►
From YouTube: Weekly Sync 2021-02-02
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.ren6doj82m8h
A
All
right,
all
right,
so
just
to
recap:
what
happened
here
was
shaw,
was
working
on
the
sk
time
model
and
ran
into
this
issue,
where
let
me
bring
up
that
that
gist
there
so
so
yeah.
I
just
took
your
your
code
from
paceman
and
I
put
it
in
a
guest,
so
you
could
see
the
diff
here
so
we've
got,
we've
got
this
model
and
exponential
smoothing.
Oh
that's
funny.
A
I
was
just
looking
at
exponential
smoothing
for
something
yesterday
and
I
didn't
put
two
and
two
together
that
I
had
actually
looked
at
your
model
also
for
exponential
spoon
that
morning.
That's
hilarious,
okay,
okay,
so
essentially
what
happened
here
was
you
have
this
code
and
you
have
this?
Is
this?
Is
the
code
to
run
it
right?
Okay,
so
you
loaded,
you
know
you
loaded
the
sample
data
set
from
from
sk
time.
You
did
the
test,
train
split
and
you
have
these
data.
A
You
know
pandas
data
frames,
so
this
is
something
that
that
has
come
up
before,
but
we
don't
have.
You
know
we
don't
have
support
for
taking
pandas
data
frames
right
now
and
and
so
what
you
tried
to
pass
a
pandas
data
frame
right,
so
y
train
is
a
data
frame.
You
pass
that
to
them
to
the
train
and
and
actually
functions
and
and
dfml
has
no
idea
what
to
do
with
it.
So,
basically,
that's
that's.
The
point
of
this
issue
is:
is
we
need
to
figure
out
okay?
A
How
do
we,
how
do
we
make
it
so
that
you
can
use
a
data
frame?
Well,
the
answer
is,
of
course,
like
we're,
going
to
make
a
data
source
right
and
data
source
will
max,
wrap
the
data
frame
and
then
there's
a
little
bit
of
of
glue
code
here
with
in
the
high
level,
so
within
the
high
level
file.
So
the
high
level
functions,
train,
test
or
train
to
train
accuracy
predict.
B
A
Okay,
so,
within
within
the
high
level,
functions
right,
the
train
and
and
actually
predict
they
call
this
records
to
sources
method.
So
let's
see
yeah
you'll.
So
you
see,
if
you
go
in
here
to
accuracy,
you
see
the
first
thing
it
does
or
here's
predict
I
scrolled
too
far,
but
basically
you
know
we
take
the
model
we
take.
You
know
some
variable
length
arguments
and
then
some
keyword,
arguments
and
our
variable
length
arguments
are
okay,.
A
So
the
variable
length
arguments
is
like
when
we're
passing
just
these
dictionaries
of
data.
That's
you
know
what
that
star.
Args
is
so
you
know
here
we're
patching
dictionaries.
We
can
also
pass
like
file
names
and
stuff
right.
Where
is
it?
A
You
know,
file
names
or
sources
themselves,
and
so
all
these
you
know
these
three
functions.
Basically,
they
have
that
same
functionality
with
the
way
that
we
pass
those
those
variable
length,
arguments
them.
So
the
first
thing
they
do
is
call
this
records
to
sources
function,
which
is
also
in
high
level.
A
So
we
go
in
here
and
we
basically
you
know
we
try
to
convert
everything
we
we
want
to
turn
whatever.
Whatever
was
given
us
to
us
into
this
sources.
Object
and
sources
is
an
array
of
data
sources
right.
A
So
if
we
were
past
a
bunch
of
dictionaries,
we're
going
to
put
those
we're
going
to
create
a
memory
source,
make
those
dictionaries
into
records
and
add
them
to
the
memory
source
and
that's
what's
going
on
here
and
so
there's
a
memory
source,
and
so
what
we
need
to
do
in
this
case
is
is
we're
passed
so
here
we're
oops.
That's
the
source
itself,
so
here
we're
past,
so
train
and
then
y
train.
A
So
instead
of
getting
a
dictionary
right
like
in
in
the
quick
start,
example
we're
looking
at,
like
you
know,
dictionaries
or
file
name
or
a
source
itself,
so
now
here
we're
getting
a
data
frame,
so
we
need
to
to
complete
this
issue.
What
we'll
do
is
we're
basically
modifying
we're
going
to
modify
this
records
to
sources
or
yeah.
A
Let's
records
the
sources
function
that
gets
called
from
train
accuracy
and
predict
to
make
it
say:
okay,
if
you're
looking
this
in
this
section,
is
work
where
there's
a
for
loop
over
the
arguments
right
and
and
here's
where
we're
checking.
If
it's
a
dictionary,
we
make
it
a
record.
Well,
we
add
this
little
block
that
says
if
it's
a
data
frame
and
to
avoid
importing
pandas,
because
we're
trying
to
keep
this
the
dfml
at
the
top
level
free
of
free
of
imports,
so
we
avoid.
A
A
If,
if
we
run
into
something
that
doesn't
have
a
class
and
a
call
name-
and
then
we
basically
check
okay
is
the
is
the
string
data
frame
in
the
in
the
class
name
right,
so
you're
gonna
see
like
pandas.something.something.dataframe
for
the
class
name
of
dataframe,
and
so
we
basically
say:
okay,
you
know
if
this
object
that
we're
being
passed
as
a
data
frame
then
create
an
instance
of
dataframe
source
right,
and
so
that's
that's
this
branch.
A
And
if
we
look
in
this
issue,
I
mentioned
that
okay,
so
we
I
gotta,
start
on
it
right
enough
to
sort
of
unblock
to
get
to
the
point
where
that
that
model
was
accepting
the
data
frame,
and
so
what
I
did
was
I
implemented
the
records
and
record
methods.
So
I
created
this
data
frame
source
right.
Here's
that
edit
here
to
instantiate
the
source.
If
we
see
a
data
frame
and
then
the
data
frame
source
that
I
created
so
far,
so
we
still
need
to
do.
A
We
need
to
implement
test
cases
for
it,
and
so
we
are
going
to
require
pandas
within
the
requirements,
hyphen
dev
file,
and
I
I
believe
we
also
require
some
other
things
like
the
numpy
or
pi
image,
or
something
in
there.
So
we're
you'll
require
pandas
in
there
and
then
you
need
to
write
some
tests
for
this
data
frame
source
to
you
know
make
sure
that
we
have
full
full
coverage
line
by
line
in
here.
A
So
we
basically
you
know
we
need
to
test,
for
you,
create
a
data
frame.
You
wrap
it
with
the
data
frame
source.
You
can
use
the
high
level
save
and
load
functions.
Let's
see
that
we
may
need
to
modify
those
as
well.
So
let's
see
here.
A
Okay,
is
this
load?
Okay,
this
also
save
and
load,
looks
like
they
also
use
record
to
sources,
so
you
can
probably
just
use
the
high
level
save
and
load
functions
to
write
your
tests.
So
basically
you
know
you
save
some
records
to
the
data
frame
load,
some
records
from
the
data
frame
and
and
this
code
here
is
your
intermediary
code,
which
is
basically
you
know
your
the
the
source
is
exposing
the
data
frame
within
the
ffml.
A
You
know
world
right,
so
so
within
dfml
we
we
interact
with
sources
so
that
you
know
that
way,
whether
we're
interacting
with
a
data
frame
or
with
a
mysql
database.
The
model
see
it
as
the
same
way
right
so
so
we're
we're
basically
just
doing
this
light
wrapper
here
with
the
source
on
top
of
the
data
frame,
and
so
I
implemented
the
records
in
the
record
method.
A
A
Now
the
there's
another
way
that
you
can
do
this
right,
so
so
I've
we've
done
this
before,
where
we
run
into
this
problem,
where
okay,
I've
got
data
frame,
what
the
hell
do
I
do
with
it,
and
the
answer
is
essentially,
you
can
take
where's
that
data
frame
stuff.
A
I
think
it
was
in
this
code
stuff,
but
no
train
file
data
frames,
oh
yeah
data
frame
from
records.
Oh
no.
This
shows
how
to
create
a
data
frame
from
okay,
never
mind,
there's
a
way
okay.
Well,
we
have.
I
thought
I
had
a
way
documented
somewhere,
but
I
don't
think
I
do.
I
thought
I
had
written
a
little
tutorial,
but
I
I
wrote
this
code
with
forecasting
one
and
it's
not
this
so
anyways.
B
Yeah,
so
you
mentioned
that
you
had
a
way
you
haven't
created
a
clothing,
but
it
would
help
if
you
would
just
show
me
the
code
base
for
that.
B
No,
no,
no,
no,
no,
the
way
you
sort
of.
A
That
is
so
that's
basically
this
branch
here.
So
if
you
look
in
this
issue,
this
branch
that's
referenced,
the
source
data
frame.
This
is
the
so
basically
start
from
here
right,
and
this
is
how
I
did
it
is
basically,
I
called
data
frame
it
or
tuples,
so
we
can
just
trace
through
it
right
now
and
just
just
so
that
it's
clear,
but
essentially,
let's
see
so
when
we
call
train
what
happens
is
that
we
end
up
in
this?
A
The
first
thing
that
happens
is
train
calls
records
to
sources
with
all
the
arguments
that
were
passed
to
it
right
and,
and
your
argument
was
y
train
right.
So
y
train
becomes
the
first
element
in
this
tuple
that
is
args,
so
we
convert
args
to
a
list.
We
go
through
the
list
of
args,
and
so
the
first
one
and
the
only
one
we're
going
through
in
this
case
is
so
arg
is
going
to
be
y
train.
A
So
arg
is
this
at
this
time
so
arc
right
here
is
y
train
and
then
we
look
and
we
say:
okay,
look,
let's
see
if
the
class
of
you
know
the
class
of
that
object.
We've
got
this
object,
which
is
arg
the
variable
arg
right
now
and
we
want
to
check
okay.
Is
this
thing
a
class
all
right
so
does
it
have
the
class
attribute
and
does
that
class
attribute?
Have
a
qual
name
attribute,
which
is
going
to
be
the
name
of
the
class
and
if
it
does
then
check
it
to
see?
A
If,
if
you
know
the
string
data
frame
is
in
there
to
see,
if
this
is
you
know,
arg
is
basically
of
type
data
frame
and
when
we're
doing
this
yeah,
so
this
is
basically
a
hacky
way
to
do,
is
instance
without
importing
panda's
data
frame
and
if
it,
if
it
is,
then
we
instantiate
this
data
frame
source
and
we,
you
know,
as
the
config
parameter
for
the
data
frame.
We
we
say:
here's
arg.
A
So
then
we
go
and
implement
the
data
frame
source
and,
let's
see
okay,
so
this
is
probably
gonna
need
to
be
changed.
So
we
probably
need
to
probably
need
to.
A
Require
not
have
a
default.
I
need
to
remove
the
default
value
of
none
here.
I'm
not
sure
I
can't.
I
don't
know
why.
I
added
that
it
doesn't
need
a
default
value.
There's
it
doesn't
do
anything
if
there's
no
data
frame
so
so
yeah.
So
you
instantiate
the
data
frame
right
you're.
You
pass
the
data
frame
here
right,
so
data
frame
equals
the
arg,
which
is
y
train
right.
So
now
what
will
happen?
A
Is
you
know,
as
we
use
the
source
later
right,
the
source
will
when
we
call
sources
with
features
that
ends
up,
calling
the
records
method
of
this
source
right,
and
so
then
we
iterate
over.
You
know
the
rows
in
the
data
frame
and
we
convert
it
to
a
dictionary
and
we,
you
know,
delete
the
things
we
don't
need
right.
The
index
becomes
the
records
key
and
the
feature
data
is
you
know
the
the
row
as
a
dictionary
minus
the
index.
A
The
similar
thing
happens
for,
for
you
know,
if
you're,
just
getting
one
record
by
the
key
and
then
so
you're
going
to
need
to
figure
out
how
to
implement
the
update
method
and
then
you're
going
to
need
to
write
some
tests
for
this,
and
that's
going
to
enable
you
to
to
to
pass
a
data
frame
and
that's
going
to
be
really
sweet.
Because
I
think
a
lot
of
people
want
to
do
that.
B
Right
right,
so
the
update
method
base.
What
does
the
update
method
do
again.
A
The
update
method
will
so
the
update
method.
Let's,
let's
take
a
look
at
the
memory
source.
A
A
A
A
No,
it
doesn't
okay,
all
right
all
right.
I
don't
think
we
have
any
good
examples
of
update.
So
essentially
oh
yeah,
we
do.
We
can
look
in
the
example
sql
light
source,
okay,
so
in
the
update
method,
you're
going
to
look
at
the
features
of
the
record.
A
So
this
is
a
convoluted
one,
because
you
know
there's
a
bunch
of
sql
going
on
here
too,
but
but
the
gist
of
it
is
that
you're
going
to
you
basically
look
at
the
record
the
data
in
the
records
feature.
You
want
to
look
at
the
record
and
you
want
to
save
the
updated
representation
right
so
so,
for
example,
with
your
pandas
data
frame
you're,
going
to
find
the
correct
index
and
you're
going
to
update
the
fields
within
that
index,
right
you're,
going
to
update
the
column
values
for
that
index.
A
So
here
you're,
seeing
you
know,
there's
basically
an
insert
or
replacement
happening
where
we
look
at
the
key
and
then
we
basically
take
all
the
feature
data
and
we,
you
know,
update
the
feature
data
in
the
sql
database.
You
know
we
update,
we
take
the
columns
and
we
update
them
to
be.
You
know
what
what
the,
what
the
value
should
be
for
those
columns
for
that
record.
Does
that
make
sense.
A
Yep,
yep
and-
and
my
guess
is
there's
going
to
be
so:
here's
the
documentation
for
the
data
frame,
my
guess,
is
you're
going
to
use
something
kind
of
like
where
was
it?
A
You'll
probably
use
the
columns,
there's
like
a
there's,
a
way
to
find
the
columns,
yeah
so
the
column
labels,
and
then
I
believe,
if
you
use,
I
said
I
at
yeah.
I
at
I
think
you
know.
You'll
probably
want
to
look
at
the
columns
of
the
data
frame.
You
want
to
find
the
index
of
the
column
that
you're
updating
based
on
the
feature
key
right.
So
if
my
feature
key
is
b,
then
I'm
going
to
look
at
the
data
frame
columns
and
I'm
going
to
look
and
say:
okay.
A
Well,
where
is
b
well,
the
index
of
b
is
1.
So
if
I
wanted
to
update
you
know
record
so
if
I
wanted
to
update
record
one
feature
b,
then
I'd
I'd,
you
know
index
at
well.
Okay.
So
if
I
wanted
to
update
record
one
feature
c,
then
I
would
index
at
you
know:
I'd
use,
I
add
one
two
equals
you
know
the
the
record
data,
so
record,
dot,
feature
c,
so
that's
you,
you'll
figure
it
out,
but
essentially
yeah.
So
basically
right.
A
A
All
right
cool,
absolutely
so
that's
yeah!
That's
the
plan
here
cool
and
then
yeah
so
see.
Recording
for
explanation.
Oh
no,
we
lost
two.
Don't
you.
B
Not
yet,
no,
I
think
I'll
tackle
the
issue
first
and
then
we
can
talk
about
that.
A
C
A
Been
asked
that
multiple
times-
and
I
just
you
know
I
hadn't-
I
hadn't-
had
you
guys,
you
guys
know
this
about
me
and
you
know
sometimes
sometimes
it
takes
me
a
long
time
to
get
around
to
something
that
that
would
be
short
to
sort
of
give
a
give
a
little
example
of
I
just
got
there's
a
lot
on
my
plates,
so
anyways,
I
think
you
you'll
be
definitely
you'll
be
able
to
run
with
that.
You
know
and
then
we'll
have
that
functionality,
which
will
be
really
great,
so
all
right,
so
natash.
A
Let's
talk
about
what
we
weren't
able
to
run
console
test.
Do
you
want
to
share
your
screen.
D
Is
it
visible
now,
yes,
yeah,
so
that's
the
s2o
model
right
and
every
test
case
is
working.
Fine,
all
these
things,
but
only
the
only
the
console
tests
are
fading.
So
let
me
show
the
error:
what's
what
was
that.
A
All
right,
great:
okay,
no
module
name;
okay.
So
let's
take
a
look
at.
Let's
take
a
look
at
your
console
test,
plugin
all
right.
Let's
take
a
look
at
your
doc
doc
string.
D
That's
the
h2o
model,
training,
okay,
great
sweet,
yeah
and
that
are
the
this.
Is
the
this
simple
configuration
nice.
That's
right.
A
Okay,
so
let's
just
run
that
doc
string
test
so
here
so
when
you
run,
can
you
let's
see
for
some
reason?
I
can't
see
like
the
very
bottom
little
bit
of
your
screen.
Can
you
do
like
ctrl
c
ctrl,
l,
okay,
there
we
go
okay,
so
let's
go
up!
Do
the
up
arrow
until
you
get
to
the
discover,
let's
see,
yeah
okay.
So
let's
change
this
instead
of
discover
to
python
dash
m
unit
test
dash
v
and
then.
A
Tests.Testmodel.Testmyslr
docstring
test
underscore
model
dot
test
minus
the
lord
doctrines
to
the
class
name
and.
A
In
the
trace
back
actually
too,
if
you
want
yeah
okay,
so
if
you
scroll
up
in
the
trace
back
just
a
little
bit,
this
is
how
I
usually
do
that
so
see:
error
test,
docs
string.
So
I
just
copy
paste
that
thing.
That's
next
to
see
error
all
caps
like
six
lines
up
from
the
one
that's
highlighted:
yeah
yeah.
I
usually
just
copy
paste
that
and
then
it's
done
so
yeah
there.
You.
A
A
Yep
so
yeah
they're
basically
make
sure
the
entry
points
match
up
here
and
then
make
sure
that
it's
installed
right
and
that
you've
changed
the
intro.
You
know
the
entry
points
in
the
setup
file
right
and
then
you
ran
the
reinstall
and
then
it
should
work.
A
Okay,
so
entry
point
not
found
so
so
yeah
have
you,
let's
run
the
re.
So
let's
check
out
your
setup.
A
Let's
see,
let's
check
out
your
setup
py
first,
because
I
want
to
make
sure
the
entry
points
are
correct:
oh
yeah,
so
that
looks
good
right
that
that
that's
the
correct
entry
point
so
yeah.
So
so
how
did
you
run
the
reinstall?
Let's,
let's
just
see
how
you
ran
the
reinstall.
A
Because
did
you
say
you
had
reinstalled
it
because
I
think
the
main
thing
that
I've
run
into
here
is:
is
that
that
I
am
dead
it
doesn't?
It
doesn't
pick
up
those
changes
to
the
entry
points
unless
you
run,
if
you
run
with
dash
dash
force
dash
reinstall
yeah.
So
it
looks
like
that's.
That's
probably
what
happened?
Here's
we
need
to
do
and
then
do
space
dash,
dash
force,
yeah,
okay,
so
this
command
just
ctrl
c.
Out
of
this,
though,.
A
Because
that's
basically,
it
won't
pick
up
the
changes
to
the
entry
points.
Unless
you
add
this
flag-
and
I
argued
I
argued
with
them
about
this-
but
apparently
this
is
the
way
it
is
so.
A
C
A
A
It's
a
good
thing:
you
have
faster
internet,
though
at
least
makes
this
less
painful.
The
other
thing,
the
other
way
that
you
can
do
this.
That
makes
it
slightly
faster.
Okay,
so
there's
a
trick
here,
but
the
problem
is
this
trick
won't
work
in
a
few
months,
so
I
opted
to
not
put
it
in
the
docks.
So
basically
remember
we
had
that
giant
conversation
about
eggs
versus
wheels,
and
so
so,
basically
there's.
If
you
run
python,
setup.py
space
egg
underscore
info,
it
will
update
the
entry
points
now.
A
The
problem
is
that
command,
apparently
as
they
move
away
from
eggs
and
towards
wheels,
will
not
work
very
soon,
so
you
can
use
it
for
now.
If
you
want
I'll
paste
it
in
the
getter,
but
just
also
know
that
it's
not
in
the
docks,
because
it's
apparently
destined
to
fail
soon.
So
I
didn't
want
to
put
it
in
there
to
sort
of
head
off
having
having
that
problem.
A
A
Oh
okay,
okay:
this
is
the
other
problem
with
force.
Reinstall
is
that
dude,
okay,.
A
A
minute
so
the
god
damn
it:
okay,
yeah
the
other
problem
with
force
reinstall,
is
that
it
reinstalls
dffml
and
it
reinstalls
it
from
the
production
version
for
some
reason,
and
this
is
why
we
had
that
hole.
This
is
why
we
have
the
whole
setup
with
the
setup
py
files
that
we
had
where
we
dynamically
check
to
see.
If
dfml
is
installed
as
a
development
package,
because
then
or
if
it's
not,
then
we
don't
add
it
to
the
list
of
dependencies
and
if
it
is,
then
we
well.
We
then
yeah.
A
If
it's
installed
in
development
mode,
we
don't
add
it
to
the
list
of
dependencies.
So
we
just
need
to
do
a
pip
uninstall
dash
y
dffml
reinstalling.
It
will
not
fix
the
problem,
because
what
happens
is
that
it's
installed
in
both
development
mode
and
in
production
mode.
So
you
see
it's
in
site
packages
that
would
be
like
the
production
version.
A
Python
will
automatically
use
the
the
production
version
over
the
development
version.
If
there's
a
development
version
present,
which
is
obviously
a
point
that
I
take
issue
with,
because
I
would
argue
if
you
have
the
development
version
of
something
football
installed,
you
would
want
to
use
the
development
version,
but
that's
the
way
it
is
so.
You
need
to
do
pip,
uninstall,
dash
y
dffml,
so.
A
A
If
you
ran
uninstall
again
it
would,
if
you
yeah,
if
you
ran
uninstall
again,
yes,
you
go
to
cd;
no
don't
so
don't
run
it.
Don't
don't
run
it
again.
Sorry,
cd
back
to
that
that
directory
that
we're
in
with
the
model
and
and
run
the
tests
again.
I
was
just
saying:
if
you
were
to
run
uninstall
again,
then
it
would
uninstall
the
development
version,
so
you'd
see
that
yeah
you'd
see
you'd,
see
exactly
sort
of
the
problem
that
it
it
treats
the
production
version
better.
A
Okay,
god,
damn
it;
okay!
No
it
did.
It
did
uninstall
it
all
right.
Okay,
things
have
changed
all
right
yeah,
so
you
need
to
go.
Do
that
dash
e1!
D
A
A
Yeah,
it's
possible
that
all
the
versions
of
packages
will
explode
everything:
yeah,
okay,
yeah,
it's
mad
about
the
chardette
thing.
Okay,
I
think
you
need
to
do
dfml
service,
dev,
install
and
then
it'll,
it'll
reinstall.
All
the
correct
versions
of
all
the
packages
at
once.
D
A
Yeah-
and
you
probably
want
you
probably
want
dash,
are
you
kidding
me
oh
ebay?
This
is.
This
has
been
happening
yeah.
The
problem
is
that
that
yeah,
the
problem
is
that
that
to
load
anything
it
it
explodes.
Oh
god,
this
is
so
annoying
god.
I
hate
the
stupid
packaging
stuff,
all
right.
Okay,
try
just
uninstalling,
aio,
http
and
chardet,
and
then
rerun
the
dev
install
so
pip
uninstall
dash
y
chart
space
aio.
A
A
D
A
A
A
If
the
development
inversion
is
installed,
because
I
think
what
happens
is
it
goes
and
yeah
I
don't
know,
I
don't
know
what
the
hell
happens,
but
something
is
messed
up
with
the
whole
packaging
thing
and
installing
to
a
virtual
environment
can
can
help
mitigate
this,
because
then
you
can
just
blow
up
the
virtual
environment
if
something
goes
wrong.
A
D
D
D
Meanwhile,
I
just
want
to
know
how
to
write
a
code
for
the
test
cases
of
this
particular
source
that
I'm
working
on
or
the
hdf5.
What's
the
flow
like
the
setup
class
and
then
teardown
class
and
then
again
a
setup
class.
D
A
Yeah,
okay,
so
this
is
basically
I
mean,
so
this
is
leveraging
that
source
test,
and
the
best
thing
you
could
do
is
is
go.
Look
in
that
util
testing
source.
It's
basically
just
I
mean
this
is
sort
of
it's
trying
to
it's.
It's
trying
to
give
you
just
sort
of
a
bit
of
an
abstraction
over.
You
know
needing
to
write
those
testing,
the
update
method
and
testing
the
records
and
record
method
themselves.
So,
and
actually
this
is
a
good
one
for
shaw
to
know
about
the
data,
the
data
frame
source.
A
So
let's
go
into
that
the
dfml
source
code
and
look
at
util
testing,
source
source
test
and
and
that'll
I
mean
that's-
that's
sort
of
your
answer
here
is
that's.
What's
going
on.
A
A
A
This
features
feature
floats
we're
just
going
to
go
through
all
these
properties
and
see.
If
one
of
these
is
a
list,
data
type,
okay,
yeah
list-
you
have
that
execute
algos.
So
what
is
it
a
list
of?
D
A
Okay
yeah,
so
then
we
need
to
change
it
to
be
list
stir
using
the
typing
model,
so
module.
So
it's
capital
list,
sorry,
so
so
just
the
first
little
letter
is
capital
and
then
brackets
stir
within
brackets.
Sorry
brackets
now
parentheses,
the
yeah
like
an
array,
they're
gonna
access,
yeah.
There
you
go
so
this
is
using
that
typing
module.
We
we're
adding
a
type
in
saying
that
there
is
a
list,
it's
a
type
string
all
right.
So,
let's
see,
let's
see,
if
that
does
it.
A
A
C
A
C
A
Some
of
what
okay,
let's
see,
yeah?
Okay,
so
you
got
this
something
with
pickling
and
unpickling
the
model,
so
something
something
to
do
with
loading
and
saving
the
model
line.
D
A
Let's
so
so
different
different
different
objects
will
or
won't
support.
You
know
pickling
with
job
lab
or
pickling
or
other
things
like
that.
So
some
sometimes
this
stuff
works.
Sometimes
it
doesn't
and
pickle
has
a
note
in
there
about.
You
know
highest
protocol
and
whatever
and
and
basically
like
the
different
levels
of
pickle
pickling,
determine
like
you
know
it
determines
like
some
objects.
Some
object,
like
a
dictionary,
easy
to
pickle
right.
A
Some
things
like
you
know
these
comp,
more
complex
models
with
different
properties
may
or
may
not
be
serializable
to
disk.
You
know,
for
example,
if
they
may
be
backed
by
c
code
or
stuff,
you
know
it
can
be,
it
can
be
complicated
and
python
may
not
know
how
to
store
them
disk-
and
that's
maybe
what's
happening
here.
So
let's
take
a
look
at
wine.
1
198
in
your
file.
D
A
Yep
you're
saving;
okay,
sorry,
let's
see
okay,
so
job
lib
load.
D
I
think
in
that
case
I
need
to
find
out
the
another
way
to
save
the
model
if
job
live
and
pickles
doesn't
support
in
s2.
Also
right
I
mean.
A
I
think,
let's
see
yeah.
Where
are
you
saving
the
model?
I
guess
let's
see,
so
that's
loading
the
model
right.
So
where
are
you
saving
it?
At
the
end
of
the
train
function,
maybe
yeah
they're
gonna,
they
will
have
some
kind
of
way
of
job,
lift
jump,
jump
yeah
they
so
so
yeah,
my
guess
is,
is
you
need
to
go,
look
at
h2o
and
see
what
they
say
about
saving
and
loading
the
model
right
and
actually
you
know
well,
let's
take
a
look
here.
A
So
let's
take
a
look
at
that
object
that
you
created
that
that
scroll
up
a
little
bit
so
that
saved
so
that
thing
does
that
have
anywhere
where
it's
specifically
saving
to
disk?
D
Because
h2o,
home
or
ml
is
like
a
collection
of
multiple
machine
learning.
A
A
This
may
be.
You
know
that
may
be
the
case
here.
You
know
what
I'm
saying
right,
because
you're
instantiating
that
h2o
automl
instance,
using
all
the
properties
of
your
config
right
and
one
of
those
properties
in
your
config,
is
the
directory
to
which
you
save
and
load
right.
Isn't
it.
A
D
A
May
you
may
be
able
to
get
like
if
pickling
the
model
is
not
what
they
want.
You
know
if
maybe
they're
using
export
checkpoint,
stir
and
project
name
to
then
create
a
you
know,
maybe
they're
creating
a
directory
within
their.
You
know
using
those
two
attributes
and
they're
saving
and
loading
for
you.
Somehow.
A
If
that's
the
case,
then
you
know
we
just
want
to
make
sure
that
we're
instantiating
that
model
with
the
same
project
name
and
export
checkpoints
so
that
it
knows
where
to
save
and
load
from
right,
and
in
that
case
you
know
we
would
we
would.
We
would
pickle
our
config
object
right
and
then
we'd
instantiate.
You
know
we'd
load,
the
the
pickled
config
and
then
just
instantiate
an
instance
of
that
h2o
ml
object
with
the
same
config
right.
If,
if
that
thing
is
handling
it,
underneath
you
see
what
I'm
saying.
D
A
A
And
and
definitely
look
into
look
into
what
they
recommend
right
actually
for
the
dolphin
pie,
stuff,
we'd
and
we
weren't
sure,
right
and-
and
we
opened
an
issue
with
them
and
we
asked
the
guys.
You
know
we
asked
the
project
team.
How
do
you
recommend
doing
this
right?
You
know,
how
do
you
recommend
saving
and
loading?
And
they
said
you
know
after.
A
They
finally
replied
and
said:
oh,
we
have
job
lib,
whatever
support
so
and
it
sounded
to
me.
I
haven't
used
job
lib
much,
but
it
sounded
to
me
that
there's
maybe
some
methods
or
something
that
you
can
define
in
your
class
that
allow
job
lib
to
you
know
better
serialize
and
unserialize
it.
You
know
the
class
and
and
so
the
alpha
pi
guys.
You
know
just
as
an
example.
They
must
have
implemented
those
these
h2o.
A
You
know
whoever
is
maintaining
h2o
ml.
They
may
not
have
implemented
that,
and
so
pickle
probably
is
trying
to
do
it
the
best
it
can,
but
it
may
not
do
it
correctly,
whereas
you
know,
if
you
they
had
implemented
these
specific
methods,
then
maybe
it
would
do
a
better
job
of
serializing
and
unserializing
or
maybe
they
have.
You
know
some
documentation
on
hey.
We
don't
support
job
lab.
You
should
do
it
like
this
or
if
they
don't,
then
you
can.
A
A
D
Yeah-
and
I
think
this
actually,
I
have
a
another
parameter
in
a
conflict
that
is
show
leaderboard
after
the
training
of
this
model.
S2O
model.
It's
it,
then
this
multiple
different
models
and
then
create
a
leaderboard
kind
of
stuff
right.
A
D
If
user,
if
user
wants
to
show
a
leaderboard
in
their
console,
so
I
just
made
a
boolean
variable,
like
user,
can
config
on
the
console
that
show
later
about
true
or
false
on
that
basis,
I'm
just
showing
in
the
logger
right
so.
A
A
Okay
right
because
the
user
is
asking
for
this
information
right,
the
user
is
not
asking
for
debug
information
info
is
probably
the
appropriate
level
here.
D
Oh,
I
think
everything
is
working
fine
for
s2o
and
for
this
part,
what
what
file
I
need
to
read,
as
I
said,
bfml
util,
oh.
A
A
Testing
source
dot,
py
and
then
the
class
in
there
is
source
test.
So
yeah
just
take
a
look
at
that:
slash
testing,
source
dot,
py.
A
A
A
Yeah
and
and
essentially
what
it's
doing
here
is
it's
basically,
you
know
it's
creating
one
instances
of
the
class.
You
know
when
when
we
you
know
instantiate
this
test
case
and
then
or
that
you
know
that
we
create
the
class
and
then
it's
using
the
same
instance
of
the
source
or
no,
let's
see
what
is
it
doing
setup
class.
A
Oh,
oh
yeah,
that
setup
source
method
returns
the
the
source
itself
setup
class
is
used
for
creating
that.
I
think
you
have
the
right
format
here.
I
would
look
at
you
know.
There's
there's,
there's
other
examples
that
you
can
look
at.
I
would
basically
be
reading
the
code
to
you.
If
I
told
you
right
now,
because
it's
been
so
long,
so
you'll
you'll
you'll
find
out
when
you
read
it,
but
and
then
just
ask
if
you
have
any
questions:
okay,
cool
all.
C
A
Well,
any
oh,
we
have
sudhanshu
now
in
sutanshu
anything
else
from
natasha
or
shaw.
Do
you
guys
have
anything
else?
You
want
to
talk
about.
A
A
Down
this,
the
changes
from
the
master
branch
and
you
either
rebase
or
merge
in
the
changes
right
yep
and
then
you
push
it
up
and
I
think
it's
probably
going
to
be
something
with
a
change.
Log
is
my
guess,
but
yeah,
it's
usually
the
changelog.
So
all
right.
C
A
Sorry
I
haven't
gotten
to
the
example
thing.
Yet
I
looked
at
it
and
I
realized
that
this
is
tricky
it's
trickier
than
I
thought
it
would
be
to
start
with
on
like
how.
How
do
we
set
this
up,
mainly
because
we're
now
involving
javascript
and
so
the
questions
become
a
little
more.
You
know
it
becomes
a
little
more
like
okay,
we
talked
about
the
web
ui
last
week
and
stuff
right,
so
it's
now
we're
getting
into
the
space
of
okay.
We're
gonna
start
testing
javascript.
A
How
do
we
do
that?
So
there
was
a
couple
things
that
I'd
looked
into,
that
there
was
this
playwright
module
from
microsoft,
which
looks
awesome
if
you
guys
ever
need
to
do
some
browser
automation.
That
thing
looks
really
cool
and
then
there's
some
other
things
like
like
cypress
that
look
cool
too,
but
I'm
not
I'm
not
entirely
sold
on
anything
yet,
especially
because
I'm
not
sure
how
that
all
interacts
with
the
console,
since
that
stuff
is
the
the
test
the
example
stuff
just
dumps
to
the
console.
A
So
I
haven't.
I
haven't
gotten
too
deep
into
that
yet
and
I'm
not
sure
does
anybody
have
any
anything
that
they've
used
for
javascript
testing
before
that
they
really
love.
A
A
I
think
I've
heard
of
selenium
before
this
is
like
this
has
been
around
for
a
while.
It's
pretty
pretty
battle
tested
one
nice
yeah.
Where
was
the
let's
see
what
the
docs
look
like.
A
A
It
may
be
worth
going
with
the
the
playwright
because
of
the
I
think
it
thinks
a
sync
interface,
but
yeah
you
see
it
looks
basically,
it
looks
very
near
near
the
same
thing,
so
anyways,
we'll
we'll
I'll,
probably
I'll.
Probably
I
want
to
do
some
more
digging
just
because,
just
because
I
see
this
being
something
that
we
use
more
and
more
of-
and
I
don't
want
to
want
to
make
a
hasty
decision
here
so.
C
Yeah,
so
actually
I
wanted
to
talk
about.
How
are
we
like
going
to
do
the
the
mctx
route
and
the
ac
so
previously,
like
we
had
decided
to
go
with
the
the
parameterized
decorator,
so
I
have
actually
implemented
the
parameterized
decorator
but
like
there
are
some
issues
which
I'm
facing
right.
A
Now:
okay,
let's,
let's
have
you
share
and
and
then
yeah
look
at.
C
So
is
the
skin
visible?
Second,
yes,
yeah,
so
what
I
have
actually
done.
So,
let's
start
from
the
beginning.
C
C
Which
is
take
the
m
label
and
I've
also
added
the
ac
tx
route
to
wrap
what
is
being
given
by
this
okay.
A
C
So
in
the
mctx
root,
so
this
is
what
changes
I
have
made.
So
I'm
I'm
not
able
to
figure
out
like
if,
in
the
keyword
arguments
I
will
get
the
the
label
keyword.
But
how
am
I
like
going
to
use
it
in
here.
A
A
C
A
A
A
These
decorators
get
really
hard
to
get.
They
get
like
really
hard
to
wrap,
wrap
ones
head
around.
When
you're
doing
these
nested
levels.
I
had
a
hell
of
a
time
trying
to
figure
out
that
op
thing
it
took
me
forever.
Okay,
so
yeah
we've
got
an
error.
Let's
try
running
just
this
one
test
so
test
so
copy
paste.
So
on
that
line
we'd
yeah
there
we.
A
A
A
A
All
right,
oh
yeah,
we
need
to
log
in
on
yeah
the
reason
why
it
does.
That
is
because
we
don't
want
to
print
stack
traces
to
the
client
from
a
security
perspective.
So
let's
see
actx
missing
positional
required.
A
A
Mctx
or
let's
see
yeah,
I
think
what
you
mean
to
do
here
is:
oh,
let's
see
yeah.
Okay
now
this
is
interesting.
So
what
we
ran
into,
I
think,
is
the
fact
that
they're
stacking
now,
because,
let's
see
I
think
what
happened
here
is
is-
is
calling
git
actx
and
it's
not
expecting
to
be
stacked
with
another.
One
of
these
you
know
underscore
root.
Decorators.
A
Well,
I'm
kind
of
thinking.
Maybe
we
want
to
make
these
a
keyword
argument,
because,
or
else
you
know
we're
in
this
positional
argument
space
here
so
or
let's
see
or
we
could
do
star
args.
I
think
what
we
could
do
is
like
get
a
ctx.
A
Let's
see,
yeah,
let's
look
at.
Let's
look
at
the
mctx
root
function,
so
it
basically
calls
handler.
It
says,
get
mctx
request
handler
and
then
it
passes
mctx.
A
A
A
Yeah
there
we
go
so
see
you
have
at
a
ctx
root
and
then
at
mctx
root
right
so
at
doing
at
and
then
the
function
name
right
for
a
decorator
is
equivalent
to
being
like
a
ctx
root.
You
know
so,
or
I
guess
so-
with
mctx
root
label
is
label.
That's
when
you
do
at
that's
equivalent
of
saying
mttx
root
and
then
calling
label
equals
label
and
then
calling
the
result
of
that
with
as
accuracy
with
passing
accuracy
score
right.
A
This
thing,
that's
already
been
wrapped
by
mctx
root
and
I
think
what
we
want
to
do
is
we
want
to
call
actx
root
after
we've
called
mctx
root
or
like
we
want
to
wrap
with
mctx
root
first,
because
atctx
root
is
expecting
that
mctx
root
has
already
wrapped
the
function
right.
A
Yeah
and
I
think
the
way
around
this
is
that
we
add.
We
basically
pass
whatever
arguments
exist.
So
if
you
scroll
up
so
don't
call
yes
er
right.
So
let's
see
wait.
A
second,
don't
call
a
ctx
root.
Did
you
just
add
those
two
prints
yeah
there
we
go
so
leave
it
like
that
yeah.
A
Root,
okay,
so
right
so
you
have
git
actx
and
then
mctx
right
or
you
have
gity
ctx
self
request,
mctx
right
online,
one
or
127.
so
yeah.
So
I
think
what
we
need
to
do
is
we
just
need
to
say:
okay
pass.
Whatever
arguments
there
are
right.
So,
instead
of
mctx,
we
need
like
star
yeah,
star
handler
args
handler
underscore
args
right,
so
we
could
do
star
args,
but
we
we
use
star
args
in
the
next
one.
So
let's
just
do
star
handler
so
star
hand
handler
underscore
args.
A
Right,
okay,
great
and
so
now,
if
we
just
add
ins
in
place
of
where
we
have
on
line
135,
where
we
do
handler,
where
we
call
handler
and
we
pass
mctx
and
actx
if
we
pass
in
place
of
mctx,
if
we
do
star
handler
args,
then
that
will
pass
any
arguments
if
they
were
there
right.
So
if
they
weren't
there
we're
not
going
to
pass
them
right
and
if
they
are
there,
we
are
going
to
pass
them
and
we
should
do
the
same
thing
in
git,
mctx.
A
And
that
way
this
way
so
on
git
mctx,
if
we
add
that
star
handler
args
and
then
we
also
pass
them
here
now,
this
way
it
won't
matter
if
we
call
if
we
decorate
with
mctx
or
actx
root.
First,
you
know
it
won't
matter
the
order
that
we
decorate
the
handler
with.
C
C
A
C
So
so
we
are
actually
trying
to
get
the
accuracy
so
in
the.
C
C
One
label
should
be
for
the
model
context
route
and
another
label
should
be
a
ctx
route
so
for
the
model
context
route
in
below
in
the
api
endpoint.
What
we
have
given
here
is
the
for
the
accuracy
route.
Our
label
name
will
be
labeled,
but
for
our
model
context
root,
our
label
will
be
m
label.
So
this
is
what
we
are
trying
to
do
here.
C
So
what
we
are
actually
trying
to
do
here
is
we
are
actually
trying
to
get
both
the
accuracy
context,
scorer
and
the
model
context
scorer
from
these
two
loops.
So.
A
C
Are
actually
trying
to
load
this
like
it
is
mct.
C
C
Are
any
so
we
will
get
the
handler
here
and
we
will
grab
the
handler
and
we
will
give
the
wrapper.
You
buy
the
wrapper,
but
if
we
are
actually
getting
some
label
then
for
that
label
we
will
actually
have
to
load
that
m
label
so
for
we
will
get
in
the
model
context
and
we
will
load
that
m
labels
context,
so
mctx
will
get.
We
will
get
that
here
and
we
will
have
to
pass
that
mctx,
but
we
also
now
have
to
load
the
actx
coder.
C
So
we
have
created
another
route
called
actx
which
will
actually
give
us
the
dx,
and
it
would
also
pass
that
too.
So
we
will
get
both
of
them
and
then
we
will
calculate
the
score
part.
A
Yep,
okay,
so-
and
I
think
the
one
thing
that
we're
missing
here
right
so
so
when
we
wrap
these
right,
we're
basically
adding
another
argument
on
to
the
end
of
of
that
handler's,
you
know
call
arguments
right,
so
we
have
usually
it
would
just
be
request
and
now
we're
basically
saying
we're
appending.
You
know
mctx
and
a6
actx
to
the
end
of
that
handler's
set
of
arguments.
So
we
need
to
go.
Look
at
the
handler
again.
I
think
we
are
missing
ac
ctx
from
the
set
of
arguments.
A
C
A
C
We
actually
have
to.
We
will
actually
get
another
a's
here.
A
Yeah,
okay,
so,
and
and
now
now,
what
we
want
to
think
about
is
I'm
not
exactly
sure
which
which
order
this
is
going
to
go
in
whether
it's
going
to
be
a
mctx
or
actx
first.
So
if
we
run
this-
and
it
says,
ac
actx
has
no
attribute
score
right.
It
tries
to
call
actx.score
and
it
blows
up,
then
we're
going
to
know
that
it
should
be
a
ctx,
comma
mctx.
A
You
know,
because
I
can't
you
know
I
can't
yeah.
We
could
read
that
if
we
wanted
to
right,
but
we
can
also
just
run
it
and
see.
That's
my
preferred
method
of
things
is
write.
The
code
run
it
and
see
what
happens.
A
That'll
tell
us
how
to
read
the
code
right.
What
the
code
actually
does.
So,
let's
see
know
how
to
produce
score.
Yeah
fake
model
context,
so
we
need
to
swap
mctx
and
actx.
So
it
looks
like
actx
is
being
added
to
the
set
of
arguments
first
and
then
mctx's
or
well.
We
don't
we
don't.
We
can
just
swap
it.
We
we
didn't
need
to
yeah.
We
could
do
it
there
too
yeah.
You
could
swap
it
there.
A
All
right,
I
don't
know
if
you're
gonna
hear,
but
I'm
clapping
all
right,
great,
exciting,
yay,
okay,
fun,
great
wow,
that's
great
okay,
so
I
guess
do
we
have.
I
think
we
had
some
more
test
coverage
stuff
right.
We
have
the
configure
routes.
Well,
you
probably
did
you
get
all
the
configure
routes
and
everything
in
the
context
routes.
A
A
Okay-
okay,
great
so
yeah
from
the
examples
test
standpoint.
Okay,
so
you
said
the
examples
were
not
working
in
their
current
state.
A
C
A
Okay,
yeah,
and
I
don't
think
I
ever
ran
that
python
example.
I
just
took
it
from
some
from
someone
who
wrote
it
and
put
it
in
here.
I
think
it
was.
They
worked
at
one
point
I
think
yeah.
No,
I
think
I
did
run
it
at
one
point,
but
it's
been
a
long
time.
So
obviously
that
thing
was
added
a
long
time
ago.
A
Well,
I
have
no
idea
when
now
it's
added,
maybe
if
yeah
well,
it
doesn't
matter.
I
guess
we'll
reset
it
at
some
point
and
it
either
does
or
doesn't
work
now
all
right.
So,
oh
yeah,
and
if
you
guys
have
seen
the
ci
is
failing.
I
think
it's
related
to
the
there's,
an
issue
with
numpy
and
auto
sklearn
and
the
in
this
static
comp,
the
compilation
that's
going
on
there.
A
I
believe
it's
going
to
be
fixed
with
the
pinning
issue
when
the
pinning
issue
gets
fixed
due
to
the
the
fact
that
we
have
numpy
at
a
1.18
or
something,
and
I
think
we'll
get
upgraded
when
we're
done
with
that.
So,
let's
see
all
right,
yay
good,
all
right.
So
next
steps
here.
A
So,
let's
see
okay
yeah
the
examples,
so
you
know
I
guess
I'm
I'm
going
to
look
at
the
examples
anyways.
So
I
think
we
can
just
say
that
that
all
handle
updating
the
examples,
so
we
can
just
you
can
just
move
on
to
the
next
thing.
Oh
yeah
coverage
yeah.
This
is
a
good
thing
to
to
check
out
here,
so
you
can
just
move
on.
Let's
check
what
the
next
phase
here
is.
So.
C
A
Okay,
update
the
http
service.
Okay,
so
are
we?
Oh,
I
think
we
have
some
command
line,
there's
command
line.
Let's
see,
it
looks
like
that.
There's
there's
bullet
points
in
the
bullet.
There's
the
bullet
points
in
the
in
phase
six
of
this
issue,
which
is
732
and
there's
command
line,
instantiation
objects
or
options
as
well.
So,
let's
see.
C
A
I
think
the
one
thing
the
let's
see,
service,
hp,
docs,
cli,
okay,
so
if
you
look
at
the
documentation
for
the
where
is
it
yeah?
So
if
you
look
at
the
documentation,
so
if
you
look
in
in
service
http,
so
you
see,
I
don't
know
what
that
modify.
I
don't.
A
One
that
just
says
modify
was
we'll
just
remove
that
it
was
probably.
A
A
C
A
A
So,
basically
you
know
you
don't
have
to
use
the
configure
route,
it's
already
pre-configured,
so
we
should
make
sure
that
the
we
should
make
sure
that
we
can
instantiate
the
sources
from
the
command
or
the
the
scorer
from
the
command
line
as
well-
and
I
think
that's
our
final
thing
in
phase
six
here
and
then
we're
basically
on
to
phase
seven
and
then.
A
Man,
oh,
this
has
been
a
crazy
project.
You
did
you've
done
a
great
job
on
this.
This
has
been
huge
wow
june
we've
been
doing
this
since
june
holy
smokes.
This
is
a
big
one.
I
mean
wow
all
right.
This
has
been
a
major
refactor,
okay
yeah.
So
let's,
let's
get
this
in
here
into
the
you
know
update
so
while
you're
at
it-
let's
just
you
know
when
we
said
phase,
seven
update
documentation,
that's
sort
of
more
than
that's
like
the
rest
of
the
docs
right.
So
this
is
this.
A
Is
you
know
the
this?
Is
we
still
want
to
update
the
http
service
while
we're
at
it
here,
because
I
think
we're
going
to
forget
so
while
we're
in
the
http
service,
let's
update
the
http
service
docs
so
that
doc
slash
cli,
rst
and
make
sure
there
are
tests
with
this,
and
that
should
be
in
test
slash
test
cli,
and
then
we
also
have.
So
we
need
to
make
sure
that
the
this
works
for
test
cli.
A
I
think-
and
you
should
see,
that
the
sources
and
the
models
in
there
in
test
cli
and
then,
let's
see.
A
It's
one
of
the
you
know
we
have
it
or
hit
or
miss
some
other
places,
but,
okay,
so
test
models,
slr
model
train
the
model,
accuracy.
Okay,
let's
see,
run
great
okay.
I
think
that
is.
I
think
this
I
think
this
is,
I
think,
we're
looking
in
the
right
place.
Let
me
know
if
you
run
into
hit
in
any
issues
here.
A
Okay,
great,
I
think
that's.
C
A
A
Okay,
great
sweet,
hey
there,
we
go
all
right,
that's
what
we
like
to
see:
hey
sweet,
all
right,
very
good,
very
good,
okay.
So
I
think-
and
I
think
you
know
I
updated
the
meeting
minutes
to
say
you
know
this-
is
the
next
step,
we're
looking
at
that
and
I'll
check
these
off
and
in
the
issue
here
and
then
I
think
that's
the
last
thing
for
our
phase.
Six
is
just
that
cli
stuff,
awesome,
okay
and
then,
eventually
we
have
to
rebase
this
beast.
A
Replace
with
okay
play.
A
A
Right,
hey
thanks.
Everyone
have
a
good,
oh
yeah,.
D
A
D
A
Because,
just
to
add
you
guys
to
this
meeting
so
if
I
I've
added
everyone's
email
to
the
meeting,
did
I
add
yours
to
the
meeting
wait
I
thought
I
did
and
now
it's
not
showing
up.
A
Great,
so
just
because
you
know
I
usually
have
to
allow
you
into
the
meeting,
but
apparently
that's
the
way
that
you
can
get
it
to
just
let
you
in.
If
I
add
people's
names
to
the
meeting,
I
thought
okay.
A
So
that
way,
you
guys
won't
have
to
wait
around
in
the
lobby
or
sometimes
you
know,
I
don't
see
the
sometimes.
I
don't
see
the
wow.
I
have
too
many
tabs
open,
okay,
so
retouch,
okay,
yes
you're
on
there
now
too
great
so
yeah,
that's
that's
all
so,
just
just
to
make
sure
that
you
can
get
in.
If
I'm
not
here
and
like
I
said
you
know,
I've
I've
been
a
little
bit
late
because
I
run
this.
A
I
run
this
meeting
before
this
now
too
apparently
tuesday
morning
is
a
good
time
for
a
meeting
or
tuesday
evenings
for
you
guys.
So
all
right
anyways
have
a
great
night
and
a
great
rest
of
your
week.
Let
me
know
if
there's
anything
else
and
I'll
be
around
all
right.
Okay,
thanks
everyone,
great
work,
bye.