►
From YouTube: IncEng MLOps Update - 2022-05-11
Description
Notebooks seem to have caused and acceleration on the growth of comments on Notebooks, Finding on MLFlow, general blockers
All Updates: https://gitlab.com/gitlab-org/incubation-engineering/mlops/meta/-/issues/16
This Update: https://gitlab.com/gitlab-org/incubation-engineering/mlops/meta/-/issues/54
A
Hello,
everyone
and
welcome
to
another
update
for
incubation
engineer.
Mlaps
today
is
may
11th
and
we're
going
to
talk
about
usage
and
progress
on
the
features.
So,
first
of
all,
let's
start
with
jupyter
notebook
devs.
We
are
doing
a
lot
of
maintenance
on
the
work
that
we
did
so
far,
so
fixing
bugs
acting
on
user
feedback
and
to
make
it
a
little
bit
more
transparent.
I
created
a
niche
and
epic
just
to
collect
all
the
bugs
all
the
user
feedback,
all
the
user
ideas
that
we
are
receiving.
A
So
this
is
taking
a
lot
of
my
time.
So
this
I
am
blocked
on
adding
new
features
because
of
capacity
most
of
my
energy
is
going
to
bug,
fixing
and
making
it
a
good
product,
but
we
are
working
now
on
graduating
this
project
into
a
different
team,
because
it's
it's
just
most
of
the
time,
I'm
using
into
maintenance
and
that's
not
the
goal
of
commission
engineer.
We
want
to
test
out
new
ideas
as
well,
so
right
now,
new
features
are
blocked.
We
don't!
A
Next,
we
don't
have
any
expectation
when
the
toggle
motif
are
going
to
be
released
into
the
mr,
nor
renderable
diff.
We
are
working
on
that,
but
it's
it's
hard
to
project
a
a
finish,
a
timeline
because
of
the
maintenance
that
has
been
going
on
on
the
mail
flow
side.
We
start
doing
a
little
bit
of
user
research.
Some
of
key
users.
A
Interesting
findings
is
that
ml
flow
or
the
model
registry,
which
is
a
registry
for
machine
learning
models,
is
not
used
only
on
packaging
or
on
deployment,
but
data
scientists
are
using
the
great
phase
already
to
start
off,
organizing
their
runs
their
experiments,
so
it
needs
to
be
available
earlier
than
just
after
ci
starts
and
speaking
with
a
platform
engineer,
one
of
the
biggest
pain
points
is
auth
authorization,
so
ml
flow
out
of
the
box
doesn't
provide
any
sort
of
authorization.
A
It
expects
the
the
user
to
integrate
this
into
their
own
platform
or
in
their
own
infrastructure.
So
it's
a
little
bit
complicated
to
add
authorization
on
top,
and
I
think
this
can
be
a
really
good
opportunity
for
gitlab.
If
we
put
a
model
registry
behind
a
repository,
then
we
would
get
a
poster
project
because
then
we
would
get
authorization
for
free
and
authentication
for
free.
A
We,
I
have
an
issue
where
I'm
collecting
all
the
feedback
that
I'm
getting
from
the
user
research
and
just
reading
user
feedback
on
the
slack
channels
and
whatnot
that
will
go
into
that
specific
issue
over
here
on
usage.
So
I
did
a
little
bit
more
analysis
on
the
usage
of
notebook
diffs
and
by
usage
I
mean
number
of
comments
and
the
growth
that
has
been
going
on
over
years.
So
what
you
see
here
on
the
green
line
is
the
green
vertical
line.
A
It
was
the
release
of
14.5,
that
was
the
release
of
the
jupiter
notebook
devs
render
tiffs
or
rather
cleaner,
divs
that
we
we
did
and
the
most
interesting
one
is:
let's
go
back,
it's
the
growth
acceleration.
So
we
see
that
after
14.5,
the
the
the
the
growth
of
comments
on
the
books
had
been
decreasing,
but
after
the
release
of
14.5,
we
see
a
big
big
growth.
Well,
growth
on
the
growth.
A
It's
it
accelerated
the
growth
on
enjoyment,
book
devs,
and
this
is
a
little
bit
sustained
after
a
while
this
is
year
over
year.
So
it's
very
interesting
to
see
the
data
just
point
that
this
is
on
the
gitlab.com
website.
It
is,
does
not
take
into
account
the
data
from
self-managed
customers,
but
this
is
really
great,
so
I
am
receiving
this.
A
This
is
great
to
see
this
is
in
quantitative
ways,
but
also
in
qualitative
ways
I'm
receiving
I'm
receiving
every
week
some
update
from
users
that
they
started
doing
code
reviews
that
they
would
not
do
that
before.
Because
of
this
feature-
and
this
is
really
really
great-
that's
what
I
had
for
today,
thanks
for
all
watching
see
you
next
time.