►
From YouTube: Weekly Sync 2020-10-13
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.cxpozdrmkkb4
A
All
right,
okay,
we're
doing
windows
for
the
recording
here,
so
dfml.
A
Okay,
all
right:
let's
go
to
the
installation
and
let's
try
this
and
watch
it
fail
miserably.
Okay,.
C
A
A
Integration
and
I
use
your
tabs.
The
tabs
thing
that
you
found
this
group
tab
thing
this
is
this
is
nice,
so
we
can
do
a
little.
A
This,
what's
it
called
the
powershell
ps.
A
Okay,
yeah,
it's
sort
of
got
a
path
there,
all
right
so
we'll
just
say,
like
c
calling
something
all.
A
A
All
right,
okay
and
yeah:
let's
try
this
in
the
command
prompt
too
or
or
I
guess.
Let
me
make
sure
that
this
is
command,
because
you
know
who
knows
some
people
will
probably
do
that
instead
of
powershell.
So
let's
make
a
new
one
of
those
all
right.
Okay,
I
think
we're
not
in
powershell
now
yeah,
okay,
great
so.
A
That
is
exactly
the
command.
I
just
typed.
Let's
see,
let's
try
user,
I'm
not
sure,
maybe
that'll
work.
Who
knows
if
that
will
actually
give
us
any
success
here,
but
we'll
find
out.
A
A
A
A
A
Okay
yeah
see,
I
think
that
that
seems
to
be
the
only
way.
I
can
consistently
get
package
installations
to
work
in
windows
which
is
kind
of
annoying.
So,
let's
see.
A
Let's
have
them
install
it
within
a
virtual
environment.
First.
A
Well,
we
do,
I
think
I
think
we
do,
but
not
in
the
main
installation
instructions.
Oh
yeah.
We
have
it
in
contributing
yeah,
so,
let's
see-
and
we
probably
want
to
make
sure
that
it's
right
on
windows.
So
let's
just
try
this
guy,
because
this
should
work
to
create
a
virtual
environment.
C
A
Oh
really,
okay,
let's
just
make
sure
that
that
yes,
okay,
goes
off
without
a
hitch,
all
right,
okay,
just
one
more
time
for
my
own
sanity
or
lack
thereof;
okay,
great
okay!
So
do.
A
Okay,
yeah.
Well,
I
think
the
thing
is
that
you
know
it
depends
what
packages
you're
using
some
things
unless
you
create
the
verge
the
virtual
enviro
with
that
dash,
dash
use
or
site
system
system
site
packages
or
something
certain
packages
that
can
only
be
installed
within
this.
You
know
like
that,
can
only
be
installed
by
apt-get
and
things
won't
be
accessible
within
the
virtual
environment,
so
so
that
creates
some
problems
with
like,
mostly
if
you
use.
A
A
I've
never
I
mean
I
don't
do
much
on
windows,
but
this
is
the
only
way
I've
been
able
to
get
it
to
work
on
windows
is
by
using
a
virtual
environment.
So
yeah.
I
guess
I'm
not
sure
if
that
really
answers
the
question
right,
but
I
think
that
so
far
you
and
me
both
can
only
get
it
to
work
with
the
virtual
environment.
So
I
think
it's
it's
two
against
one
right
now.
A
B
B
A
Agree:
yeah,
let's
see
when
does
make
sure
you
may
want
to
first
create
a
virtual
environment
to.
A
A
All
right,
we'll
flip
back
over
here
for
the
time
being.
Okay,
these
guys,
oh,
I
think
there
may
have
been
an
issue
with
single
quotes
and
windows.
Somebody
told
me
if
I
remember
correctly,
so
we
should
probably
update
these
guys.
A
Yeah,
I'm
not
sure
if
windows
does
windows
do
black
backslashes.
Do
you
guys
know,
I
think
we
can
find
out
python
pip,
install
nope,
okay
windows
does
not
do
backslashes.
A
Let's
find
out
we're
to
learn
a
lot
about
windows
within
the
next
few
minutes
here,
all
right,
so
windows,
equivalent
of
backslash
of
t
line
with.
D
A
A
Okay
and
then,
oh
yes,
we'll
just
move
this
up
here
and
do
it
with
a
carrot.
A
A
A
A
A
Now,
oh
for
now,
yeah.
The
one
thing,
though,
is
now
we
really
gotta
get
okay,
so
now
windows
support.
If
I'm
putting
windows
in
the
main
installation
dock
here,
then
we
should.
We
should
really
make
sure
that
windows
support
works
before
we
do
this
release
all
right.
A
A
A
I
had
squashed
bugs
a
few
while
ago,
but
there
are
now
more
bugs
than
there
used
to
be
issues
with
dfml
service
dev
install
command.
A
A
Whereas
they
do
for
linux-
and
we
should
also
note
that
this
will
probably
be
also
an
issue-
this
may
also
be
an
issue
with
os
x
or
mac
os,
whatever
the
hell
they're
calling
it
nowadays
so
yeah.
We
should
probably
make
a
note
in
the
documentation.
A
This
is
another
good
thing
for
the
are
you
freaking
kidding
me
could
not
install
a
package
due
to
an
environment
area.
The
tag
present
in
the
reparse
buffer
is
invalid.
I
put
this
through
this
stupid.
One
drive
crap,
they
made
it.
They
say
packages
god.
I
hate
this
stuff.
A
They
made
everybody's
stuff,
be
on
one
drive,
and
I
don't
I
don't.
I
just
did
it
python
python
python,
python,
python.
A
A
Okay,
oops:
it's
not
windows,
all
right!
Okay,
I'm
sorry!
Let's
just
go
over
here
and,
let's
say:
docs
installation.
A
A
B
Only
support
like
when,
when
we
run
dxml
service
install
so
did
you
just
add
that
skip
flag
for
that
particular
thing
or
like
are
we
checking
when
it
fails
and
then
skip
set
all
together.
A
B
A
You
have
to
manually
add
the
skip
flag,
mainly
because
you
we
don't
want,
you
know
yeah,
so
you
have
to
manually,
add
the
skip
flag.
So
it'll
tell
you
which
ones
failed
and
then
you
have
to
you
have
to
say
which
ones
you
want
to
skip
actually.
D
B
For
both
of
the
systems,
the
python
version
was
pi,
one
one
had
three
point:
twenty
eight
point,
three,
the
other
had
three
point:
eight
one
five
and
both
of
them
had
failed,
the
install
and
in
the
8.3
version
I
had
to
skip
like
five
packages
and
eight
point
five
version.
I
had
to
skip
two
packages
only
weird,
it's
very
inconsistent
in
windows
and
what's
happening
like
the
pip
version,
was
the
same.
The
python
was.
A
That's
see,
that's
really,
weird!
That's
really!
Weird!
Okay,
let's
make
a
note
of
that.
Okay.
So
when
running
dffml
service,
dev
install.
A
With
different
3.8.x
versions
with
same
pip
version,
sometimes
different
sets
of
plugins
fail
to
install,
and
this
is
where
I
think
what
we
really
need
to
do
is
we
need
to
sort
of
we
need
to.
We
probably.
A
A
And
then
take
note
of
which
plugins
fail
to
install
within
the
ci
and
use
that,
as
our
you
know,
ground
truth
list
of
what
works
and
what
doesn't,
because
if
it
works
in
the
ci
we
can
basically
say
hey.
You
know
it
does
work.
You
know
you
gotta,
you
gotta
mess
with
your
system
until
it
works.
So
that's
that's
probably
what
what
what
we
should
do
here
to
get
this
to
get
to
get
our
list,
because
I
can
show
you
basically
lists
or
exclusions.
B
A
No,
you
usually
wouldn't
have
a
separate
branch
for
every
os.
I
mean
right
now
we
may
have
a
separate
branch
like
we.
If
there
was
like
a
ton,
a
ton
of
work
to
do
where,
like
the
master
branch,
doesn't
yet
support
that
os,
then
you
may
create
a
separate
branch.
Do
that
ton
of
work
and
then
merge
it
into
master.
A
Somehow
so,
usually
you
just
try
to
you
know,
beat
the
code
base
into
the
point
where
it
works
for
both
right
so
but
for
for
with
the
with
the
ci
jobs
right,
we'll
yeah,
we'll
probably
just
we'll,
probably
enable
them
in
the
matrix,
somehow
and
and
then
just
go
from
there,
I'm
not
exactly
sure
what
not
yeah
we'll
we'll
figure
it
out.
A
But
basically,
you
know
like
this
sort
of
well,
okay,
so
we'll
turn
it
on
right,
we'll
turn
it
on
for
all
of
them
right,
for
example,
let
me
go
in
here.
B
I,
what
I'll
do
is
actually
just
install
it
in
my
system
again
and
both
of
the
systems
and
I'll
just
post
it
in
an
issue.
Okay,.
B
A
Because
what
we
want
to
do
is
we
want
to
look
in,
we
want
to
basically
go
and
we
want
to
figure
out
what
installs
and
what
doesn't
install
right
and
then
in
here
this
is
the
list
of
plugins
right
and
right
now
we
know
that
dolph
reply
down
for
pi
doesn't
support
3.8,
so
I
we
don't
add
it
unless
it
it
is.
You
know,
unless
the
the
python
version
is
3.7.
A
Yeah,
because
this
stuff
stuff
that
gets
typed
in
here
gets
lost.
Usually
unless
I
go
back
through
and
you
know
we
go
back
through
and
we
look
at
it
right,
but
we
don't
often
go
back
through
all
the
meeting
minutes.
You
know
you
gotta,
remember
that
you
need
to
go
check
the
meeting
minutes
for
something
first.
I
did
recently
go
through
it
because
I
saw
I
I
moved.
We
changed.
A
We
talked
about
changing
the
headers
for
the
tutorials
or
the
examples
and
usage
and
tutorials,
and
so
I
did
go
through
and
I
changed
those
and
I
moved
the
should
I
tutorial
for
the
operations
under
under
the
the
examples,
because
you
know
it's
sort
of
way
longer
and
the
rest
of
the
stuff
is
under
the
examples
for
that
so
and
then
the
tutorials
is
is
now
it's
a
little.
I
think
it's
a
little
more.
A
It
makes
sense
what
the
hell
is
going
on
here,
but
I
think
there
was
still
a
little
bit
more
refactoring
work
to
do
on
that
documentation
itself,
but
that.
A
A
Core
plugins
for
windows
within
dffml,
plugins.py
and
I'll
link
to
that.
B
Yeah
I'll
I'll
actually
try
it
myself
again
and
I'll
just
open
an
issue.
Okay,
let
me
start
with
yeah:
whatever
tests
are
failing
I'll
discuss
it
with
you
again,
okay,.
A
All
right,
so
that
sounds
good
okay
yeah,
because
that
would
be
cool.
Let's
see
anything
else,
you
wanted
to
talk
about
yash.
A
A
About
okay,
cool,
so
sudhanshu,
okay,
so
we
wanted
to
talk
about
the
removal
of
the
accuracy
method
from
model
context,
and-
and
you
know,
maybe
we
will
do
that
in
the
future.
Maybe
we
do
it
now
and
then
is
there
anything
else
you
wanted
to
talk
about.
A
Okay
and
then
himanshu
will
move
that
for
whenever
he
comes
next.
E
Meeting
so
so,
actually,
like,
I
have
fixed
you,
the
two
models,
the
spacey
model
and
the
tensorflow.
A
Okay,
you
fix,
let's
see,
let
me
write,
let
me
get
the
notes
going
here.
E
B
E
A
Okay,
yeah,
and
so
this
is
sort
of
what
I
was.
What
I
was
thinking
is
based
off
that
so
let's
open
that
stuff
up
here
so
wrap
the
score
evaluation
methods
within
accuracy,
plugins,
so
yeah,
okay,
so-
and
I
think
I
think
what
we're
seeing
here
and
what
we're
seeing
and
let
me
bring
up
the
code
for
that-
I
wanted
to
talk
about
here
so.
A
I
think
we
have
a
couple
cases
now
where
you
know
so
we
implemented
the
accuracy
method
right.
We
got
rid
of
the
accuracy
method.
We
basically
said:
okay,
it's
implemented
in
the
in
the
base
class
right
and
then
we
okay
here
we
go
yeah.
We
said
it's
implemented
in
the
base
class.
Really
it
didn't
take
me
to
the
file
and
so
now
we're
relying
on
the
base
class
to
to
to
call
the
score
right.
A
Yes,
but
the
thing
is
right
with
this:
one
that
you
just
implemented
or
the
one
that
got
modified
here,
where's
this
guy
right,
so
you
needed
it
to
take
the
model
context
right.
So
now
the
scorer
takes
the
model
context,
but
first
the
model
context
took
the
scorer
basically
and
started
to
think
you
know.
A
I
think
actually,
the
reason
why
that
was
was
because
we
needed
the
features
right
and
so,
but
the
thing
is
right:
if
we,
if
the
score
takes
the
model
context,
this
score
can
grab
the
features
from
the
model
context
right,
because
features
should
always
be
be
features
or
we
can,
you
know,
implement
some
kind
of
method
and
and
within
the
model
context.
This
is
give
me
the
features
that
you
care
about
right,
so
I'm
kind
of
leaning
towards
you
know.
A
Maybe
maybe
we-
and
this
is
where
you
know
it's
good-
to
have
that
high
level
accuracy
method,
but
maybe
we
need
to
a
you
know.
A
Maybe
it's
not
within
phase
four
here,
maybe
it's
afterwards
after
we've,
we've
moved
everything
around,
but
maybe
we
need
to
make
it
so
that
this
is
actually
the
signature
of
the
the
score
method
right
of
all
of
the
score
methods
and
because
you
know
then
that
in
that
case
now
the
score
has
accuracy
access
to
the
model
context,
and
it
can
just
call
you
know
it
can
just
call
predict
right
and
or
it
could
it
could
reach
in
and
it
could
say.
A
Oh,
like
you
are
a
spacey
model,
you
know
I
know
you're
going
to
have
these
certain
things
right.
Yes,
I
think
I
think
it
seems
like
I
mean
correct
me
if
I'm
wrong,
but
that's
just
sort
of
what
I'm
observing
from
the
code
that
you've
the
the
patterns
that
I'm
seeing
in
the
code
that
you've
been
writing.
E
Actually
like
for
this,
for
this
thing
right
here
so
like
the
spacey
has
its
own
scorer
method,
so
in
the
model.
B
E
In
the
moral
context,
an
object
of
that
is
created
actually,
so
that's
why,
yes,
so
that's
why
I
thought
like.
Maybe
I
can
just
call
that
function
in
the
moral
context
to
actually.
A
A
Yeah
yeah
and
I
think,
and
that's
what
I'm
thinking
you
know
is.
I
think
that
if
we
go
through
and
you've
been
changing
things,
you
know
like
you've,
been
changing
some
test
cases
like
this
and
where
we've
done
that
it
may
make
sense
to
go
through
and
actually
just
look
at
wherever
you've
changed
test
cases
like
this
and
have
them
call.
A
You
know
the
high
level
version
of
accuracy
and
then
that
way,
so,
if
you
go
through
and
you
look
at
all
the
places,
you
did
this
type
of
thing
where
you
added
the
score,
this
is
really
just
you
know.
This
code
is
sort
of.
You
know
it's
it's
old
right
from
when
we
didn't
have
the
high
level
accuracy
function
right.
A
Could
change
it
to
the
high
level
and
then
we
make
the
high
level
do
the
switch
on
on
how
it
actually
gets
called?
Does
that
sound
like
a
good
plan.
A
But-
and
maybe
you
know
and
that's
like
you
know,
if
things
are
working
right
now,
we
maybe
make
this
phase
five
or
something
it
doesn't
have
to
be
right
now,
but
I
think
we
know
we've
learned
and
this.
E
A
I'm
sure,
as
you
go
through
this
big
refactor
right,
this
is
a
serious
job
right.
So
so
yeah
I
mean
yeah,
you
you're,
because
this.
A
In
right
and
there's
just
so
many
different
variables
going
on
that
that
you,
you
learn
as
you
go.
What
what
really
the
way
the
code
needs
to
flow
best
right.
A
Yeah
we've
got
so
many
models
right
now
that
it's
like
the
good
thing
is
see
this
and
there's
there's
an
upside
in
a
downside,
right,
the
upside
or
so
so
the
downside
is,
you
know
it
you
end
up
with
now.
You
have
to
go
back
and
do
more
stuff
right.
Well,
the
upside
is
that
if
we
didn't
have
all
these
models,
we'd
be
finding
this
out.
You
know,
maybe
six
months
from
now
when
we
did
this
and
now
we've
got
all
these
models.
A
More
code
that
relies
on
it
right
so
yeah.
I
think
this
is
really
cool.
I
was
talking
to
somebody
about
this
accuracy
scoring
stuff
and
how
this
will
be
sort
of
a
demarcator
of
the
beta
release.
I
think-
and
so
I'm
really
excited
to
see
this-
let's
see
cool.
Is
there
anything
else?
Let
me
make
sure
that
we
write
this
down
here
and
then
is
there
anything
else
you
wanted
to
talk
about
related.
To
this
I
mean,
I
think.
E
Yes,
so
like
I
will
have
to
like,
create
an
accuracy
method
in
the
high
level
and
make
the
call
to
the
high
level
right.
A
A
A
C
A
A
Okay,
so
let's
see
yeah
and
then
yeah,
so
this
makes
good
sense
is
like
a
phase
five
before
we
get
to
things
like
the
http
service,
and
so
how
are?
Let's
see
I
haven't
looked
at
the
actual
output
of
the
test
recently
did
you
is
it?
Are
things
starting
to
pass
again
like?
Did
they
go
from
failing
to
now
passing
again
or
are
we
seeing
more
more
issues
in
the
in
the
the
ci
here.
E
So
in
the
ci,
like
the
tests
are
passing
actually,
oh.
E
A
Okay,
great
so,
let's
see
and
then
model
scikit.
I
feel
like
we
talked
about
this.
Did
we
talk
about
what.
E
A
I
think
this,
you
remember
the
thing
that
you
said
where
you
said:
if
you
change
the
feature
name,
I
think
that's
related
to
the
same
issue
that
this
guy
is
seeing.
A
A
Basically,
the
issue
that
he's
trying
to
fix
is
that
the
the
classification
or
sorry
the
the
psychic
models
when
they
do
a
prediction
they
sort
of
just
leave
it
as
like
a
floater
and
into
whatever
comes
out
of
scikit
and
but
the
you
know
that
we've
got
the
prediction
feature
and
it
tells
us
what
kind
of
fee
you
know
what
kind
of
a
data
type
it
should
be
so
sometimes
we're
trying
to
predict
a
float
but
scikit's
giving
us
in
an
into,
or
vice
versa.
A
A
He
tried
to
he's
in
progress
trying
to
do
that,
but
what
he
found
is
that
basically
he's
ending
up
with
a
problem,
and
I
believe
it
is
with
the
clustering
ones
too,
where,
where
it
ends
up
comparing
a
string,
it's
looking
for
a
string
within
this
list
of
with
within
this
list
of
ins,
so
that
tells
us
something
about
the
data
types
is
wrong
and
and
when
I
went
in
there
and
I
looked
at
it
and
I
looked
at
the
test
gate
data-
I
also
noticed
the
thing
what
you
were
talking
about,
where,
if
you
flip
the
feature
name
and
stuff
like
something
something
changes
when
you
flip
those
feature
names-
and
I
can't
exactly-
I
haven't-
figured
it
out
yet
either
but
yeah
that
is
there's.
A
I
think
it
may
be
something
with
the
way
that
the
test
case
is
written.
I'm
sure
the
test
case
is
is
right
right
now.
I
think
I
think
well,
okay,
I'm
not
sure
that
the
test
case
is
right.
Right
now,
I
think
is
is
where
this
boils
down
to,
and
it
would
be
good
if
we
could
talk
to
him
honshu
about
this,
because
he
wrote
the
test
case,
and
you
know
we
could
all
spend
more
time
on
it
or
we
could
we
could.
We
could
see
if
he
could
answer
our
question
right.
E
A
So,
let's
see,
let's
see
change
so
that's
yeah
and
we'll
try
to.
I
tried
to
ping
him
on
an
issue
that
I
opened
today
and
I
don't
know
if
he's
got,
he
hasn't
gone
back.
So
let's
make
a
note
in
here.
A
A
C
A
Okay,
sorry,
let
me
just
sort
of
get
organized
here,
okay,
so,
and
then
we
also
talked
about
so
there
is
an
issue
with
scikit
or
yes,
scikit
clustering
models.
A
It
might
be
related
to
the
issue
that
is
being
faced
in
and
the
other
thing
is:
if
himanshu
comes
back
and
tells
us
what's
going
on
that's
great,
if
not,
who
knows,
maybe
this
guy
will
fix
we'll
figure
it
out
when
he
fixes
this
pull
request
so
we'll
see,
and
if
no
one
figures
it
out,
then
we'll
we'll
we'll
just
take
a
look
at
it
together
in
the
next
meeting.
A
E
A
So,
let's
see
we
will
see
if
pr
8
64
fixes
the
issue
as
well
as
talk
to
himanshu,
to
see
if
he
can
help
us
figure
out.
What's
going
on
there.
A
Okay-
and
let
me
just
add
a
little
add
a
little
thing
here,
just
to
remind
me
to
message
him:
okay
and
then,
let's
update
yeah,
let's
update
phase
five
so
go
okay,
so
change,
remove
accuracy,
method
from
model
and
change.
Okay,.
A
Score
and
then
self
so
the
other
thing
was
that
you
did
model
context
which
is
good,
that's
probably
more
descriptive,
but
sort
of
everywhere
else.
We've
been
doing
this
shorthand
with
the
sort
of
first
letter
of
whatever
it
is
and
then
ctx,
so
I
I
think
it's
probably
probably
best
that
we
just
keep
that
unless
that's
not
clear
is
that
I
mean
I
know
this.
This
may
not
be
clear.
This
is
a
good
thing.
They
wear
it's
clear.
I
thought
it
was
clear.
It
may
not
be
clear.
A
A
A
A
Yes,
all
right
well,
we'll
assume
that
josh
thinks
the
same
thing.
Just
it's
for
all
three
of
us
here.
A
B
A
Yeah
no
worries
so
basically
we're
talking
about
this.
This
method
signature
here
and
so
so
we
can
type
out.
We
can,
we
can
say,
model
context
or
model
ctx
or
we
can
say
mctx.
B
D
D
A
A
A
Well,
so
this
is
the
score
method
right,
so
so
we
need
to
change
all
the
score,
func
or
score
methods
within
all
the
accuracy
scores,
so
that
their
signature
is
like
this
right,
because
this
is
the
score
method
within
the
accuracy
score
yeah,
and
then
we
need
to
remove
the
accuracy
method
from
all
of
the
model
or
contexts.
A
Okay,
okay,
so
there
should
only
be
the
one
in
the
base
model.
A
Context,
I
just
thought
I'd
seen
one
somewhere
else,
so
that's
why
it
said
that,
but
I'm
not
I'm
not
sure
on
that.
So,
let's
see
I
just
want
to
make
sure
I
I
wanted
just
wanted
to
wear
it
like
that.
Just
in
case.
A
A
A
A
Base
funnel
context-
and
there
should
be
another-
remove
in
text
classifier,
okay.
So
now
we
need
to
modify
all
of
the
tests
that
used
that
don't
use
the
high
level.
A
Accuracy
function
to
now
use
that
function
so
anywhere
we
were
doing
and
then
let
me
just
sort
of
link
to
it.
The
double
context,
entry
type
of
thing
we
want
to
replace
that
with
the
usage
of
the
new
high
level
api.
A
Okay,
so
so
anywhere
we
were
doing
the
double
context:
entry
from
pre,
high
level
accuracy
function
days
or
from
when
we
didn't
have
the
high
level
function,
for
example,
here.
A
Functions
so
that
they
call
the
score
method,
appropriate
lee,
okay
and
then
so
I
think
I
think
this
should
get
us
to
where
we
need
to
be.
Do
we
have
it
see
any
problems
with
this
plan.
A
Yeah,
maybe
right
yeah,
okay,
maybe
not
at
the
moment,
but
you
know
we
can
always
go
over
it
later.
So
let
me
just
link
to
this,
and-
and
we
can
you
know
we
can
always-
we
can
always
realize
that
we
need
to
change
things
later
so
so
yeah.
That's
why
I
wanted
to
make
sure
we
write
it
out
as
detailed
as
possible,
so
all
right,
okay.
So
let
me
just
make
sure
we
got
everything
here.
A
Yes,
okay,
do
you
see
any
issues
with
that
at
all
or.
A
E
A
And
I
think
that's
where
that's
also
where
this
is
is
is
is
powerful
is
the
fact
that
we
can
now
what
we
can
do
is
we
can
say,
okay,
if
the
accuracy
score
receives
a
model
context
that
you
know,
doesn't
have
these
certain
properties
right
or
does
not
derive
from
spacey
model
context,
or
something
like
that.
You
know,
then
we
can
just
say
hey.
I
don't
work
on
this
right
instead
of
the
model.
Half
the
same,
I
don't
work
with
this
score.
A
I
think
that
makes
more
sense.
So
I
think
I
think
you've
found
a
good
good,
good
good
way
of
doing
this
so
yeah.
Anything
else.
Then.
E
A
E
A
Yes,
me
too
yeah
there's
there's
there's
there's
not
that
there's
not
as
many
of
us
yeah
yep
it's
I
like
the
computer
engineering.
Do
you
have
any
sort
of
at-home
circuit
projects,
you're
doing
or
anything.
E
A
A
Yeah
yeah
well
yeah
that
could
be
cool
because
I
had
a
friend
of
mine
he's
also
doing
computer
engineering.
He
recently
ordered
this
sort
of
like
robot
car
that
he
was
trying
to
play
around
with,
and
you
know
it
didn't
it
didn't
go
as
planned.
Basically,
the
thing
will
only
turn
left.
He
can't
even
get
it
to
go
straight,
so
it
just
turns
left
in
circles
forever,
but
you
know
yeah
electronics.
As
soon
as
you
get
down
to
stuff,
that's
not
software!
It
just
is
weird.
It's
just
it's
all.
A
Right
yeah,
it's
fun,
though
all
right
well
have
a
good
one.
Let
me
know
if
you
need
anything
and
I'll
talk
to
you
next
week,
then.