►
From YouTube: Weekly Sync 2021-06-22
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.vmzdnj9ofgsn
A
There
was
all
right:
oh
that's
what.
A
A
A
A
A
C
I
just
completed
all
the
bugs
on
land
comment.
One
there's
still
one
one
one
serial
error
that
I
want
to
talk
about.
That
is
a
around
the
mac
os.
It
is
not
getting
some
child
watcher.
I
don't
know
it
should
be
having
an
event
loop,
but
it
looks
like
it
doesn't
have
something.
C
Other
thing
is
that
I
like,
as,
as
you
told
I
changed
the
article
thing
into
an
operation
and
I
have
pushed
a
sample
code,
some
sample
code,
so
that
I
can
just
get
some
input,
and
I
also
want
to
talk
about
how
to
test
it,
because
there
are
two
ways:
one
is
like
patching
everything.
Another
would
be
like
creating
actual
files
for
testing
and
trying
to
compress
them
and
decompress
them.
C
A
Okay,
all
right
and
then
let's
see
hashem,
so
what
are
you
or
anything
else
from
you.
C
A
Sounds
good
sounds
good,
yeah,
good,
good
luck
with
finals;
it's
always
tough,
all
right!
So
hashim.
What
do
we
got
for
you
today.
D
D
I
I
showed
final
predictions
instead
of
the
accuracy
and
stuff
okay,
so.
A
A
A
D
Because
how
we
cannot
effectively
tune
the
models.
A
Okay,
okay,
so
all
right,
let's
see,
I
need
to
figure.
A
A
D
Yeah
there
were
some
cr
errors
that
I
wasn't
sure
about
on
the
test.
Notebooks.
A
Okay,
there
you
go
okay
seriously,
you
weren't
sure
about
on
the
test.
Notebooks.
E
A
A
A
A
Really
it
named
it
with
txt
still
all
right,
great
good
job
windows.
Let's
see
all
right
so
testing.
A
A
A
A
A
A
I
bet
it
is,
I
bet
there's
just
a
okay.
What
happened.
A
All
right,
so
let
me
make
a
note
of
that,
so
we
it
looks
like
js
is
formatting
code
differently
than
it
used
to.
A
Let's
see
it
looks
like
there's
a
new
version
from
what
I
had
so
quite
quite
possible
that
it
did
indeed
get
weird.
So,
let's
see
where
is
this
so
yeah?
A
A
All
right,
okay,
it's
failing
for
me
too!
So
my
guess
is
that
there's
an
updated
yeah,
it
looks
like
they've
updated
something
so.
A
A
That
was
really
annoying.
They
had
this
bug
where
you
would
say
just
put
the
new
line
at
the
end
of
the
file,
and
it
wouldn't
do
that.
So
that's
annoying
all
right.
So
we'll
just
commit
this
to
the
master
branch
style.
A
A
Well,
let's
see,
I
think
if
we
just
update
it
throughout,
let's
see
june
or
june,
14th
yeah,
it
must
be
to
be
a
cache.
What
do
you?
I
guess
I
don't
quite
understand
what
you're
saying.
A
All
right,
so,
let's
grab
that
commit
and
push
it
so.
A
A
Fixes
this
rebase
master.
A
All
right
to
branch,
okay,
great!
So
if
you
rebase
master
into
the
pr
branch,
then
that
should
be
fixed
okay,
and
so
this
is
the
mac
os.
Okay.
So
now
we
have
the
mac
os.
One
could
not
add
child
handler.
Oh
great,
this
book.
Okay,
this
thing
is
annoying
all
right.
So,
oh
god,
I
hate
this
stupid
thing.
A
Okay,
this
happens
when
you're
trying
to
run
sub
processes
from
an
event
loop,
that's
not
in
the
main
process,
usually
or
is
not
the
main
event
loop,
which
is
weird
that
we
haven't
run
into
this,
but.
A
Yeah,
but
it's
different
different,
it's
different
across
just
because
it's
passing
on
linux
doesn't
necessarily
mean
that
it
would
pass
on
on
mac
or
windows,
because
because
this
is
I
mean
so,
okay,
so.
C
C
A
A
A
A
If
we
don't
do
it
within
a
generator,
let's
see:
let's,
let's
just
do
this
real
quick,
should
validation,
value.
A
A
Okay,
just
for
shits
and
girls,
I
doubt
actually.
I
doubt
this
is
the
case,
but.
A
But
I'm
curious
now
so
so
mac
os.
A
A
All
right
so
yeah,
okay,
great
and
we
don't
have
a
mac
os
system
to
really
test
on
here.
What
could
we
do
so
could
not
have
child
handler?
Child
watcher
does
not
have
a
loop
attached.
Okay,.
A
A
A
A
All
right,
so
you
know
be
written
okay,
so
these
would
these
should
be
written
in
all
caps.
Was
there
a
case
where,
like
changelog,
should
always
be
in
all
caps?
So
is
there
a
case.
C
A
C
A
And-
and
let's
just
also
so-
let's
not
get
too
far
down
into
this
too
right,
because
I
think
I
think
that
okay,
so
the
functions,
slash
class
names,
you
know
we
don't
want
those
to
fail
to
fail
a
validation
right,
the
the
test,
so
the
test
one.
We
need
to
make
sure
that
we
do.
A
We
need
to
make
sure
that
we
do
number
two.
We
need
to
make
sure
that
we
do
number
three.
I
think
we
can
wait
on
four
just
because
that
and
we
need
to
do
five
so
actually
yeah.
We
need
five
so
because
the
the
other
thing
is,
you
know,
there's
there's.
A
This
is
just
like
a
most
like.
We
want
to
catch
90.
You
know
we
want
to
catch
a
lot
of
things
right.
We
don't.
We
don't
need
to
catch
everything,
because
at
the
end
of
the
day
like
we
can
just
check
the
ci
job
and
see
why
it's
failing
right
and
we
can
just
still
merge
a
commit,
even
though
it
fails,
but
I
think
that
I
think
that
so
do
you
think?
Do
you
see
this
as
something
like?
What
do
you
like?
C
Probably
three
would
take
some
time
one
and
two
could
be
done
pretty
fast.
Okay,.
A
So,
let's
just
do
one
and
two
then
and
then
we'll
merge
this
in
and
we'll
create
an
issue.
So
let's
see
so
we'll
create
an
issue
to
track
the
implementation
of
three.
So
okay.
A
Create
an
issue
for
we'll
implement
one
and
two
within
the
pr.
A
Will
track
implementation
of
three
through
five
within
the
issue
you
can,
I
mean
you
can
just
you
know,
make
it
like
a
checklist
right
and
then
so.
Okay,
let's
make
pick
the
issue
or
I'll
just
create
the
issue
right
now,
so
we
see
what
I'm
talking
about
so.
E
A
Details
all
right
and
we'll
call
this:
what
service
dev
lent
commits
right.
A
Okay,
so
we'll
track
these
guys.
A
A
A
All
right,
I'm
just
gonna,
look
at
this
here.
Okay,
great,
so
we've
got
this
and
let's
see
so
we'll
check
this
out.
Actually,
okay,
if
you
submit
this
issue,
then
you
can.
If
you
submit
this
issue,
then
you
can
check
them
off,
but
if
I
submit
it
you
won't
have
permissions
to
check
them
off.
I
just
realized.
So
let
me
just
paste.
C
A
A
Great,
so
all
right
so
and
then
okay,
so
we're
still
so
we're
seeing
this
mac
os
issue
right.
So
let's
take
a
look
and
see
what
happened
there
in
that
ci
run.
A
C
C
That's
why
I
was
asking
like
what's
the
solution.
It
was
probably
in
some
of
the
download
progress
volunteers
that
I
had
the
same
issue
with
mac
os
and
later
on.
I
changed
some
something
like
I
just
refactored
the
same
test
with
same
functions
and
it
worked.
A
A
So,
okay,
let's
see
so
why
are
you
mad?
It
is
matte,
obviously,
because
they
create
some
process
stuff.
So.
A
A
A
C
So
I
I
think
it
may
be
like
in
the
test
itself.
We
are
making
an
object
of
the
comment
class
which
might
not
have
a
loop
attached
to
it,
because
it
is
not
being
called
from
the
run
rather
than.
C
Can
that
be
a
thing
you
need.
A
A
Yeah
I
saw
what
you
were
talking
about.
Let's
see
yeah
you
create
an
instance
of
this
class.
Where
is
that,
where
you're
just
saying
right
at
the
class
scope
in
the
test
case.
A
An
issue:
okay,
yeah
see
that.
A
A
A
A
A
So,
let's
see
okay
yeah,
I
doubt
it's
that
generator.
A
Thinking
thinking,
we
might
want
to
try
to
go
ahead
and
create
a
new
event
loop
at
the
class
scope
here.
A
A
A
A
A
A
A
A
A
A
A
A
Let's
think
about
this
a
little
harder,
so
it's
mad
because
the
we
cannot
add
child
handler
child
watcher
does
not
have
loop
attached,
okay.
So
the
question
really
in
my
mind,
is
why
have
we
not
run
into
this
before
because
we're
calling
out
subrasus's
other
places,
because.
C
A
A
A
Okay,
but
we
patched
subprocess
exec,
that's
why
that
explains
why
we
didn't
run
into
this
issue.
So
it's
something
wrong
with
the
test
harness.
A
I
wonder
if
switching
to
let's
just
command
sequence,
yeah,
so
we're
patching
so
process
exec
here
all
right.
We
spent
a
lot
of
time
on
this.
Let's
see.
A
So,
let's
see
for
a
second,
why
do
we
need
let's
check
out
why
we
need
to
run
those
git
commands.
A
Good
log,
all
right,
let's
see
what
ls
tree
so,
and
this
is
the
extension
so
yeah,
that's
right,
didn't
we
talk.
I
think
we
talked
about
this
last
week
too.
So
so
we
obviously
you
know.
Ideally
we
fix
the
bug
with
the
test
cases,
but
they
think
io
related
bugs
can
be
very
tricky,
especially
when
there's
something
like
this.
That
is
not
even
the
same
across
platforms,
and
I
would
assume
none
of
us
have
a
mac.
So
unless
that
assumption
is
wrong,
does
anybody
have
a
mac.
A
B
A
Yeah,
so
in
that
case,
let's
try
to
fix
the
bug
by
you
know:
changing
the
behavior
so
or
work
around
the
bug.
So
this
get
all
extensions.
Basically,
you
grab
every
file
and
you
see
good,
let's
see
well,
let's
yeah,
so
you
grab
give
constitution.
C
C
A
A
Okay,
so.
A
A
A
A
A
A
Okay,
yeah,
it
would
be
really
great
if
we
didn't
need
to
run
git
to
validate
a
commit
message.
C
A
Yeah
so.
C
A
Yeah
in
our
glob
and
then
grab
all
the
extensions
yeah.
So
let's
see
youtube
exists,
I'm
just
thinking.
Is
there
a
way,
so
mutated
paths
execute
funk
list:
okay,
okay,
okay,
so
and
then
you
have
a
tuple
like
that.
A
Could
you
work
down?
No,
you
can't
work
down.
Okay,
yeah,
I
don't
know
yeah.
Let's
just
go
for
our
glob
here
and
grab
the
list
of
all
the
extensions,
because
you
should
just
be
able
to
do
yeah
our
glob
and
then
yeah
you.
You
can
just
do
our
up
and
grab
all
the
extensions
because
get
ls
tree.
A
Yeah
yeah,
okay:
let's
just
do
that
and
so
we'll
just
say
that,
because
debugging
this
without
an
os
x
system
is
going
to
be
a
huge
pain
in
the
ass.
So
but
let's
maybe
submit
an
issue
to
track
it
or
something
so.
C
But
in
itself
it
is
an
issue
right
so
other
than
this,
like
in
general.
This
is
an
issue.
A
A
Yeah,
it's
probably
an
async
test
case.
So,
let's
just.
A
A
All
right,
util.
A
Testing
sync
test
case
issues
on
os
x
with.
A
A
A
So
we'll
use
pathway,
dot
path,
dot,
r
glob
to
get
file
extensions
instead
of
using
git
ls
files
or
ls
tree.
What
was
it
whatever?
It
was
yeah
that
was
true,
yeah.
A
We'll
track
os
x
async.
I
o
testing
issue
in
this
issue.
Great
all
right,
so
all
right
anything
else
on
that
or
you
you
can.
You
can
make
progress
on
that
now.
C
Yes,
I
can
one
with
that.
The
other
thing
just
about
test
is,
I
mean,
related
to
a
project.
Okay,
11
28,.
B
A
A
A
A
You
don't
want
the
input
directory
path
in
the
output
file
path
to
both
be
the
same
definition,
because
the
way
that
this
works
is
you'll,
create
you'll
end
up,
creating
an
infinite
loop
or
oh,
wait.
That's
not
the
output,
okay,
yeah,
so
the
thing
is
input
directory
path
would
be
you
you
need
it
to
be.
Okay,
so
the
way
that
the
definitions
work
right
now
is,
let's
see,
where's
the
link
on
that.
A
Okay,
so,
okay,
so
we
want
to
use
let's
use
different
definitions
for
input,
directly
path
and
output
file
path.
B
A
Yes,
so,
and-
and
that
would
be
just
like
a
string
in
this
case,
because
when
we
think
about
primitives,
it's
a
primitive
data
type
is
not
like
path:
lib.path
object,
it's
just
a
string,
something
that
could
be.
You
know
if
we
were
to
transport
it
over
the
network.
What
would
we
serialize
it
to
right?
A
We
would
serialize
it
to
a
string,
so
let's
use
different
definitions
for
input,
directory
path
and
output
file
path,
so
just
because
so
you're
not
thinking
so
so
you
want
to
you
want
to
try
to
make
the
definitions
as
specific
as
the
as
specific
as
possible.
Essentially
so,
because
the
the
input
directory
path
right-
that's
it
you
wouldn't
if
the
input
directory
fast,
is
different
than
the
output
file
path
right.
A
You
see
what
I'm
saying,
because
you
would
want
to
use
so
each
each
thing,
that's
a
definition
should
be
able
to
be
used.
You
know
wherever
that
is
right.
So
if
I
were
to
define
this
as
a
zip
file
right,
if
I
say,
output
directory
path
is
a
zip
file,
then
anywhere
like.
If
I,
if
I
had
a
zip
file
in
the
in
the
in
the
data
flow,
it
would
say:
okay,
where
is
a
zip
file
used
right
and
it
would
say,
oh
well
I'll
put
file
path.
Is
a
zip
file?
A
Okay,
so
this
now
I
know
I
need
to
pass.
You
know
that
that's
where
I'm
going
to
use
that
zip
file
input
object,
but
if
you
just
have
a
path,
it
would
say:
okay,
oh
there's
an
input
directory
path
and
there's
an
output
file
path.
So
let
me
just
take
anything:
that's
a
path
and
put
it
there
and
then
you're
going
to
end
up
with
so.
A
Yeah,
you
would
probably
want
to
make
it
different
for
different
formats
yeah.
So
so,
let's
just
say,
and
options
is
also
too
vague,
so
yeah
and
let's
not
do
options
right
now.
So,
let's,
let's
not
do
options
at
the
moment
because
remember
we're
going
to
wait
on
those
and
decide
whether
we
wanted
to
use
operation
configuration
like
the
standard
way
of
configuring
things
or
whether
we
want
to
use
them
as
inputs
right.
So
we'll
wait
on
those
for
now.
A
Definitions
for
inputs
as
specific
as
possible,
if
that
def
think
about
it
like.
A
In
the
same
way,
somewhere
else
by
another
operation,
for
example,
a
operation
that
so,
for
example,
your
unzip
operation
would
take
a
zip
archive.
However,
it
wouldn't
know
what
it
wouldn't
be
able
to
unpack
a
directory.
A
Therefore,
the
you
know.
Therefore,
we
know
that
a
definition
of
a
path.
A
A
A
All
right
and
then
writing
test
wise.
I
would
say
you
can,
let's
see,
extract
all.
A
Okay,
so
okay
and
then
this
stuff
is
the
old
stuff,
so
so
yeah.
So
let's
write
some
test
cases
and
let's
see
we'll
write
some
test
cases
and
we
will
use
so
mock
or
not
to
mock
or
not
mock.
A
Yeah,
you
know
you
could
use
unit
test
mock.
You
know
just
verify.
C
A
Yeah,
let's,
let's
use
mock
so
for
testing.
Let's
use
unit
test.mock
rather
than
creating.
A
Archives
reasoning
behind
this
that
some
archive
formats
can
be
tricky
to
make
reproducible.
A
A
This
this
fake
process
stuff,
so
here
we
build
a
tar
file
and
I
think
we
had
to
mock.
Oh
no,
we
didn't
mock
the
when
there's
gzip
compression
on
the
tar
file.
You
have
to
like
mock
the
time
library
which
is
annoying
and
I
think
zip
file
might
have
something
similar.
A
Yeah,
who
knows
not
a
big,
oh
test
net,
this
one.
D
On
all
right,
can
I
share
my
screen
and
show.
D
A
D
Okay,
so
this
is
the
notebook
that
I
just
pushed.
I
actually
used
a
different
approach
and
since
you
were
talking
about
keeping
it
precise,
I
excluded
a
lot
of
things
in
the
notebook.
D
The
data
it
goes
straight
to
you
know,
assembling
and.
D
I
I
just
wanna
just
wasn't
sure
if
we
wanna
go
down
that
road
and
you
know,
try
to
tune
different
models
and
see
if
you
know
we
get
the
better
accuracy
and
try
different
data
sets.
But
I
can
do
that.
A
You
know,
I
would
say
yeah
I
mean
so
so,
let's
wrap
up
the
tutorial,
as
is
let's
make
sure
that
we
include
the
accuracy
stuff
right,
even
though
it'll
show
that
it's
worse
and
then
we
can
pick
a
better
data
set.
You
know
in
the
future
right.
So,
let's,
let's,
let's
see
so
what?
What
is
our
timeline?
What
week
is
this.
D
It's
it's
not
an
issue
for
me.
I
can.
I
can
do
it
right
now
as
well.
You
know
yeah.
If
you
want
to
go
down
that
road
might
as
well
just
do
it.
A
Yeah,
I
think
that
it
would
be
good
because
we're
into
week
this
is
week
three
now
right
so,
and
you
have
you
have
like.
Does
that?
Does
that
put
you
behind
on
your
schedule?
I
think
you're
still
pretty
on
track
with
that
right,
because
don't
you
have
you,
you
combined
two
of
them
into
one,
and
then
this
is
the
third
one,
so
you'd
still
be
on
track.
If
you
worked
on
this
this
week,
right.
A
Okay,
so
let's
do
that
then:
let's,
let's:
let's
try
to
find
a
data
set
that
that
you
know
performs
better
right
and
you
can
always.
D
D
A
Yeah,
I
think
that
was
good
right,
because
then
we
could
say
hey
you
know
we
want
to
create.
You
know
we
we
tested
a
few
and
we
saw
that
these
are
probably
going
to
screw
it
up,
and
you
know
now
I
don't
know.
Maybe
I
mean
you
know,
there's
a
possibility
that
they're
accurate
in
ways
that
the
other
ones
are
not.
I
don't
know.
A
D
On
you
know
what
type
of
tutorial
you
want,
because
I
mean
we
have
already
gone
through
the
accuracy
part
other.
D
But
if
we
wanna
include
it
here,
it's
okay,
you
can
do
that
as
well.
A
I
mean
I
mean,
I
think,
that
the
moral
of
the
story
is
right.
Let's
have
an
example
that
shows
how
the
accuracy
increases
with
the
ensemble
model
right,
so
because
that's
kind
of
the
take-home
point
here
right,
so
so,
whether
that
means
including
them
or
not,
including
them
or
whatever,
right
or
changing
the
data
set,
let's
just
figure
out
a
way
to
make
it
make
that
you
know
make
it
make
sense
to
the
end
user
because
they
see
that
the
actually
goes
up
right.
D
D
It
says
that
the
model
was
not
found
and
the
list
only
contains
slr.
Okay,
when
it
runs
the
cell
containing
the
loading
models.
Part.
A
Okay,
oh
so
larv's
not
found
nslr.
Okay,
it
looks
like
the
scikit
models
aren't
installed.
Oh,
I
think
this
is
just
a
oh
yeah.
We
probably
so
there's
the
test
run
twice,
so
it
runs
once
without
any
models
installed
and
another
time
with
with
all
the
plugins
installed.
Let's
see,
where's
the
documentation,
let's
say
testing
all
right
so
yeah,
so
we
just
need
to
use
that
required.
Plugins
thing.
A
So
it's
just
this
call
to
you.
Basically
change
modify
the
test
case
to
say.
Take
a
look
at
that
link.
I
just
sent
in
the.
A
A
So
this
will
essentially
raise
a
skip
if
it's
not
installed
and
then
because
the
main
package
runs
test
twice
so
the
first
time
it
runs
tests
it.
It
doesn't
have
any
of
the
plug-ins
installed
just
to
make
sure
that
you
know
the
main
package
works
without
plug-ins
installed,
except
for
skips,
like
you
know,
intentional
skips,
so
we
don't
have
any
errors
and
then
it
installs
all
the
plugins,
and
then
it
runs
the
test
again.
So
this
is
see
that
required.
Plug-Ins,
cfo,
config
png.
A
A
A
A
A
So,
let's
see
okay
yeah
because
test
nope-
oh
I
see
so
because
the
test
notebooks,
oh,
like
the
test,
notebooks
quite
likely
require
model
psychic.
Okay,
so
I
see
I
see
where
the
confusion
is
okay,
so
what
we
should
do
here
is
let's
put
in
that
class
test
notebook,
okay,
and
where
is
this
so
okay,
yeah?
And
that's
not
documented.
So
let's
document
that
okay,
so.
A
A
Okay,
I'm
updating
the
docs
okay,
so
you
need
to
do.
Let's
see.
Where
is
that
other
one?
So
in
test
notebooks,
I
think
we
have
this
required
plugins
yeah.
So
if
in
test
notebooks,
you
would
want
to
say,
like
the
test,
notebooks
class
you'd
want
to
say
required.
Plugins
equals
dfmo
model
scikit,
and
I
put
that
in
the
chat
too,
like
as
a
class.
A
I
variable-
and
this
is
like
a
basically
a
way
of
saying
so
instead
of
at
a
test
case
basis,
it's
at
a
class
level.
So
if
so
I'll
just
write
this
in
the
docs
too.
So
if
all
if
all
functions
or
all
methods
within
a
class
require
a
plugin
or
set
of
plugins.
A
A
All
right,
I've
updated
the
docs.
So
let's
see
so
yeah
you
would
just
add
that
to
test
notebook,
because
your
notebooks
are
going
to
require
psycit
and
so
it'll
it'll
make
sure
to
run
them.
It'll
it'll
skip.
D
D
A
So
actually
boost
could
not
be
loaded
okay,
so
I
think
this
is
the
same.
Is
this
the
same
thing?
Yeah
might
be
the
same
yeah,
so
we
might
just
want
to
add
it
to
is
this
also
in
a
notebook
test
case
is
this:
is
this
notebook?
That's
maintained,
yeah
yeah,
all
right.
So,
let's
add
that
yeah
that
set
of
required
plugins
well
done.
D
And
I
think
you
were
just
discussing
this
one.
A
And
it's
fixed,
it
should
be
the
style
one
yeah,
so
that
would
be
a
pull
from
master.
Yes,
so
or
a
rebase
from
master.
So.
D
Okay,
as
a
follow-up
to
our
101
one-on-one,
I
also
wanted
to
discuss
a
little
about
multi-output
models.
If
we
have
time
yeah
sure.
D
So
the
list
was
more
about
multi-class,
multi-output,
classifier
models
and
and
then
you
have
some
multi-auto
models
that
only
support
binary
classification
and
there
are
some
models
that
support
multi-class
classification-
okay
in
multi-output.
So
these
are
the
only
models
that
support
multi-class,
multi-output
classification,
okay,.
A
So
I
mean
yeah,
I
think
that
you
know
the
the
strategy
stays
the
same
here
right,
it's
it's!
Basically,
you
have
a
different
entry
point
for
each
one.
If
you
can
get
them,
you
want
to
target
target
the
ones
that
you
can
wrap
with
the
same
abstraction
right.
So
if
you
can
wrap
six
of
those
right,
then
that's
going
to
be
your
your
target
is,
is
six
of
those
you
know
using
the
same
sort
of
code
right.
A
E
E
E
Okay,
so
so,
let's
start
so,
I
will
get
get
you
through
the
code
because
you
haven't
probably
seen
it.
So
this
is
the
base
class,
the
scorebase,
where
we
have
the
scikit
scorer
config.
E
So
all
these
methods,
I'm
actually
creating
a
context
for
it
and
I'm
then
subclassing
it
from
the
base
class
and
then
I'm
adding
it
to
the
entry
point
great
great,
so
the
I've
also
written
tests
for
it.
So.
A
E
Is
actually
one
of
the
tests
I
have
written,
but
actually
I
have
written
like
there
are
like
300
tests,
which
I
have.
E
Yes,
I
get
right
model,
so
the
problem
which
I'm
facing
here
is
when
so
so
here
I
am
facing
the
problem.
E
A
Yeah,
so
I
would
say:
oh
man
there's
several
ways
to
do
this.
I
can't
remember
which
one
is
the
best
way.
Does
anybody
have?
Does
anybody
have
a
let's
see
so
one
way
I
usually
do
is
is
funk
tools
partial,
let's
see,
there's
a
way
to
unbind
it
from
the
class
cell,
but
I
can't
remember
which
one
is
better
so.
E
A
Adding
functions
to
classes
dynamically,
so
let's
see
so
that
score,
I
think
so.
Where
do
you
add
self.score
to
the
to
the
class.
E
Okay,
so
I
actually
added
to
this
test,
so
here
I
am
providing
the
class
here
and
when
it
actually
sets
up
the
mod
well
and
it
changed
it
so
here
I
am
doing
it,
providing
it
so
class
dot
score
and
then
in
the
accuracy
we
have
providing
it
here.
A
So
the
fourth
line
up
in
the
stack
trace,
so
the
very
very
top
line
in
the
stack
trace.
We're
looking
at
check
consistent
length,
not
not
not
the
one
we
could
see
so
basically,
the
okay.
So
let's
see
so
self.score
y
true
y
pred.
A
So
now
we're
into
scikit
mean
absolute
error,
check
consistent
length.
So,
let's
check
out
the
score
class
itself.
A
A
Let's
avoid
doing
the
import
lib
import
module
stuff.
So
let's
avoid
doing
that,
we
had
started
to
do
that
at
one
point
and
we
realized
it
wasn't
it
wasn't
something
good
to
continue
doing
so,
let's
see
so.
A
Okay
yeah
and
that's
a
holdover
that,
like
that,
still
exists
in
the
psychic
stuff.
So,
okay,
so
self
score
equals
parent
circuit
score.
Okay,
and
is
that
a
function
then
so
yeah?
It's
just
a
function?
Okay,
so
I
would
say
when
you
do
try
this
try
doing.
Self.Score
equals
static
method.
All
one
word
static
method,
so
self.score.
So
sorry,
online
27.
A
Yeah
so
static
method,
and
then
you
know
put
self.parent.socket
score
within
that
yeah
perfect.
So
hopefully
this,
let's
see
what
happens
now
so
just
make
sure
line.
46
doesn't
have
that
equals
yeah
all
right
great.
So,
let's
see
what
happens.
A
Static
method
object:
it's
not
callable,
okay,
so
that
doesn't
work
all
right.
I
would
just
say
funktools.partial,
if
you
just
say
instead
of
static
mesh
method,
say
funktools.partial
and
usually
that.
A
B
B
A
A
A
A
Where
do
we
have
this,
it
could
be
a
df.
I
don't
think.
A
Okay,
it
might
be
in
the
operation
stuff,
no,
because
I
think
we
just
pass
funk
there.
A
I
know
we
do
this
somewhere,
you
know
what
you
can
do
is
you
can
just
create
a
method
that
calls
scorer,
let's
see
so
selfless
score,
so
you
could
say:
let's
see.
A
A
A
Yeah,
I
don't
know,
I
would
just
say:
let's
use
some
kind
of
helper
method
here.
I
know
that
we
do
this
somewhere
else,
so
this
is
silly,
but
let's
just
do
some
kind
of
helper
helper
method.
That
says
like,
if
you
say
self,
you
know
you
could
say
here
wait
a
minute
now
this
is
going
to
run
in
the
same
issue.
So
self.score
is
a.
E
I
also
tried
like
providing
keyword,
arguments.
A
Yeah,
because
why
pred,
I
think,
is
taking
the
first
keyword
argument
right
now,
because
they're
not
using
like
this
strict
keyword,
argument
thing,
so
I
would
say
because
they're
doing
the
the
way
it
works
right
now.
I
believe
it's
doing
keyword
or
positional
for
why
pred,
and
so,
if
you
pass
self
like
self
is
self,
is
actually
why
true
and
then
why
true
is
actually
wipe
red
and
then
why
pred
is
actually
some
keyword
argument.
So,
let's
see
I
mean
what
else
can
we
do
here?
What
this
is
ridiculous?
A
Why
I'm
it's
I'm
blanking
on
this.
Let's
just
put
it
self
score.
What
if
you
do
self
funk
tools,
dot,
partial
so
right
on
the
return
on
line
47
do
funk
tools,
dot,
partial!
A
So
so
you
still
says
controls
yeah,
wait,
wait,
wait,
don't
call
sorry,
don't
call
so
funktools.partial
will
return
a
it
should
return
a
call
to
self.score
that
doesn't
involve
self
and
then
we
can
pass
yeah.
Now
we
can
pass
y
true
and
why
pred,
hopefully,
and
let's
delete
the
self
thefunktools.partialwrap
up
there
on
line
20.
A
A
Partial
okay,
so
if
you
pass
partial-
but
here
we
go,
is
this-
it
look
at
the
method.
It's
super
much
value.
A
How
do
we
unbind
this
stupid
thing?
I
would
just
okay,
so
why
don't
you
just
set
it
as
a
property
of
a
dictionary,
and
you
just
set
it
as
a
property
of
a
dictionary
for
now
and
like
create
a
dictionary,
make
it
the
only
property
in
the
dictionary
and
then
call
it
from
the
dictionary
and
it
shouldn't
try
to
bind
to
the
dictionary
it
shouldn't
try
to
put
itself
in
it.
It
shouldn't
try
to
put
the
dictionary
in
where
self
is.
So,
if
you
say
self,
for
example,
if
you
were
to
say.
A
Let's
see,
let's
see
like
this,
so
yeah
self.score
yeah,
so
so,
but
don't
do
app
property
right.
So
you
can
just
do
this
on
line
28
and
say
self
score
equals
you
know,
or
let's
see,
why
actually
could
we
just
say
self.parent.psychid
score
like
do
we
even
need
to
say
self-score.
E
E
A
A
Okay,
I
don't
know
we're
gonna
have
to
look
more
at
this
because
I
don't
want
to
take
any
more
time.
So,
let's
see
okay
yeah,
let's,
let's.
A
E
A
A
Yeah
in
the
in
the
wherever
the
base
classes
are
getting
created
for
the
circuit,
so
psychic
score,
psychic
score.
Yeah
second
scores.
A
A
A
A
E
Let
me
restate
so
there
are
two
scorers
in
this
ik
one
is
which
actually
uses
the
models.
Score
method
default
score
method
to
give
the
accuracy.
A
So
I
think
so,
okay,
let's
all
right,
I
think
this
might
be
one
of
those
where
we
need
to
expose
what
model
type
it
is.
Is
that
is
that
what
you're
thinking
here
as
well,
to
check
if
it's
cluster
and
then
basically
throw
an
exception?
Damn
it.