animations are jerky
5a2f781
to
52050a6
Compare
Co-authored-by: Julien Vignoud <[email protected]>
Co-authored-by: Julien Vignoud <[email protected]>
0b51367
to
b90b193
Compare
on the way to v3, we have some stuff to iron out, tracked in #686.
taking over some from there.
{Trainer,Model}.fit
to return async iterators to follow progression of every part of processing. allows for more reactive UI as it refreshes on every batch rather on every round.