a quiet week this week. I did some work on streamlining our deploy of the similarity code in AB
finished up an old PR and got that ready to merge
spent some time reading through iliekcomputers' PR for artist recommendation, which he managed to get working before I published my version of the same thing (thanks)
that's it. Freso?
Freso
o/
yvanzo
never seen Freso second after alastairp first before...
Freso
Also another quiet week from me. I had a fever and a headache most of last week, with some other cold symptoms coming back and forth. Still have smell and taste, so hoping it isn’t covid. Other than that, summering and skulking about.
yvanzo: There’s a first for everything, eh. :)
zas: Go!
Sophist-UK joined the channel
alastairp
what's the probability of reviews being given in a given order?
yvanzo
1?
Freso
Or… yvanzo, go!
yvanzo
hi!
Mr_Monkey
…something somthing about infinite monkeys…
yvanzo
Last week I mostly tried to catch up with PR reviews, that’s about it.
Go Mr_Monkey!
Mr_Monkey
Hi!
Freso
alastairp: I try to randomise the list every week, but some people do seem to have biases in who they call next. :) (My choice to randomise the list was exactly to counter any bias I might have in who to call first.)
Mr_Monkey
Slow week for me too. It's starting to get hot in here (inset song snippet).
ruaok
so better take off all your fur!!
Mr_Monkey
I did finish my PRs for LB and BB pagination, but still finding some small tweaks (thanks to CatQuest too for finding one)
CatQuest
:)
ferbncode joined the channel
Mr_Monkey
I also worked a bit on improving ElasticSearch configuration; it's going to take some time to find out what to improve and how.
alastairp
reosarevok: bravo
Mr_Monkey
Also a bit of PR reviews
And some repo upkeep
That's it! prabal go !
Freso
Or… ruaok, go!
ruaok
k
last week was doing some recommendation data set hosting and writing tests for the various components.
and then it was about the coming new lockdown in BCN, so I decided to high tail it to cooler climate and chill up north for a bit.
the rest of the week wasn't as productive, dealing with lockdown in bcn and communicating with friends and so.
I did some timescale test fixing, dump improvements and dump testing.
we finally got the timescale DB synched and receiving all the data -- the influx server and timescale server are now running in parallel.
today I started the process of planning the migration -- in under 48 hours we'll actually migrate to timescale in production.
shivam-kapila
Woo!!
Mr_Monkey
!m
!m ruaok
BrainzBot
You're doing good work, ruaok!
ruaok
shortly thereafter we're going to have another release to already make some improvement we already know about and have implemented, thanks to Mr_Monkey's help.
it will be good to put this project to bed -- timescale is a game changer I feel.
(People still up: bitmap, reosarevok, diru1100, jmp_music, _lucifer, shivam-kapila, CatQuest, Cyna[m] – anyone else who wants to give review, let me know ASAP!)
ishaanshah
Also worked on optimising the query for calculating daily activity stats
Will try to complete the PR by tomorrow
Thats it for me, fin!
bitmap: next?
bitmap
hey
last week I did a bit of code review and worked more on bug fixes to the incremental sitemaps to handle deleted rows better
and spent more time bug-fixing a branch to change how we pass data to the template renderer, since that's been causing issues lately that require hacks to work around
unrelatedly, also wrote a SQL function that can generate Flow types for a particular edit type, which will help us work around deficiencies there a bit more safely and perhaps clean up the data later
reosarevok
So. Many. Hacks.
bitmap
not much interesting to report on otherwise. I submitted the PR to speed up Data::Release::find_by_artist last Monday and it was deployed by reosarevok yesterday, and seems to have helped with the load issues on floyd a lot
reosarevok: yupp
but there there are other slow queries we still need to work on
anyway, fin
reosarevok: go!
reosarevok
Hi!
Last week I chilled for the first half, then worked on more React conversion, plus small bugfixes, plus released a hotfix as bitmap mentioned
Working on more React bits right now, will keep at that for a while I think
Fin
CatQuest mentioned he might not be around, but just in case: CatQuest, around?
If not, _lucifer can go
CatQuest
Moja opinia jest dziś w języku polskim
bo Pewna Osoba był dupkiem do mnie, gdy byłem zdenerwowany i próbowali prosząc o pomoc dla czegoś.
Nie mam nic innego dzisiaj.
następny może być diru110
diru1100
wow, that was fast :)
shivam-kapila
(I didnt understand a single word)
diru1100
Hello All !!
I only evaluated the model till now, last week I worked on predict function, tested it with different inputs and got decent results.
Will be making regular PRs for review this week and make good progress in the project.
Side note: I got shortlisted for online hackathon for jobs conducted by JP morganchase, excited about that
Thank you
jmp_music: ?
yvanzo
!m diru1100
BrainzBot
You're doing good work, diru1100!
jmp_music
Hi everyone!
Freso
(People still up: _lucifer, shivam-kapila, Cyna[m] – anyone else who wants to give review, let me know ASAP!)
jmp_music
Busy last week for me. I completed the ML data shuffling and k-fold training in the same way gaia does. I also refactored some code in order to be easily expandable to applying any ML model from sklearn later on.
I completed also the classification task manager for each task of the ML (similar to gaia). Now a GridSearch algorithm runs and exports the best model among the other best models that occurred from each preprocessing method.
The best model parameters are saved in a JSON file locally and are then called from evaluation via a cross-validation training kind. Furthermore, the best parameters from each of the other tasks are also saved.
shivam-kapila
(diru1100: I have seniors at JPM in case you need any help)
mostly everything along with these is redesigned but these are the mocks I have now
jmp_music
Finally, the mean accuracies from each fold, the standard deviation, confusion matrix, and the classification report are exported in separate files locally to disk too. Additionally, some plots for each preprocessing step (normalize, gaussianize, etc.) are exported and saved locally in order for someone to have the chance to study the data for
research purposes. A distribution plot of the accuracy values from each fold is also created.
that's all for now
_lucifer?
alastairp
jmp_music: nice work! can't wait to see it tomorrow
shivam-kapila
SomalRudra: thanks I will take a look over them
_lucifer
Hi! all
I worked on adding pagination to the search results last week
Freso
(Only shivam-kapila and Cyna[m] left - last call for anyone else who want to give review!)
jmp_music
alastairp: thanks!
_lucifer
I also improved a few unit tests
thats it for me
shivam-kapila: next
diru1100
(shivam-kapila : Thanks, will ask you later :) )
_lucifer
SomalRudra: I have fixed the UI bug. You had called the wrong overloaded method for onCreate. I'll push that a few other fixes so you can apply them
SomalRudra
_lucifer: Thanks! can you push them now?
_lucifer
!m diru1100
BrainzBot
You're doing good work, diru1100!
_lucifer
yes
TOPIC: MetaBrainz Community and Development channel | MusicBrainz non-development: #musicbrainz | GSoC students start here: https://bit.ly/2NUaxgb | Channel is logged; see https://musicbrainz.org/doc/IRC for details | Meeting agenda: Reviews, musicbrainz-android (_lucifer)
shivam-kapila
Oops I didnt see the call
Hello everyone
Last week wasnt productive for LB work. Had a lot fo coolege projects, reports and assignments going on
MajorLurker joined the channel
I tried to work on LB for all the time I was free in between and managed to finish the right part of the redesign
Will make PRs this week and I expect the redesign to finish by next week
Apparently I am a lil lagging behind but will catch up soon.
Thats it I guess. Thank you everyone
Cyna[m]: next?
Freso
No Cyna[m]?
And also no reemerge zas or prabal.
*reemerged
So I guess that concludes reviews. Thank you all who gave one. :)
One more item on today’s agenda:
_lucifer: musicbrainz-android
_lucifer
sorry for the short notice!
i wanted to say that musicbrainz android has been progressing well
and its time we think of a name for it
we have been using musicbrainz and picard in it
Now, the app supports tagging, exploring musicbrainz and submitting listens
MajorLurker has quit
Freso
I personally don’t like the idea of using Picard for it, since that makes it easily confusing with desktop Picard.
_lucifer
in future it may as well support data submission to acousticbainz