►
From YouTube: Weekly Sync 2020-06-23
Description
Oh no! John forgot to start recording until about halfway through :,(
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.wt0a12bx9ka2
A
B
A
A
A
And
an
async
iterator
of
record
objects
so
and
then
the
other
thing
we
need
is
that
we've
got
to
have
like
you
know
what
the
health
features,
what
we're
working
on
here.
So
second
input
is
list
actually
yeah.
Second
input
is
list
of
features
because
I
think
I
mean
for
now.
All
our
models
predict
one
feature,
but
you
know
who
knows
maybe
we're
going
to
have
a
model
in
the
future.
That
predicts
multiple
features.
B
A
Okay,
because
all
you're
working
from
now
I
mean
at
this
point,
all
you
have
to
work
from
is
like
all
you
can't.
You
can't
work
from
anything
within
the
model
right
you're.
Only
working
from
like
you
don't
have
the
models
configure
anything
so
you're.
Just
looking
at
you
just
have
the
list
of
features
that
were
or
well
yeah.
This
should
work
fine.
The
attitude
work
line
because
you're
just
going
to
look
at
the
record
feature
data
and
then
the
record
prediction
data
for
the
value.
C
A
A
A
A
A
A
A
It's
probably
gonna
be
very,
very,
very
similar
to
how
we
did
the
the
the
decide
kit
models.
Only
basically
we're
just
gonna,
like
you
know,
throw
out
you're
gonna
you,
it's
basically
gonna
be
like
that
number
I
create
config,
and
then
we
throw
out
anything
that
doesn't
have
a
default
value
for
the
config
we
use.
We
use
things
that
do
have
default
values
within
the
config.
A
A
C
A
A
A
A
D
C
D
A
D
A
A
D
D
A
D
D
D
C
A
D
A
E
A
E
A
D
A
Basically,
if
you
wanted
to
change
the
definition,
so
so
the
reason
why
is
if
you
wanted
to
change
something
in
the
definition,
you
would
have
to
go
through
and
change
every
instance
of
it
within
every
operation.
So
if
one
operation
right,
if
we
have
like
a
linked
data
flow,
that
was
you
know
like
say,
we
did
Auto
flow
right
and
just
all
the
for
example
like
we
should
I,
if,
with
the
should
I
data
flow
here,
let's
see.
A
A
We
have
pi,
PI
package.json,
goes
to
pi
PI
package
contents
and
safety
check,
and
and
and
so
if
you
look
at
the
code-
where's
the
code
yeah.
So
if
you
look
at
the
code,
pi
pi
package
context
is
defined
or
no
snooze.
It
always
gets
me,
so
the
input
is
auto
discovered
here
as
the
URL
right.
So
we
just
specify
the
output
is
going
to
run
bandit
right
and
then
we
use
the
input
here
or
wait
and
wait
a
minute
five
of
hackage
contents.
A
The
input
is
the
URL
okay
yeah,
so
we're
gonna
use
a
package.json,
outputs,
the
URL
right
and
so
basically
right.
So
in
the
case,
where
we're
using
definitions
from
other
operations
like
this
right
rather
than
redefining
them
all,
then
we
have
to
go
through
in
the
exported
yeah
mole
and
change
every
single
one
of
these
right
in
each
place
and
if
I
didn't
change
them
now,
the
whole
thing
is
gonna,
be
right:
it's
not
gonna
work
correctly
right.
A
So
if
you
do
it
linked,
then
it's
always
going
to
reference
it
by
name,
which
means
you
only
have
to
change
the
information
about
the
definition
in
one
place.
Right
so
say,
I
wanted
to
make
one
of
these
resources
locked
then
I,
just
change
it
yeah
up
top
Thanks,
okay,
so
yeah
see
uh-oh
I
think
we
just
realized
that
this
might
not
actually
clean
up
that
directory.
A
A
A
A
A
A
You
should
subclass
from
integration
CLI
test
case,
and
this
is
what
we're
talking
about
live.
Do
you
guys
remember?
Last
Friday
I
was
saying
that
we
really
need
to
make
this
the
base
class
for
everything,
and
this
is
why?
Because
you
know,
then
we'll
end
up
with
everything
having
a
temporary
directory
and
we'll
end
up
with.
You
know
the
fact
that
we
can
do
these
required
plugins,
for
example,
if
you
need
a
good
grab.
A
A
A
A
A
D
C
D
A
A
A
A
D
A
A
A
A
A
D
A
A
D
Working
I
tried
that,
but
the
thing
is
that,
if,
if
in
lis,
if
the
day,
if
the
directory
source
is
labeled
like
label,
four
folders
are
there
then
and
in
label
folders
this
file
names
in
each
label.
Folder
is
one
dot,
PMG
till
hundred
dot
PNG.
So
for
that
create
an
issue
because
then,
if
there
are
ten
levels,
then
there
will
be
ten
records
with
same
keys
as
one
dot
PNG.
A
A
Okay,
what
is
the
input
data
for
these
models?
Look
usually
look
like
like
what
is
a
model
that
you're
looking
at?
Is
there
you?
Basically
okay?
So
let
me
just
let
me
just
recap
on
what
we're
doing
here,
because
I
want
to
I
want
to
make
sure
that
I
understand
what
what
you're,
what
you're
doing
so,
you
were
going
to
take
classification
model
like
a
psychic
classification
model,
and
you
are
going
to
take
these
directories
of
labelled
images
right
where
the
it's
just
a
list
of
images
in
numbers.
A
D
A
D
C
A
A
D
If
there
is
unlabeled
data,
if
the,
if
there
are
no
label
folders-
and
they
are
just
n
number
of
images,
so
according
to
that,
if
condition
the
label
will
be
deleted
from
features,
but
if
there
is
a
CSV,
sometimes
there
is
a
CSV
source
with
a
label
data
giving
file
names
and
labels
mapping
them.
If.
B
A
D
D
A
But
it
sounds
like
this
is
the
case
to
me
where
we
need
to
add
that
parameter
to
sources
where
we
tell
it
not
to
merge
the
records,
because
or
else
you're
just
going
because
right
by
default,
the
source
sources
goes
through
and
it
merges
all
the
records
right.
So
it
takes
the
first
source
and
it
says:
okay.
Is
there
any
more
feature
data
for
this
record
in
any
of
the
other
sources
right?
And
but
it
sounds
like
in
this
case
you
have
some.
A
D
So
there
there
are
two
things
happening
here:
if
the
directory
source
is
segregated
into
a
label
folders,
this
will
work
fine.
This
will.
This
will
go
fine
and
label
the
images,
and
if
there
is,
if
the,
if
they
are
not
segregated
into
label
folders-
and
we
provide
another
CSV
source
with
key
as
file
name
and
the
labels,
it
will
merge,
it
will
map
them.
Yes,.
A
D
A
Yeah
yeah,
okay,
now
I
see
what
you're
saying
so,
but
then
that's
where
I
think
you
could
just
have
a
switch
that
says,
use
the
image.
Hash
use
the
hash
of
the
image
right
so
because
we're
talking
about
the
first
case
is
where
we
have
two
sources
and
we're
appending
one
source
to
the
other
source
and
we're
using
all
of
it
right
is
that
case,
one.
D
A
A
A
D
D
A
D
A
D
D
A
A
A
A
A
C
A
A
D
A
A
I
mean
it's
possible
right,
but
how
long
is
it
going
to
take?
You
is
the
question
right,
so
I
mean
what
is
more
important
here
is
look
you
look
at
the
timeline
of
your
project,
right
and
and
and
what
you
want
to
do
and
make
sure
that
you
like,
because
this
is
so
basically
the
thing
is,
you
can
write
config
files,
but
so
you
can
write
config
files
or
you
can
write.
You
know
just
dictionaries
that
contain
all
the
properties
that
you're
going
to
need
right,
but
then
you're
gonna
end
up
in
this
similar
situation.
A
That
kind
of
test
wrapper
that
we
talked
about
earlier
right
where
you're
gonna
have
to
you
know
everything's
gonna,
be
you
know,
you're,
probably
gonna
find
variations
on
things.
Right
so
probably
makes
sense
to
go
through
and
do
a
few
more
of
these
and
understand
what
are
what
are
the
types
of
things
that
you're
going
to
need
to
happen
here?
Let's
see,
let's
just
see.
D
D
D
C
A
D
A
And
yeah,
okay,
so
yeah
I
mean
this
is
really
the
really.
The
question
here
is
like
how
much
work
you're
gonna
have
to
do
and
you're
you'll
figure
that
out
right
is.
Does
it?
Is
it
more
work
for
you
to
write
this
thing
that
you
parse
some
config
than
to
just
go
through
and
write
each
each
one
right,
because
the
goal
here
is
we
want
to
be
able
to
expose
a
lot
of
these
things
right?
Yes,
so
yeah
all
right,
cool,
let's
just
say.