#metabrainz

oh, sure!

of course, if you're evaluating on everything that you trained on I would hope that you get 100% ;)

this is the evaluation of fitting and predicting to the whole amount of thedanceability's data collection

data collection --> dataset

yeah, understood

I'm trying now to see what I can do with gaia

you said before: `replace this with just a single item pointing to the history file`

you mean something like that, right?

i made a code directory inside the data dir of AB, and I put the file there

right

and by saying `add anything you want to the 2 items under mergeValues`

so, `svm_models: [ /data/svm_models/voice_instrumental.history, /code/data/jmp_results_danceability ]`

yes

in fact, you can delete the whole mergeValues section

make sure you get the path correct for your second item

the `acousticbrainz-server` folder on your computer is mapped to /code inside the container

oh, ok!

but sure, you could also make a new folder in /data, but keep in mind that when you close the session (and shut down the container) that data will disappear

and the json value I want to predict, how AB-server can read it?

the low-level file?

yes

just download it and put it in your `acousticbrainz-server` folder somewhere, and then when you are inside the container you can access it as `/code/wherever/you/put/it`

Thanks! And finally, which is the `[profilefile]`

?

jmp_music_: profilefile is that file that you just made

the configuration file

ishaanshah: hey

Hi

How are you

?

good, good. I'm gonna deploy the daily activity backend today.

i just want to pull in the landing page changes as well.

Cool, I worked on the frontend today

The line graph is done

But very messy right now

messy as in it doesn't look good, or as in the code is messy?

doesn't look good

[listenbrainz-server] paramsingh merged pull request #966 (master…daily_activity_backend): LB-658: Add "Daily Activity" statistics (Spark + DB + API) https://github.com/metabrainz/listenbrainz-serv...

LB-658: Add "Daily Activity" statistics (Spark + DB + API) https://tickets.metabrainz.org/browse/LB-658

but thats because of less data

hmm.

I'll trigger a calculation today, let's see how it looks then

Lets check again after calculation

I have a feelling it would be better for others

yep

I didn't work on fine tuning the graph much

Like toggleable legend and stuff

yeah, makes sense. let's see how it looks first

if we have to switch to bar then it would have gone waste thats why

yeah exactly

I came across this site today

I like the first graph they have used

I think we should replace the stream graph in sitewide entities with this one

sure, it looks cleaner1

Btw the whole site is built using nivo

!

that's pretty cool

cool, just wanted to ask your opinion on it

Let me know when you calculation is complete

I will be around for some more time, I will post some screenshots

you -> the

sure, ok. i want to merge a couple of pull requests that are building. i'll keep you updated.

Cool

[listenbrainz-server] paramsingh merged pull request #977 (master…remove-queue-stats): Remove the broken stats from the status page https://github.com/metabrainz/listenbrainz-serv...

i don't have much else, do you?

Nah nothing else

cool!

[listenbrainz-server] paramsingh merged pull request #975 (master…param/landing-page-text): Improve landing page text https://github.com/metabrainz/listenbrainz-serv...

@alastairp: just ran the extractor for the same track

cool. 99.9% looks very close to the 100% of acousticbrainz

that's right!

I have to check my code again. maybe i 've done something wrong when i call the model

i.e. in the predictions

sure thing

have you tried to make a model specifically with the C/gamma values from gaia?

I 'll proceed with this ones

one*

I'll work on it tomorrow

no problem

thanks

I have already made a script that loads gaia's params

so it will not take long to me to test it

will you be available tomorrow?

to discuss any results?

ishaanshah: you up?

yep

So I wanted to dicuss about the graph on the my listens page

What do you think we can show there

Listening activity over past 30 days?

Thats what I was gonna say

I dont want to make a bar graph

We can have a line graph without the points

You know for the sake of variety

Yep I think the same

Yeah we are in the same boat

Do we have something like past 30 days listen activity in backend right now. I mean a spark job for that

the graph you have shown in the mockup seems good already

jmp_music_: I'm not sure if I'll be around tomorrow, I have some things to do. however, send me some messages and I'll respond if I have time

@alastairp: Ok! Thanks again!

Yeah. Since the whole dashboard focuses on smoothness I would like this one to be smooth

[listenbrainz-server] release v-2020-07-23.0 has been published by release-drafter[bot]: https://github.com/metabrainz/listenbrainz-serv...

Yep, we can do that

So its basically the last month plus the month till now

Yeah

Cool we have the data

But we should skip the last month

From what I was thinking is: say today is 23 july so we show from june 24 to july 23

What do you think

Yep, that makes sense

shivam-kapila: what would happen if stats were not calculated for a month, due to some downtime?

we can have the same message that we have on stats page?

For a month? Arent the stats jobs triggered regularly

Yeah but there are bugs some times

they are, but part of engineering is trying to handle failure cases gracefully

The pipeline is not very stable right now

TBH I was gonna just tag you because of your wider sense of figuring things out

Yeah I understand

we solved the issue for stats by using the latest listen timestamp in spark for all calculations.

Yeah I remember seeing that in the PR

on the frontend we show data based on that, but we also explicitly state the time range (Jan 23 - Jan 30, for example)

iliekcomputers: What I had in mind is, we show slice the array to show past 30 days

Yeah show the date expicitely

right, that makes sense.

expicitly

Actually the graph comes just after user's current listen count etc

ishaanshah: https://api.listenbrainz.org/1/stats/user/iliek...

oh dammit

error on sentry

i didn't do the schema change lol

So thats I was considering the last 30 days

Whoops :D

shivam-kapila: Instead of past 30 days, lets just show the time range that we are calulating it for below

Thag makes it somewhat like a line copy of a bar graph already in charts section if I am not wrong

That*

iliekcomputers: iiuc the edge case you mentioned is somwhat like a user has no listens in the past 30 days. Am I right?

I mean, show 24 june to 23 july

I understand that

not really, the edge case here is that you might not have the stats for the past 30 days

the user could be submitting data and stats not getting calculaed