►
From YouTube: CDF SIG MLOps Meeting 2020-06-04b
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
B
A
A
So,
just
very
quickly
with
the
status,
Andrea
and
I-
probably
don't
want
to
also
see
from
your
side
right
where
we
are
so
one
of
the
things
you
know
is,
let
me
find
out
if
we
have
it
running
yeah
so
right
now,
you
know
we
had
at
a
point
where
you
know
in.
If
you
look
at
this,
you
know
the
chip
flow
pipeline.
Let
me
see
if
I
have.
A
This
so
this
is
our
you
know,
just
the
starter.
You
know
Morton
pipeline,
which
we
test
the
right.
If
we
can
run
into
it,
or
at
least
you
know,
subset
of
the
end-to-end
just
to
you,
know,
train
and
store
model
and
then
deploy
using
model
machine
learning
and
this
one
right
now
you
know
we
are.
Let
me
try
to
see.
A
Okay,
so
that
is
getting
initialized
right,
but
so
the
there
is
one
which
we
have
run
before
and
by
the
way
this
is
you
know,
chip
flow
pipeline
is
running
on
top
of
Tecton
right.
So
the
goal
is,
you
know
exactly
to
get
the
same
behavior
right,
which
we
would
more.
So
this
is
the
one
right
which
we
ran
before.
Well,
the
current
one
is
getting
initialized.
A
The
fact
that
you
know
we
are
getting
the
logs,
as
is
you
know,
as
things
function,
the
next
of
the
things
which
we
are
working
on
is
you
know,
integrating
the
ml
metadata
and
the
artifact
artifact
is
almost
getting
very
close
ml
metadata
is
what
we
are
integrating
right,
so
I
think
so
far,
so
good
right
and
behind
the
scenes.
If
you
bypass
this
right.
Yes,
there
is
the
checked
on
UI
as
well
right.
A
So
if
you
see
you
know,
it's
triggering
on
check
bound
right
now
behind
the
scene
is
the
current
pipeline
which
I
launched,
but
you
can
actually,
you
know,
enter
through
this
world
right,
which
is
essentially
the
cue
flow
pipeline
world,
or
if
someone
wants
to
jump
directly
to
the
Techtron
experience
right,
they
can
get
it
from
that
side
because
that's
essentially
what's
happening
there.
So
so
far,
so
good
I
think
this
is.
A
You
know
now
showing
us
that
you
know
this
is
working
to
the
extent
we
wanted
as
part
of
the
MVP,
and
we
exactly
should
be
able
to
give
the
same
experience
right
to
the
chip
flow
pipeline.
Back-End,
we
checked
on
right
which,
which
the
folks
with
you
know
the
or
go
back
in
our
used
to
write.
So
the.
A
The
second
part,
which
is
you
know
the
looping
functionality
right.
These
are
the
two
big
blockers.
Obviously
you
know,
I
saw
both
of
them
were
in
some
sort
of
proposal
stage
right,
so
I
wanted
to
get
a
sense
from
you
like.
What
is
the
typical
checked
on
community
timeline?
Where
you
know
proposal
comes
in,
and
at
least
there
is
a
version
of
the
implementation
right,
even
if
in
a
branch
in
a
PR,
the
right
which
can
be
a
candidate
or
have
you
seen
any
of
these
two
actually
moving
further
from
the
design
proposal,
phase.
C
B
C
We
do
support
conditionals,
so
it's
it's
an
alpha
version
API,
but
we
have
some
form
of
conditionals
I,
don't
know
if
that
enough,
for
I
mean
it
should
be
enough
for
getting
basing
conditional
functionality
but
yeah.
So
we
may
need
to
clarify
that
offline
in
terms
of
designing
the
new,
the
beta
version
of
conditionals.
C
That's
really
just
started
so
just
like
the
proposal
was
was
made,
unfortunately
missed
the
last
the
last
meeting,
because
it
was
a
bank
holiday
for
me,
but
so
I'll
I
don't
have
the
latest
status
on
that,
but
I've
not
seen
any
any
update
3d
on
that.
Yet
so
usually
I
mean.
If
there
is
an
agreement,
it
pretty
depends
on
how
fast
the
PRS
are
contributed
and
maybe
review
bandwidth.
So
it's
it's
important
to
have
people
available
for
for
reviewing
the
changes
and
having
quick
feedback
loops.
C
C
C
A
Alpha
right,
but
from
the
perspective
right
so
when
we
are
taking,
for
example,
you
know
the
output
of
the
chip
flow
pipeline,
compiler
right,
we
are
producing
a
pipeline,
runs
back
right,
so
that's
when,
once
you
take
it
to
the
back
end.
If
something
is
in
alpha
versus
something
is
in
beta,
then
you
know:
how
do
you
leverage
you
know?
The
CLI
is
again,
you
need
to
take
multiple
of
these
fix.
A
C
Still
one
discussion
phase
right
now,
but
yes
in
terms
of
tooling
and
so
forth,
I
mean
that
the
CLI,
for
instance,
for
Tecton
works
currently
with
the
beta
version
of
the
API
for
pipelines
and
everything
else
and
the
and
the
other
version
of
pipeline
resources
and
conditionals.
So
you
don't
have
to
change
tools
to
use
different
version
of
API.
So
it's
it's
a
combination
that
is
supported
to
it.
Now
by
protecting
I
mean
the
main
disadvantage
of
the
beta.
C
The
alpha
version
is
that,
of
course,
changes
may
happen,
so
you
have
to
follow
those
changes
even
in
alpha.
What
we
do
we
before
we
make
changes.
We
usually
deprecated
them
in
one
release,
so
we
always
deprecated
them
in
one
release
and
remove
them
in
the
next
one.
So
you
don't.
You
have
less
to
release
time
to
cut
up
to
two
breaking
changes,
while
in
beta
it's
it's
quite
different.
It's
several
months
that
we
ensure
I
think
it's
nine
months
before
we
can.
C
When
we
started
predating
a
feature
before
we
can
actually
change
it
in
the
API.
So
that's
the
main
disadvantage.
Of
course,
one
thing
is
getting
conditions
to
beta.
The
other
thing
is
also
extending
the
functionality
of
conditions
and
improving
it.
There
are
some
drawbacks
and
how
conditions
are
implemented
today
or
some
limitations,
and
this
is
also
something
that
this
proposal
and
that
you
have
open
here
is
a
triangle.
A
That
okay
I
mean
check
check
dive
on
that
side.
Then
so
the
loops
part,
you
said
right,
which
Gregg
was
finding
right.
So
that's
also
you
were
saying
right
now
is
at
this
point
you
know
in
the
reason
I'm
getting
you
know
that,
that's
probably,
if
you
see
the
point
of
you
know
it
was
in
discussion,
phases
and
some
bit
of
issues
and
then
it
moved
to
that
design
pocket.
Even
there
I
see
people
sort
of
you
know
going
and
posting
multiple
opinions.
You
know.
A
Maybe
you
need
the
loop
so
that
it's
a
pipeline
level
where
there
is
a
task
level
right.
So
my
main
concern
is
you
know
it's
it's
like.
Is
there
like
a
time
limit
now,
hey
I
mean
you
cannot
have
something
in
discussion
for
more
than
two
months
right
now.
Shall
move
have
a
bias
for
action
right.
So
what
triggers
that
right
and
I'm
hoping
you
know
something
triggers,
therefore,
for
the
looping
part
right
I
mean
I
mean
the
general
understanding.
A
C
Well,
the
thing
is,
as
I
was
saying:
it's
rather
a
major
change
from
compared
to
what
we
do
today.
So
we
just
want
to
ponder
carefully,
especially
because
we
are
now
in
beta
and
if
we
introduce
a
new
API
or
a
new
capability,
we
want
to
make
sure
we
we
can
we're
going
to
live
with
that
for
the
next
nine
months.
C
So
no,
that's
that's
the
reason
for
for
the
discussion.
Well,
I,
don't
think
that's
going
to
be
much
more
discussion
on
that.
I
think
it's
getting
to
a
point
where
we're
probably
ready
to
move
ahead,
but
I
need
to
I,
probably
need
to
circle
back
with
Craig
and
see
what
what
the
things
on
that.
It's
also
a
matter
of
yeah.
C
A
C
Terms
of
contributors,
its
evolving,
so
maybe
it's
like
link
I,
think
it's
it's
slightly!
Slowly
increasing!
Now
it's
getting
a
bit
of
momentum
in
terms
of
people
that
is
actually
focusing
on
contributing.
So
there
are
a
few
a
couple
of
new
folks
that
are
technically
spending
a
lot
of
time
to
quality
reviews
so
which
is
great
to
see
and
that
usually
you
can
see
because
there
are
new
features
coming
out
and
other
features
of
the
moving
faster
forward.
C
We
have
a
lot
of
things
in
scope.
One
of
the
things
that
is
a
high
point
of
focus
and
importance
for
protecting
community
now
is
the
the
catalog
and
the
hub,
which
is
a
way
for
people
to
discover
tasks
and
to
have
kids,
something
like
a
docker
hub
or
if
you
want
help
repos
well,
we
left
different
tasks
that
people
can
pick
with
versions,
so
you
can
rely
on
a
certain
version
of
a
task,
not
changing
for
you
and
different
level
of
support.
C
B
A
Know,
there's
always
you
know
loop
and
sauce
trajectory.
You
know
the
users
play
much
faster
than
the
contributors
right.
I
mean
you
always
want
that
healthy
balance
of
more
and
more
users,
converting
back
to
contributors
right
and
helping
on
so
on
the
catalog
and
when
you
say
the
hub
right,
so
so
typically
I
mean
my
understanding
was,
and
you
know
there
were
a
couple
of
issues
where
we
went
back
and
forth.
A
That
check
down
is
going
to
provide
the
framework
right,
which
is
this
is
the
yamen
for
defining
a
camera
of
tasks
etc,
right,
but
not
necessarily
stand
behind
any
tusks,
alright.
So,
for
example,
in
the
queue
flow
pipeline,
there
are
certain
components
where
vendors
stand
behind
those
companies
right,
so
google
has
its
own
homes.
We.
A
Come
up
around
those
companies
right
AWS
has
their
components.
Is
that
the
notion
which
you
are
going
to
move
forward
and
checked
on,
like
as
a
community
to
begin
with,
you
will
stand
behind
certain
core
tasks
or
no?
The
philosophy
is
that
you
know
we
have
the
framework
and
we
are
not
going
to
stand
behind
any
implementation
of
the
task.
That's
you
know
more,
a
user
responsibility
so.
C
The
current
proposal
is
that
we
could
have
three
three
levels
of
support
so
to
say
so.
We
can't
work
like
efficient
tasks
that
are
maintained
by
Tecton
owners
and
that
are
regularly
test
cells,
come
from
CDE
and
yeah,
so
they're
maintained
by
the
Tecton
community,
then
we're
going
to
verify
type
of
tasks
that
are
not
maintained
by
Tecton,
but
they
are
also
tested
in
the
same
way.
So
they
are
contributed
by
other
third
parties
but
yeah,
so
they
are
tested.
C
A
C
There's
going
to
be
tasks,
they're
going
to
be
maintained
so,
as
I
was
saying,
we're
considering
freely
three
levels
of
support
for
tasks,
so
tasks
that
are
maintained
with
what
we
call
official
tasks.
That's
a
time
attained
by
Tecton
by
the
Technical
Community
does
have
tested,
verified
and
maintained
by
by
us.
Then
we
intermediate
level,
which
are
tasks
that
are
also
tested
automatically,
but
not
maintained
by
the
Technical
Community,
and
then
there
is
going
to
be
community
files
that
are
contributed
by
third
parties
and
not
tested
by
us
in
any
way.
So.
A
C
So
we
this
going
to
be
two
components,
so
the
catalog
itself
is
get
based,
so
the
tasks
are
stored
in
gate,
but
there
is
also
going
to
be
a
hub
which
is
a
web
UI
and
two
that
allows
it
to
discover
this
task
and
that's
I.
Have
not
sin
advanced
hop
design
yet,
but
there
may
be
an
API
as
well
to
discover
them.
So
there's
going
to
be.
A
A
A
So
I
think
yeah
I
level
is
again
as
I
said
right.
Technically,
there
is
one
project
which
is
actively
bringing
Tecton
and
QL
pipelines
together
in
a
more
seamless
fashion.
Right-
and
that's
where
you
know
also.
My
question
was
around
the
the
hub
perspective
right
because
we
are
planning
and
we
have
implemented
a
version
of
the
components
of
or
cue
flow
pipeline
components
of
right
which
we
are
planning
to
to
figure
out.
A
Is
this
more
seamless
single
entry
point
where,
whether
it's
a
component
or
a
cue
flow
pipeline
component
or
a
Tecton
task,
you
know
and
user
experience
should
remain
the
same
in
terms
of
using
it
within
their
pipelines,
right
open
in
the
cue
flow
pipelines
right.
So
we
had
a
version
of
the
UI,
which
was
very
q
flow
pipeline
centric
right
for
the
hub
and
the
API
is
right
now
the
next
set
is,
you
know,
so
I
want
to
that's
why
I
want
to
understand
the
components
part
of
it
but
yeah
the
main
technical
project.
A
Is
this
and
then
Terry
drives?
You
know
much
larger.
You
know
MLS
roadmap,
where
he
is,
he
will
working
with.
You
know:
users
and
consumers
and
customers
on
his
side
what
he
is
seeing,
as
surfacing
as
the
requirements
right
where
this
field
is
going.
There
actually
has
done
very
extensive
and
good
job
of
listing
you
know
and
defining
that
roadmap.
So
that's
how
we
are
sort
of
balanced
at
this
point.
D
Well
so
I
have
you
know
fairly
deep
interest
in
ml
ops
at
this
point,
so
I'm
involved
in
a
couple
of
communities
for
that,
so
that
would
be
an
area
of
major
student
I.
Don't
I,
don't
really
know
very
much
about
Tecton
and
I.
Don't
really
have
a
good
feel
for
the
time
horizon
on
the
roadmap
for
that,
as
far
as
the
integration
with
cue
flow
so.
B
A
The
compiler
right
so
this
this
part,
which
is
essentially
you,
know,
compile
and
run
right
so
fairly
well
test
it.
You
know
being
documented
right
so
like
we
are
hitting
80%
of
the
queue
flow
pipelines
which
can
be
compiled
things
as
I
mention
right.
Not
every
feature
is
going
to
be.
One
is
to
one
mapping
between
what
our
book
provides
versus
what
you've
Tecton
provides.
So
those
are
the
areas
right
where
we
continue
working
with
the
czech
town
committee
community.
A
Now
it's
andrea
is
our
IBM
committer
and
you
know
part
of
the
governing
board
or
Techtron
site
right,
so
he's
our
eyes
into
that
community
from
an
IBM
perspective
right
who
looks-
and
you
know,
prioritizes
the
requirements
we
have
right
and
there
are,
you
know
quite
a
bit
of
requirements
right.
So,
as
I
mentioned,
you
know,
I
was
going
through
this,
so
the
loops
and
the
conditions
are
the
big
ones
right,
but
for
the
most
part
you
should
be
able
to
now
run
using
this.
A
You
see
that
what's
essentially
triggered
on
the
Techtron
side,
the
same
pipeline,
which
you
are
running
on
chip
and
then
you
get
the
same
experience
so
yeah
I
mean
if
you
are
interested
in
using
it
by
all
means.
You
know
it's
in
the
first
stage
where
you
can
actually
get
started
and
start
running
it
yeah.