►
From YouTube: Weekly Sync 2021-04-27
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#
A
All
right
this
fail,
or
is
this
lgtm
failing
again,
okay,.
A
A
And
27th,
I
think
we
had
a
few
things
that
looked
like
they
were
pretty
much
done
here.
So
we'll
start,
let's
see
so
the
cash
download
stuff.
This
looks
like
it
was
almost
done
right.
B
Yeah,
actually
there
was
another
issue
after
we
thought
it
was
done
and
it
wasn't
popping
up
so
yeah.
I
made
those
comments
about
it.
Oh
weird,
it
took
me
a
long
time
to
figure
that
one
out.
B
I'm
not
sure,
but
that's
what
I
thought.
A
And
and
so
the
file
name
that
we'd
used
was
oh,
was
this:
was
the
temp
file
being
created
with
that
at
the
end
or
wait
was
the
file
name?
Let's
see
return
the
absolute
path.
Now
it's
directed
yeah,
okay,
yeah.
I
thought
this
was
because
I
saw
we
removed
that
absolute
and
I
was
wondering
what
happened
there:
yeah,
okay,
oh!
This
is
great.
B
B
I
removed
it
because
it
was
a
string,
but
I
made
it
absolute
before
we
make
it
a
spring.
Okay,
yeah.
B
Yeah
we
were
before
this.
The
download
path
was
a
path
right,
but
now
it's
spring,
so
you
can't
absolutely
respect.
Okay,
okay,
so
yeah,
I'm
returning
the
absolute
path
from
the
way
back.
A
Okay,
because
it
cache
download
returns
with
string,
okay,
yeah,
that's
right:
okay
and
cache
download
we'll
do
okay,
yeah
and
cash
down
already.
It
already
turns
returns
the
absolute
path-
okay,
great
great,
fantastic,
all
right
and
well,
let's
see
so
now,
I'm
seeing.
A
A
A
Let's
see
because
we
do
we
shouldn't,
we
don't
want
to.
We
don't
want
to
mix
the
two.
You
know,
let's
see,
because
yeah
this
guy
will
return
this
right
now.
A
A
Yeah,
let's
maintain
its
status
as
a
path,
lib.path
object,
just
because
we
end
up
using
those
so
many
times
that
you
know
having
the
first
call
after
the
returned
value,
be
you
know,
create
a
pathfit.path
object.
Out
of
this
string
will
reduce
you
know:
it'll
reduce
code,
we're
right
in
the
future.
A
Object:
let's
maintain
that.
A
A
Oh,
this
doesn't
even
do
anything
with
it
because
it
has
okay
and
this
had
a
target
path
and
it
was
given
the
target
path.
Okay,
so
this
one.
A
A
A
A
Validate
then
return
file
name,
and
I
think
we
just
changed
that
to
one
statement,
but
either
way
that
will
we
need
to
yeah,
let's
just
maintain
it
as
a
path
live
object,
and
I
think
if
you
just
change
it
with
sync
url
retrieve
and
validate
because
it'll
end
up,
basically
just
returning
the
same
target
path
that
it
was
passed
right.
So
it's
not
yeah
it's
just
this.
It's
almost
just
like
a
shortcut
for
return
target
path.
A
You
know
it's
almost
like
a
shortcut
for
just
saying,
sync,
url
retrieve
and
validate
and
then
return
target
path.
If
we
just
return
the
same
target
path
from
retrieve
and
validate,
which
would
just
be
this
cool,
I
wonder
yeah.
I
think
I
don't
think
anything.
Much
will
need
to
even
change
here,
and
so
this
is
really
funny.
So
it
didn't
recognize
zip
as
a
file
when
using
ch2
directory
path.
A
B
A
Great
yeah,
because
that
because
yeah
you
would,
it
would
have
broken
these
guys
if
it
didn't
end
up
with
the
correct
structure,
so
that
must
work
great
great
nice,
very
nice.
A
Do
you
want,
do
you
want
me
to
just
commit
that
one
line,
change
of
just
returning
the
the
path,
instead
of
the
string
to
see
if
it
works
here
by
just
running
the
ci
and
doing
that
one
commit
or
do
you
want
to
go,
make
the
change
yourself.
B
A
Okay,
well,
we
I
mean,
let's
just
we
can
just
try
this.
I
think
it
should
work.
Those
famous
last
words
right,
but
if
not,
we
can
just
roll
back.
We
can.
You
can
just
you
know,
get
rid
of
this
commit
right
so
that
way,
hopefully
you
don't
have
to
go
mess
with
it
again
right.
So
I
don't
want
to
make
no
reason
to
to
to
do
this.
Two
places.
A
All
right,
okay-
and
it
was
passing
before
that,
except
for
was
your
passing
before
that.
A
A
Yeah
that
that's
just
that
that
whole
thing
is
confused,
I
don't
know
what's
wrong
with
it.
Okay,
great!
Let
me
just
write
that
in
minutes
here
I
want
to
that.
One
needs
to
get
done
here.
Okay,
so.
B
Can
we
also
merge
that
csv
delimiter,
oh
yeah,
where'd,
that
go
let's
see.
A
Oh
yeah,
this
was
weren't
you
working
on
that
one
file
or
what
happened
here.
This
came
from.
A
B
Good,
I
didn't,
I
didn't
see
this
one.
I
I
also
pushed
changes
onto
the
use
cases.
Pr
and
all
this
left
is,
I
think,
to
test
it
now.
A
All
right
great
so,
let's
see
so,
which
one
was
that
this
one.
A
What
does
that
wonder
what
that's
meant
about?
I
wonder
if
we
can
say:
okay,
maybe
it
doesn't
like
how
many
values
it
sees
here?
Okay,
I
can
look
at
it
through
here.
A
A
That's
okay.
I
can
ask
okay.
A
Okay,
oh-
and
this
is
the
this-
is
the
movement
between
model
stimulus.
That's
right:
okay,
we
can
probably
get
rid
of
this
warning
stuff
or
well
wait
a
minute.
You
can't
you're
just
running
it,
never
mind.
We
need
to
get
rid
of
that.
That's
a
separate
issue.
I
think
we
have
it
okay,
so
we
got
this
all
right
and
you
have
the
features
and
you
protect
the
quality
and
you
set
the
directory
nice.
A
All
right,
great
great
use,
accuracy,
okay,
oh
and
I
think
we
need
to.
We
need
to
try.
I
need
to
try
rebasing
sudanshi's
accuracy,
stuff.
I
think
we're
pretty
close
on
that.
I
think
it's
almost
ready.
Okay
and
then
we
just
got
a
bunch
of
a
bunch
of
predictions
here,
great
nice,
looking
good
looking
good
and
then
did
you
did
you
have
you?
Have
you
been
able
to
look
into
the
testing
stuff
more
on
this.
A
A
I'll
get
on
to
it
now
awesome
awesome,
yeah,
that's
right!
That
was
the
whole
reason
for
that
cash
download,
stuff.
Okay,
let
me
just
nice,
nice
work.
B
A
A
A
Okay,
we'll
just
wait
just
in
case,
so
this
one
got.
D
A
So
we
need
to
keep
old
path,
lab
dot
path,
return
type
other
than
that.
Looking
good
waiting
for
ci
to
pass.
A
B
Regarding
the
adventure
just
a
light
reminder
on
the
confidence
bro.
A
Oh
yes,
thank
you!
Oh
yes,
I
know.
Okay,
sorry
about
that.
Yes,
oh
my
gosh
there's,
so
I'm
so
behind
see!
This
is
this
is
why
we
need
to
get
this
other
org
going
on
okay,
great
great
yeah.
That
one
is
important.
Let
me
put
that
in
here
yeah
this
is
long
ago.
Jesus,
sorry
pr.
A
D
E
F
A
E
A
Accuracy-
okay,
great
great
great,
great,
great
great-
this
is
fantastic.
Okay,.
A
So
now
everything
okay
great-
and
we
have
these
guys
in
your
async
test
case.
Oh
great,
and
now
we
have
the
stacks
okay
gosh,
because
this
is
such
a
pain.
This
is
such
a
pain
to
have
to
do
that,
to
create
a
new
stack
on
everything,
okay,
fantastic!
So
then
did
we
let's
see
and
you
removed
those
classes.
You
removed
that
class.
You
removed
that
class
great
great
great
great,
and
I
think
test
kicks
async
test
case.
A
A
C
A
So
I
think,
wait
what
happened
with
this,
because
this
one
yeah,
why?
Why
did
we?
Why
did
we
end
up
just
making
this
repo
root?
Because.
E
A
So
I
think
I
think
the
the
thing
is
that
self.package
won't
always
be
dot.
Sometimes
it
is
like
one
of
these
model.
Subdirectories
I
mean.
Ideally,
it
will
just
be
dot
if
we
move
to
that
second
party
plugin
architecture,
but
just
in
case
we
ever
still
have
things
because
right
now,
we
haven't
moved
that.
A
Okay,
package,
repo
root,
you
know
what
what
is
really
so
it
should
it
be
relative
to
the
repo
route,
because
if
we
end
up
making
this,
if
we
end
up
making
this
I'm
just
thinking,
if
we
end
up
doing
when
we
end
up
doing
the
second
party
plug-in
stuff,
then
it
shouldn't
be
relative
to
repo
root.
Actually,
after
all,
because
if
we're
in
a
different
directory,
then
we
won't
want
to.
E
I
think
for
second
party
there's
a
test
in
a
ci
right
and
if
it
is
passing,
then
it
means
it's.
It's
working.
A
Well,
I'm
saying
when
we
go
to
do
because
this
release,
automation
will
also
work,
for
you
know
we'll
be
using
this
for
the
second
party
plugins
as
well.
What's
the
rest
is
code,
okay,
yeah,
it
checks
the
version,
it
checks
on
pi
pi,
it
creates
the
archive
and
it
runs
the
upload
so
and
what
is
this
cd
to
directory?
A
What
were
we
doing
here?
The.
A
A
Well,
that'll
still
be
in
the
plugins
okay.
Maybe
we
should
just
throw
away
this
whole
thing
and
put
this
right
in
the
cis
bash.
A
The
archive
stuff
is
kind
of
useful,
so
because
you
see,
do
you
see
what
I'm
saying,
though,
because
do
you
see?
Do
you
understand
what
I'm
saying
just
did
I
I'm
not
sure
if
I
explained
what
I
was
trying
to
explain
correctly,
I
was
trying
to
say
that
that
right
now,
what
we
have
is
this
self.package.resolve
right
and-
and
this
basically
says
so-
if
we're
this
is
relative
to
the
current
working
directory
and
the
current
working
directory.
A
In
this
case
it
used
to
be
the
root
of
the
repo
right
and
so
dot
meant.
You
know
the
very
you
know
the
path
above
dfml
here
and
if
you
pass
like
model
psychic,
it
would
go
model
psychic
right,
but
if
we
just
make
it
repo
root,
then
it's
always
going
to
be.
You
know
just
dffml.
F
F
E
D
A
A
And
let
us
not
do
that,
because
that
would
be
bad
okay
so
and
the
other
thing
is
it's
gonna.
It's
gonna
say
that
the
version
is
already
on
pi
play
now.
Actually
you
know
this
will
tell
us
right
away.
It's
diff
mouse
service,
dev
release
and
then
we
say
dot
right
and
we're
in
you
know,
model
transformers.
A
So
different,
oh,
is.
A
A
It's
not
about
get
oh
because
it
tries
to
do.
I
tries
to
it
fails,
anyways,
okay,
so
I
guess
it
fails
for
a
different
reason
than
I
thought
it
was
going
to
fail,
because
but
but
the
point
is
the
same
right,
so
it's
it's
pointing
at
it's
pointing
at
a
different,
it's
pointing
at
the
wrong
place
right.
So
it's
trying
to
say
that
you
know
this.
This
directory,
which
is
not
the
one
we
just
tried
to
release,
is
outside
of
the
repository.
A
So
if
we
had
it,
as
you
know,
okay,
let's
see
you
know,
I
think
that
it
was
right,
probably
the
way
that
it
was
so,
let's
see,
it's
probably
just
should
be
path.
Looked
up.
You
know
a
path
object
to
self.
package
or,
let's
see
what
like
was,
was
it
before?
Yes,
self.packaged
ever
solved,
it
might
might
have
been
right
before
and
I
think
we
just
might
need
to
change
the
test
case.
A
Debug,
so
it's
just
now,
it
points
to
okay,
now
points
to
that,
but
relative
to
okay
and
see
now
here
it's
trying
to
look
up
the
name,
and
this
is
what
we'll
need
to
change
when
when
we
go
to
actually
do
the
other
ones,
so
so
the
point
being
right
now.
So
if
we
were
to
do
this
in
here,
let
me
go
back
to
the
main
source.
So
this
is
the
problem.
This
is
the
other
problem
that
I
was
just
talking
about.
Okay,
so
right
now,
log
debug
right.
So
this
should
do
the
release.
A
Okay!
Well,
obviously,
yeah!
Don't
don't
get
mad
about
that,
though
this
will
try
to
do
the
release
on
on
the
main
package
right.
Okay,
we
have
to
delete
all
these
checks,
so
this
will
try
to
do
the
release
on
the
main
package
right
and,
and
then,
if
we
did,
you
know
if
we
passed
like
model
scikit,
I
would
do
a
release
on
on
model
scikit,
but
if
we
had,
you
know
if
we,
if
we
had
it
like,
we
had
like
we
just
had
it
here
with
repo
root.
A
Oops
then,
you
end
up
with
doing
the
release
on
the
main
package,
see.
A
So
I
think
I
think
it
might
just
be
best
if
we
look
at
that
test
case,
and
we
just
make
it
so
that
that
test
case
does
not
change
directory,
since
it's
meant
to
be
run
since
that
command
is
meant
to
be
run
from
there.
A
A
Okay,
yeah,
we
might
want
to
just
do
the
kiss
async
test
case
now.
A
See
so
async
test
case
now
changes
the
directory,
and
so
this
is
basically.
This
is
part
of
what
needs
to
change
here
when
we
do
the
second
party
ones
right
so,
and
I
think
part
of
this
is
right
now
we
have.
This
is
set
up
py.
A
So
what
are
we
doing?
Okay
package
name
to
directory
okay?
So
what
we
need
here
really
is
this:
we'll
just
put
it
to
do
on
this,
that
this
needs
to
change,
and,
let's
put
second
party.
A
So
save
if
we're
hoping
we're
hoping
that
the
next
release
will
be
the
one
that
we
do
second
party
plugins
in,
but
if
it's
not,
then
you
know
this
will
break
not
having
not
having
the
path
relative
like
this
will
break
that.
So,
let's
just
see,
let's
just
do
for
now.
In
this
test
case,
we
won't.
We
will
change
directory
back
to
the
to
the
you
know,
to
our
original
location.
F
A
So
here
yeah,
because
it's
just
since
this
one
is
actually
one
that's
meant
to
run
and
part
of
the
point
of
the
you
know,
part
of
the
point
of
the
temporary
directory
is
that
it
it's
supposed
to
identify
when,
when
test
cases,
you
know
are
written
in
such
a
way
that
they
are
dependent
on
the
root
of
being
in
the
root
of
the
repo,
because
most
things
will
most
of
the
time
when
we
run
some
of
the
functions
and
command
line
stuff
and
all
the
things
we're
trying
to
test,
they
probably
aren't
going
to
be
in
the
root
of
the
repo.
A
Therefore,
we
want
to
set
up
the
test
cases
in
this
way,
like
just
like
you
did
in
this
pr
such
that
they
they
they
catch
it
because
they'll
just
fail
when
they
aren't
in
the
root
and
this
one
way.
I
think
we
actually
do
want
to
be
in
the
root,
which
is
the
unique
situation
there
and
you
know.
Maybe
we
just
grab
that
repo
root
from
from
service
dev.
A
A
Okay,
alright,
so
now,
hopefully,
okay,
so
now
we
should
have
the
test
case
pass
and-
and
we
should-
and
we
should
be
okay
if
we
end
up
in
a
situation
where,
where
we
have
to
release
before
we
get
those
second
party
plugins.
A
A
Okay,
let's
see,
let's
look
at
this
again
all
right.
A
All
right
so
anything
else
that
that
you've
been
working
on
natash.
A
Yeah,
okay,
so
what
happened
there?
So
I
ran
into
something
when
I
was
doing
that
so
the
h201
this
one.
What
happened
here?
Let
me
look
at
the
the
ci,
because
I
remember
I
pushed
it.
Everything
looks
like
it
worked
and
then
also
there
was
something
weird
happening,
or
maybe
I
didn't
push
it.
Maybe
it
was
oh,
yes
and
in
the
docks
here
something
weird
you
know:
maybe
we
just
need
to
re-run
it,
because
it
did
this,
which
I
could
not
find
anywhere
on
the
internet.
A
That
knew
what
this
was
so
this
past
this
I
think
it
has
to
do
with
the.
I
think
it
has
to
do
with
something
related
to
light.
Gbm
got
updated
and,
and
it's
it's
not
for
some
dependency,
it
looks
like
das
f,
ffs
spec,
which
is
using
import
metadata
like
it
looks
like
maybe
some
kind
of
something
somewhere
might
have
a
malform
malformatted
entry
point.
This
was
a
while
ago,
like
none
of
this
made
sense
and
it
wasn't
anywhere
on
the
internet.
A
So
I
kind
of
thought:
maybe
we
just
leave
it
and
try
it
later,
and
so
now
it's
good.
You
remind
it.
So
we'll
see
we'll
see
if
maybe
because
sometimes
this
happens,
where
somebody
pushes
something
to
one
way
sub
dependency
and
it
doesn't
work
for
a
while,
so
it
might
have
sorted
itself
out
but
we'll
see
because
doesn't
make
any
sense
because
it
it
passed
the
light
gbm
build
itself.
A
So
it's
only
the
docs
build
that
failed
and
it
seemed
to
be
some
kind
of
weird
entry
point
load
thing.
So
I'm
not
sure
what
could
have
been
wrong
there.
You
know,
I
don't
think
I
think
it's
some
kind
of
dependency
issues,
so
we'll
see
what
it's
what
it
says
here.
A
A
There
is
an
exception
on
load
re-running
ci,
to
see
if
it
was
a
one-off.
A
I
think
I
tried
to
rerun
it,
but
I'm
not
I'm
not
sure.
So,
let's
see.
A
So,
okay
yeah,
I
think
you
just
needed
to
maybe
rerun
the
the
format
formatting
on
that
one
it
looks
like
that
is
just
in
need
of
a
of
a
style
update.
I
think
we
forgot
to
write
that
in
the
media
minutes.
I
forgot
to
put
that
in
sorry.
So
let's
see
this
fail.
A
A
A
Okay,
it
looks
like
that
was
I
don't
know
what
what
that
seems
completely
arbitrary
to
me.
But,
okay,
I
don't
understand
why
it
does
that
sometimes.
A
A
F
A
Reformatted,
okay
and
then
we
said,
consolidate
systems
update
test
cases,
so
I
just
needed
to
keep
release
to
service
dev.
A
All
right
great
and
then
let's
see
we
have
s
d
cindy
how
how's
things
going
with
you.
C
Yeah
I
wanted
to
discuss
about
or
two
pr's
one
was
pr1031
just
like.
Is
it
ready
to
be
months
now?
Okay
and
another
was
some
queries
related
to
pr
one:
zero,
nine,
zero.
A
Okay,
let's
see
so
what
were
those?
What's
the
title
of
those
for
the
notes
here,
one.
C
One
was
added
a
function
to
split
the
data
into
enforce
that
was
one
zero.
Nine
zero
and
replace
master
with
main
was
one
zero.
Three
one.
A
Okay,
all
right,
I
think
this
one
has
a
tag
on
it
that
says
that
it's
on
hold
too
yeah
so-
and
I
think
I
put
a
comment
in
here
so
basically-
and
this
is
what
I
was
talking
about-
and
soon
we
have
suit
haunches.
So
we'll
look
at
that,
so
this
yeah
we
need-
and
this
one
you
know
and
pressure
put
that
on
the
issue
itself,
but
this
one
we
still
need
to
do.
A
A
A
A
So
and
then,
let's
see
okay,
great.
C
Actually,
for
the
for
this
part
like,
as
you
have
mentioned,
that
the
sources
have
an
async
records
method
and
it
hits
one
record
at
a
time
so,
according
to
this
function
that
things
may
pile
up,
this
is
the
issue
right
yeah.
The
data
might
get
piled
up.
So,
like
my
approach,
I
thought
that
I
would
go
with
this
approach.
The
similar
one,
which
was
done
in
async,
define
accuracy
method.
C
A
Well,
okay,
so
see
I
don't
I
don't.
I
don't
think
this
is
gonna
be
very
so,
let's
see
I
don't
think
this
is
is
exactly
what
you're
looking
for
here.
So
I
think
that,
let's,
let's
flip,
let's
flip
over
to
your
code.
C
A
So,
let's
scroll
down
for
a
little
bit,
okay,
so
you
see
you
have
so
you
have
fold
size.
Is
you
know
len
data
divided
by
the
number
of
folds,
so
we
have
so
there's
an
open
issue
to
to
add
the
to
add
the
length
of
a
source
to
each
source.
A
A
So,
let's
see
so
you
could
use
the
sources.length
method
and
but
you'd
need
to
go
complete.
Let's
see.
A
Okay,
let
me
just
I'm
putting
in
the
notes
here
so,
okay,
so
okay
folds,
so
you
could
use
and
let
me
send
the
link
to
the
docs
here.
So
if
we
look
in
there's
a
source.
A
Source
source
there
should
be
wait
a
minute.
What
happened
to
that?
Did
we
not
go?
Do
that
we
should
have.
Let
me
find
the
issue
for
you
issues.
This
was
done.
It
was.
It
was
sort
of
half
half
half
done
last
gsoc.
A
Implement
length
method
here
we
go
870.,
okay,
use
source.,
so
you
could
use
source.length
to
get
the
length
of
a
source.
A
You
can
go
to
you
can
continue
presenting
so
here
and
if
you
go
to
the
google
meet,
I
I
sent
a
link.
If
you
flip
back
back
to
the
google
me
window
and
then
you
can,
you
can
check
out
the.
A
A
All
right,
so
we
we
added
this
length,
property
and-
and
we
did
let's
see
yeah
we
need
to
implement
within
all
the
sources
like
those
sources,
okay,
yeah.
So
basically
you
could
use
this
if
you
wanted
to
do
so
with
with,
with
with
the
algorithm
that
you're
using
right.
If
you
want
to
know,
if
you
want
to
split
the
data
into
into
you,
know,
n
mini
and
mini
buckets,
then
you
you,
then
you
need
to
know
that
the
length
of
the
source
right
and
then
you
know
how
many
buckets
right.
A
So
so,
if
you
want,
if
you
want
to
continue
doing
that
way,
you
could
use
this
dot
length
method,
but
then
you'd
have
to
go
implement
it
off
of
off
of
you
know,
you'd
have
to
go,
implement
it
on
every
source.
Right
and
the
downside
is
some
sources
since
you
you're
in
the
streaming
approach.
Let
me
just
write
this
down,
so
you
need
to
do
this
first.
A
And
so
you
could
do
this
or
you
could
you
could
you
know
slightly
tweak
things
and
and
basically
accept.
You
know
the
number
that
the
the
number
of
records
in
a
bucket
instead
right,
which
is
this
you
know
slightly
different
but
another
option.
Right
so
so
say
say
say
we
have
say:
we've
gone
with
either
of
those
right
and
and
so
you
could
go
either
path
right.
A
So
you
could,
you
could
go
implement
the
length
on
all
the
sources
which
shouldn't
be
shouldn't,
be
too
bad,
it's
another
that
would
be
completing
this
issue
right,
and
so
basically,
you
just
go
implement
that
length
method
on
every
source
that
we
have
right
now
and
you
know
most
of
them.
That
would
be,
you
know
relatively
straightforward.
A
Now
the
the
other
option
is
that
you
just
take.
You
know
how
many
records
in
in
that
you
want
in
a
bucket
as
a
parameter
right
instead
of
taking
the
number
of
buckets,
and
so,
let's
flip
back
to
your
code
here.
A
Which
is
now
you
know
so
you're
taking
fold
size
now
instead
of
fold
right.
That
would
be
the
other
option
here
and
then
you
can
so,
let's
see
sample
data.
Okay,
let's
see.
A
A
Okay,
so
you're
selecting
from
the
data
all
right.
So
then
what
you
could
do
is
you
could
you
could
iterate
over
over
the
data
right
itself
and
the
problem?
So
the
one
of
the
problems
with
this
is:
let's
see
you
could
do
some
kind
of
okay.
So
if
you
shuffle
it
okay,
if
you
don't
shuffle
it,
let's
say:
split
data.
A
A
So
if
you
don't
shuffle
the
data,
this
is,
this
is
easy
enough
right.
You
basically
just
do
an
async
4,
just
like
you
would
do
that
with
with
use
writing
the
little
calls
to
load
and
stuff
right
where
we
do
async
for
record
in,
you
know,
load
source
and
you
would
be
doing
like
sources.records
or
or
whatever
right,
and
then
you
would
just
basically
create
your
folds
and
you
would
append
the
pieces
of
data
that
you
want.
A
You
know
to
the
fold
if
you're
shuffling
it
then
you'd
need
to
work
this
a
little
bit
differently
and
like
maintain
a
little
bit
of
a
some
kind
of
a
you
need
to
maintain
some
kind
of
a
a
cache,
and
then
you
know
because
you're
iterating
it
over
the
data
one
by
one
and
so
you'd.
If
you
wanted
to
like
take
chunks
of
it
and
and
put
those
the
so
flit
split
data.
Is
your
array
of
folds
and
each
fold?
A
Has
you
know
a
chunk
of
data
right
and
and
you
and
if
shuffle,
does
shuffle
this
shuffle
within
this
context?
Do
you
want
that
to
mean
that
your
the
data
within,
like
all
of
the
data,
is
shuffled
or
do
you
want
that
to
mean?
Let's
see,
where's
shuffle
used.
A
Okay,
yeah,
it
looks
like
you're
not
using
shuffle
right
now,
so
you
could
always
worry
about
that
later
right.
So
I
think
you
know
your
solution
here
is
really
iterate
over.
You
know
async
for
records
in
in
in
source.records,
right
and
and
then
the
other
thing
you
need
to
do
is
you
know
you
need
to
inherit
from
inherit
from
async
test
case
and
start
using
the
dffml
apis
into
here
and
then-
and
you
know,
start
playing
with
with
how
that
works.
A
When
you're
using
the
dfml
apis
right
and
you're,
using
an
actual,
a
dfml
source
right
and
you
could
play
you
could
play
with
the
you.
Could
you
probably
just
want
to
do
a
memory
source
and
the
memory
source?
Let's
see
yeah
the
memory
source?
Is
there
stocks
for
that?
So
if
you
just
go
source
memory,
there's
the
docs
for
that.
You
can
see
how
to
instantiate
that
and
you
could.
A
A
A
Yeah
no
problem
try
instantiating
a
memory
source.
There
are
some
plenty
of
test
cases
that
do
this.
A
The
records
using
async
4
and
since
you
didn't
do
shuffle
2,
since
we
don't
have
shuffle
now,
since
we
weren't
using
shuffle.
A
We
won't
worry
about
it
for
now,
but
I'll
put
a
note
that
you
could
use
some
kind
of
cash
cash
to
keep
enough
records
to
randomize.
Okay,
all
right
anything
else.
On
you're
in
there.
C
A
My
doubts
are
completed
all
right
great.
So,
let's
see
here,
let's
check
in
back
with
these
guys.
A
So
it
did
that
hashem
did
that
cash
download
ci
run
pass
or
did
that
screw
things
up.
B
A
A
A
I
just
went
ahead
and
I
tried
to
I
tried
to
do
the
rebase
here
we
did
get.
We
got
some
minor
mine.
It
looks
like
some
minor
things
going
on.
Let's
see.
A
D
A
A
A
D
Okay,
so
so
we
have
actually
three
methods:
one
is
for
getting
the
temperature,
one
is
for
getting
the
population
and
one
is
for
getting
the
sales.
Okay.
A
D
So
there
is
this
create
command
which
will
actually
create
a
data
flow
source
from
the
data
set.
So
I
have
actually
created
a
data
set
in
the
csc
format
and
I've
actually
pushed
it.
Okay,.
A
A
D
Will
actually
call
the
the
the
three
methods
to
get
the
values
of
the
temperature
population
and
the
sales?
Okay.
A
D
Then
I
have
this
merge
command,
so
this
merge
command
will
actually
so
we
have
this
data
set
and
we
have
this
create
command,
so
it
will
run
the
create
command.
A
A
D
Whatever,
like
values,
we
are
getting,
we
we
will
have
to
like,
we
will
have
to
dump
them
to
a
different
csv
source
yeah.
So
that's
what
we
are
doing,
but
because.
A
D
A
A
So
and
so
this,
so
basically,
we
should
end
up
with
these
values
here,
and
these
should
be
the
columns
in
this
csv
file
that
we
have,
but
we're
not
getting
yes,
okay,
yes,
and
let's
see
so,
and
and
this
is
the
same
thing,
but
just
in
python
right
or
let's
see
this
is
the
operations.
And
then
you
also
have
the
same
thing
with
the
data
flow
in
python.
A
A
Okay,
wait.
A
D
A
A
A
All
right,
docs,
wait,
tax
examples.
C
A
For
now,
we'll
run
it
from
the
sh
file,
but
then
we'll
put
it
in
the
we'll
put
it
in
the
we'll
put
it
directly
in
here.
We
can
put
it
there
actually
here
I
can
show
that
so
basically,
what
we
would
want
to
do
is
probably
read:
data
flow,
this
game
sales
data
flow.
A
Here,
okay,
I
know
we
shouldn't
do
the
same
thing:
okay,
great
and
then
we
need
to
do
the
same
thing
here
so,
and
this
is
pre-process.
A
And
the
other
thing
that
I'm
not
sure
if
this
was
an
issue,
but
you
see
you
have
a
trailing
backslash
at
the
end
of
the
file.
We
will
just
put
this
in
another
code
block
and
see
what
happens
here
so
reprocess.
A
A
A
A
And
we
won't
run
this
for
a
second
just
to
so
this
should
be
in
oh
well.
Now
it's
going
to
be
so
dataset.csv.
B
A
A
D
You
see,
oh.
A
Okay,
yeah
oops,
my
bad
pre-processed
csv.
D
A
All
right,
yeah
pre-processed,
that's
yes,
okay,
so
and
generate
sales.
Okay,
so
and
then
we
got
the
sales
and
pre-processed
okay
saved
84
records
so
and
I'm
actually
gonna
would
just
cut
some
of
that
data
short.
A
Let's
see,
let's
cut
some
of
that
data
short
so
that
we're
not
data
flow,
so
we
don't
run
quite
as
long
I'll.
Just
I'll
go
reset
these
changes
later.
Let's
just
do
one
record.
A
Okay,
so
yeah
stupid
tag.
Okay,
I
mean
to
get
rid
of
the
attack.
Okay
turned
out
not
to
be
as
good
of
an
idea
as
we
thought
it
was
okay,
so
yeah.
So
so
clearly
the
records
don't
have
temperature
population
and
sales.
A
So
let's
take
a
look
at
the.
What
is,
I
think,
saksham
had
done,
that
the
flower
17
demo
that
uses
the
data
flow
pre-processing
source.
So
let's
take
a
look
at
what
he
did
there.
As
our
example.
A
A
Okay-
and
this
is
just
a
data-
flow
all
right-
okay,
so
that's
not
really
what
we're
interested
in.
So,
let's
take
a
look
at
the
train.
A
A
A
Oops,
okay,
so
feed
search,
temp
features.
Oh,
let's
see,
okay,
so
I
think
oh,
here's!
What
it
is,
I
think,
is
this.
It
should
be
actually
source
text,
because
this
is
the
output
features.
Here,
oh
wait!
Now
it's
not
happy
with
this
because
it
looks
like
he
has
source
images
features
and
that
looks
like
maybe
or
wait
source
images
features.
No.
This
is
the
data
that
we're.
A
A
Okay,
so
I
think
I
think
that's
probably
what's
going
on
here-
is
that
okay
generate
sales,
look
of
population
okay,
but
we
didn't
have.
We
don't
have
an
output
operation
in
here.
A
A
So
we
want
yeah,
let's,
let's
grab
those.
Let's
grab
those
operations
read:
okay,
where
was
that
that's
in
examples,
data
flow
ice
cream
operations.
A
So
our
outputs
are
just
like
okay,
they're,
just
auto
deft,
okay,
so
they're,
auto
deft
outputs
on
all
these.
So
it
would
just
be
look
up
temperature
lookup
operation
and
generate
sales.
A
A
Right
because
you
have
because
we
need-
I
think
I
think
what's
going
on
here-
is
that
we
do
an
update.
Let's
look
at
the
code
to
be
sure,
but
we
basically
want
to
make
this
get
single.
Or
what
does
this
get
single
spec
all
right,
because
this
is
the
input
good
single
spec
is
the
definition
of
the
input
to
the
get
single
operation.
A
Okay,
let's
see
oh,
this
is
the
input.
A
A
So
we
should
just
be
saying
this
should
say
that
we
want
the
okay
yeah.
These
are
saying,
select
from
seed
input
city
month,
state,
okay.
So
that's-
and
this
says
this
says
the
get
single
spec.
A
A
A
And
we're
adding
so
we're
adding
a
new
input
of
type
git,
single
spec
and
we're
saying
that
the
value
is
this
array
and
the
array
should
be.
You
know
the
results
of.
Let's
see,
do
we
reference
them
yeah,
we
reference
them
by
their
operations.
Column
thing
here,
I'm
pretty
sure.
Let's
just
go.
Look
at
the
the
data
flow
source,
real
quick.
I
know
the
problem
is
not
there,
but
just
to
confirm
it
we're
doing
the
right
thing
here,
because
we
should
basically
take
yeah.
A
So
if
we
can
add
the
output
operation,
then
then
then
it
seems
like-
and
I
believe
this
will
work,
but
I
think
I
have
the
syntax
slightly
wrong
here.
Somehow
so,
let's
see
either
that
or
it's
a
console
test
problem,
so
I
might
try
to
run
it
without
console
test.
Let's
see
we
wrote
it
could
be
a
console
test
problem.
This
is
why
I
was
trying
to
do
it
with
console
test
right.
A
One
do
one
thing
at
a
time
all
right:
let's
see
some
data
flow
and,
let's
just
see
what
happens
if
we
do
this
in
data
flow.
A
A
A
A
So
the
data
flow
looks
like
this,
and
so
what
do
we
have?
So
we
have
city
month,
state
yeah,
city,
month,
state
and
so.
A
A
If
we
look
up
population
city,
we
should
accept
definitions
that
match
that
are
coming
from
seed,
the
seed
origin
city,
the
here's,
the
list
of
definitions
that
we
should
accept
from
the
seed
origin
for
the
city,
input
of
local
of
lookout
population,
and
so
then,
when
you
add
it
yeah,
so
when
you've
added
your
docs
examples,
ice
cream
sales.
A
So
when
you
say
here
feature
city
state
month,
so
that
reads
from
the
basically
anything
from
that
csv
file,
so
it'll
it'll
pass
that
steady
state
month
columns.
They
turn
into
the
city
state
month,
definitions
which
you
you've
got
that
correct
here
and
then
so,
let's
see
what
is
this
input
generate
sales
population?
Okay
and
this
is
linked
to
the
other,
so
these
guys
are
correctly
linked
to
the
other
outputs.
A
So
then,
so
good
single,
I'm
screwing
something
up
with
get
single
right.
So
what
we're
saying
forget
single
is
that
the
spec
wait
a
minute.
No,
this
is
this
is
weight.
This
is
the
flow.
Oh,
this
is
what's
going
on,
I'm
adding
this
in
the
flow
okay.
That
makes
way
more
sense
now,
so
I'm
adding
this
in
flow,
and
I
should
not
be
doing
that.
So
I
should
be
adding
this
in
the
inputs.
A
It's
like
something
doesn't
make
sense
here,
I'm
not
modifying
the
flow,
I'm
adding
an
input
right.
Okay,
that
makes
way
more
sense.
I
was
starting
to
get
concerned,
I'm
like
what
what's
going
on
here,
I'm
trying
to
add
an
input,
but
I'm
not
adding
an
input.
I'm
modifying
the
flow
all
right.
Okay,.
A
All
right,
let's
see,
definition,
lot
of
context.
Okay,
so
we
ran
get
single
operation
so
now
get
single
ran
and
it
says:
look
at
temperature,
outputs
result,
lookup
population,
outputs,
result,
generate
sales,
outputs
results
and
it's
mad
because
it
says
operations,
lookup,
temperature,
outputs
result
or
is
it
results
with
an
s?
Let's
see
so.
A
A
A
A
A
A
We
should
probably
be
doing
something.
We
should
probably
do
something
that
when
we
run
a
data
flow
or
we
load
a
data
flow
that
doesn't
have
an
output
operation
in
it,
we
should
print
some
kind
of
warning,
because
I've
had
this
happen
to
me
a
million
times
too.
I
forget
that
I
need
to
to
actually
select
which
outputs
I
I
care
about.
Well
now
it
works
great,
so
good
status
and
check
out,
let's
just
check
out
the
data
set
and
run
the
whole
thing.
A
A
So,
and
now
I
now
you
now,
we've
got
the
skeleton
for
the
tutorial
here
so
and
we'll
make
sure
that
whole
thing
runs.
D
A
A
A
So
this
is
the
way
that
I
was
running
this,
and
this
is
I'll
just
put
this,
and
this
is
see
I
use
that
node
mod
tool.
I
I
love
that
thing.
Okay,
we're
only
slightly
over
time
for
what
ended
up
being
it's
always.
It's
always
things.
I
was
just
complaining
to
my.
I
was
complaining
the
other
day
that
that
so
this
will
basically
says
reload
on
any
changes
to
python
shell
files
or
csv
files.
A
And
then
you
know
run
that
run
that
demo.
I
was
complaining
the
other
day
that
the
worst
things
are
like
spelling
mistakes
and
stuff
like
that
and
and
and
like
we
just
spent
so
much
time
trying
to
modify
the
flow
and
we
weren't
trying
to
modify
the
flow,
we're
trying
to
add
an
input.
So
you
know
a
good,
a
good
output
or
a
good
output
of
this.
A
As
a
as
the
side
note
is,
let's
just
make
an
issue
to
say
we
should
at
least
print
a
warning,
because
this
is
this
is
something
I've
run
into
a
lot
of
times.
So
let's
say
df
memory,
or
maybe
this
should
we
do
it.
We
should
probably
do
it
on
yeah.
We
should
do
it
when
we
try
to
run
the
data
flow
because
we
instantiate
a
data
flow.
A
We
might
you
know,
call
dot,
update
and
add
an
output
operation
to
it
so
say
log
warning
when
trying
to
run
a
data
flow
without
an
output
operation,
an
easy
one
to
miss
all
right,
great
cool,
sweet,
good,
stuff,
good
stuff.
A
It's
it's
fun
to
see
fun,
to
see
your
your
progress
on
that
ice
cream.
Demo,
because
that's
something
that
we've
been
talking
about
for
a
long
time.
A
The
accuracy
stuff
I
mean
yeah,
it
looks
like
I
ran
into
a
few
conflicts
but
I'll
just
I
don't
think
there
should
be
too
many
conflicts
I'll
try
to
squash
those,
and
let
you
know
when,
when
I've
pushed
it
I'll
push
it
back
to
that
accuracy,
staging
branch
and
and-
and
I
think
you
there's
just
a
couple
things
to
update
at
the
end
of
that
right.
D
A
Be
great,
I
think
that
should
be
good
time,
hopefully
with
the
hopefully,
when
we
get
those
second
party
plugins
going,
then
we
can
have
we
can.
We
can
coincide
that,
for
a
I
mean,
we'll
probably
do
4.1
first
and
then
you
know
beta
will
probably
be
soon
all
right,
so
ice
cream
sales
demo
found
we
were
missing
a
output
operation-
oh
you
know,
and
and
what
else
it's
gonna
it's
gonna
get
mad
about
this
too.
So,
let's
just
add
this,
because
we
have.
A
A
A
E
One
one
more
thing:
john
yeah,
the
ci
is
taking
too
much
time
to
start.
I
don't
know.
E
Yeah
it's
at
h2o
and
I
think
there
is
a
conflict
in
async
test
case
console
to
consolidate
the
test
resistance.
A
Oh,
that's
annoying
gotta
hit
that,
let's
see
resnet,
what
did
we
touch
in
resident?
Oh,
it
must
be
cash
download.
I
just
said
that
all
right,
can
you
fix
that
conflict
and
and
push
it
up
so,
okay,
it
should
just
it
should
be
probably
a
simple
conflict
to
fix
with
the
it's.
Probably
just
you
know
some
similar
lines
changed
around
a
setup
function
so
just
make
sure
that
we're
we're
doing
here
we
can.
What
is
this
right?
So
this
is
how
ghpr
check
out.
A
E
A
Okay,
oh,
I
grabbed
the
one
that
was
auto
merged,
okay,
yeah,
so
basically
oops
yeah
right.
So
the
weight
is
the
syntax
that
we
want
right.
So,
oh
and
it
looks
like
you
know,
so
we
we
want
async
test
case
right.