►
From YouTube: Weekly Sync 2020-11-10
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit?pli=1#heading=h.3z3847d1k5dw
A
Hello
all
right,
perfect!
Thank
you.
I
don't
know.
That's
weird.
I
don't
know
why
it
wasn't
things
weren't
jumping
earlier
anyways,
the
just
the
the
audio
knobs
weren't
working.
B
A
Okay,
see
here,
okay,
so
I
so
I
let's
see,
I
think
I've
got
something:
that's
just
ready
to
be
merged
here,
yeah
right,
so
I
think
I
got
pretty
much
all
the
test
cases
tested
or
all
of
everything
tested
at
this
point.
So
that's
the
that's
the
good
news,
so
I'm
gonna
do
it
right
now,
so
you
guys
can
all
get
the
results
of
that
review.
You're,
not
seeing
what
I'm
doing,
because
I'm
on
a
different
computer
but
it'll
come
up
in
a
second
and
test.
Docs
fix
it.
A
A
Sweet
so,
let's
see
what
did
I
just
do
here?
Take
a
look
back
at
it.
Let's
see,
okay,
all
right,
so
we
talked
about
this.
I
can't
remember
if
this
was
yeah
shoe
and
we
talked
about
this
or
if
all
of
us
talked
about
this,
but
we
moved.
Okay,
let
me
and
let
me
bring
open
some
notes
here.
A
A
A
A
A
A
Actually
we'll
go
through
and
we'll
grab
everything,
and
then
I'll
start
telling
you
guys
about
this,
but
hopefully
that's
hopefully
now
that
I
put
it
on
the
agenda.
It
didn't
actually
change
and
crash
this
time.
This
passed
on
a
different
branch,
but
it's
always
a
mystery
as
to
whether
it's
going
to
run
or
not
all
right.
Yes,
I
saw
you'd
been
making
some
new
issues
related
to
windows
stuff.
How
is
what
what
do
you
want
to
talk
about
today?.
C
D
D
A
A
A
D
Cool
scikit
is
working,
fine,
orte
sqlan
isn't
supported
on
windows
and
I
reviewed
the
pr
that
was
open
for
autism
too.
A
Okay,
sweet.
We
got
a
pr.
A
Okay-
and
I
think
I
remember
seeing
that
one
so-
let's
see
let's
jump
over
there.
A
A
A
Yeah,
thank
you,
that's
very
helpful.
So,
let's
see,
I
don't
think
I
think
at
the
moment
where
really
ogen
got
sidetracked
here,
so
we
need
to
he
needs
to
he's
working
on
this,
and
this
is
sort
of
turned
into
something
that
I'm
doing
as
well.
This
guy,
I
think,
and
this
guy
he
just
needs
to
run
black.
I
think
or
oh
no
hebrew.
Oh
that's
right.
There's
an
issue
with
the
psychic
models
here.
A
A
A
C
C
A
D
A
D
A
A
D
That's
it
like,
I,
I'm
just
running
the
test
and
I'll.
Let
you
know.
A
Yeah,
just
just
let
us
know,
and
then
why
don't
you
I
don't
know,
I
can't
remember
if
we
have
an
issue
to
track
this
as
a
whole
or
if
we're
just
doing
issues
so
just
issues
for
each
thing
that
doesn't
work
but
yeah,
just
you
know,
just
put
it
in
some
kind
of
issue
somewhere
and
we'll
we'll
track
it.
That
way.
If
we
find
one
that
doesn't
work.
Okay,.
A
I've
noticed
that
yeah
there's
something
wrong
there.
Okay,
so
and
that's
and
that's
a
bug
we
have
like
overall,
so
that's
something
that
we
need
to
track
as
a
whole.
So,
let's
see
bugs
found,
so
the
transformers
keeps
re-downloading.
A
When
tess
are
running
right
yeah,
this
is
this
is
some
kind
of
problem.
We
need
an
issue
to
track
this,
I'm
not
exactly
sure
what's
going
on
there,
but
I
noticed
that
in
the
ci.
I
think
it's
been
like
that
for
a
while.
So
it's
not
just
a
windows
thing
yeah,
it's
definitely
slowing
down
our
ci,
and
sometimes
it
fails
because
we
get
you
know
random
network
errors
that
fail
right
there
and
then
it
fails.
So
that's
annoying.
So
we'll
want
to
try
to
figure
out
what
the
hell's
going
on
there.
D
A
D
D
A
Okay,
that's
really
interesting
yeah
some
of
the
tests
aren't
working
locally
but
are
working
in
the
ci.
Okay,
let's
see
actually
this
there's
another
bullet
point
for
this,
so
that's
really
interesting
yeah.
We
may
want
to
count
when
we
want
to
do
the
unit
test
skip
on
that.
I
think
it's
probably
the
right
way
to
handle
that
for
now
right
and
note
that
this
is
a
windows
issue,
I
mean
we
can
go
fix
it
too.
A
That
would
work
fine
right,
but
I'm
not
sure
I
can't
I'm
I'm
not
sure
it's
interesting
to
me
that
it
works
in
the
ci
environment
and
not
locally
because
yeah,
you
know
what
the
hell's
up
with
that.
So,
let's
see,
let's
see
if
we
can
find
out
what
happened
here.
A
C
All
right,
we're
looking
for
tests
test
functions
all
right.
B
A
B
A
A
A
D
A
A
Oh
yeah,
that's
right,
yeah,
so
yeah.
It
could
very
well
be
that
because
yeah
windows
doesn't
like
it
when
two
files
have
two
files
have
a
temporary
file
or
a
file
open
at
the
center.
Two
processes
have
the
same
file
open.
It
doesn't
usually
like
that.
A
That's
interesting,
okay
yeah
we
might
want
to
just
mark.
We
probably
want
to
mark
this
one
with
the
unit
test
skip
the
fact
that
it's
working
in
the
ci
is
weird,
but
maybe
that's.
A
D
A
And
that
should
I
mean
that's,
usually
good,
for
what
is
it
usually
good
for
that's
good
for
when
we
call
the
subprocess
and
then
we
want,
because
I
think
my
understanding
of
is
in
those
situations.
A
We
can
actually
close
the
file
before
we
run
the
sub
process,
but
in
this
case
to
run
the
sub
process
we
need
the
file
open
is
my
understanding
of
of
how
this
works,
and-
and
I
mean
you
I
would
say
you
know
it
might
be-
it's
probably
worth
it's
worth
trying,
but
I'm
pretty
sure
that
it
won't.
It
won't
work
like
you
won't
even
be
able
to
get
to
pass.
It
won't
run
the
commands,
I'm
pretty
sure,
like.
A
Yeah,
the
question
is
really
how
the
hell
is
it
working
on
ci
and
I
think
the
the
answer
lies
somewhere
within
it's.
This
windows
container
that
they're
using
some
this
some
kind
of
windows
container
that
they're
using
that
that
doesn't
have
this
limitation,
which
is
interesting.
So,
let's
see,
let's
see
python.
A
C
C
A
A
A
This
is
yeah,
I
don't
know
okay,
I
guess
this
will
take
some
digging.
Weird
though,
what's
what's
the
exact,
can
you
post
the
exact
trace
back
somewhere,
yeah
and
actually.
B
C
You
can
see
that,
yes,
all
right.
B
A
Okay,
I
think
we
have
different
we're
on
a
different
commit
here.
Now,
let's
see
I
can
do
a,
can
you
do
git
status,
real,
quick.
A
A
C
A
A
D
A
A
Okay,
let's
see
yeah,
I
wonder
if
there's
something
else
we
can
use.
That's
like.
A
Yeah
I
haven't,
I
haven't
seen
it
either.
It's
a
good
question.
A
Oh
hell,
yes
great,
to
see
I
passed
on
the
latest,
run.
E
A
Yeah,
I
haven't,
I
mean
I'm
not
I'm
not
aware
of
which,
what
version
of
anything
it
is
either.
It
would
be.
Let's
see,
let's,
let's
find
out
just
so.
We
know
github
actions,
windows
latest
shell,
environment.
A
Tools,
kit
doesn't
list
scrap,
but
it
must
have
it
because
it's
yeah
this.
A
A
A
A
Okay,
yeah
so,
okay!
Well,
that's
that's!
Let's
just
throw
that
link
that
you
found
in
that
in
our
meeting
minutes
just
so
we
keep
that.
A
Excuse
me.
Okay,.
A
There
we
go
all
right,
sweet
all
right
anything
else
here
or
you're,
just
you're
still
just
gonna
you're
gonna
go
through
the
rest
of
these
and
and
sort
of
find
find
more
bugs
and
stuff
right.
Yeah.
A
E
I
haven't
been
doing
much.
I
just
dropped
to
see
what
everyone's
been
up
to
cool.
B
A
A
Oh
there's,
a
battery.
I
needed
that
battery.
E
Remember
so
maybe
I'll
take
a
look
at
pie,
torches,
new
versions
and
see
what
else
is
failing?
E
Then
yeah
I
have
to
I
have
planned
on
adding
the
config
class
for
the
other,
the
optimizer
from
oh.
A
A
A
A
A
A
Okay
and
then
I
wanted
to
duplicate
this
stuff
down
here.
A
Sweet
and
I
told
somebody
about
your
colorization
demo
and
they
were
very
excited
to
see
that
so
it's
it's
being
eagerly
awaited
by
more
than
just
me.
Okay,.
A
A
Yeah,
it's
nice
that
it's
nice
that
we're
we're,
starting
that
with
your
with
your
image
stuff.
That's
always
it's
nice
that
we're
starting
to
get
that
because
you're
doing
those
things.
Because
that's
you
know,
people
love
people,
love
images
right!
It's
like
it's!
They
can
really
yeah.
It
really
makes
the
connection
to
what's
going
on.
Okay,
so.
E
Maybe
I
was
thinking
about
playing
with
some
data
sets
regarding
the
recent
elections.
Oh.
A
All
right,
oh
yeah,
thank
god
that
went
well
well.
You
still
have
to
over
I'm
holding
my
breath
on
on
scene.
How
it
goes.
I
mean
the
guy's
got
two
months
left
to
completely
screw
us
all.
So
it's
we'll
see
right
all
right,
let's
see
see
okay,
so
I
wanted
to
just
it's
anybody
either.
You
guys
have
anything
else.
You
wanted
to
talk
about,
or
else
I
just
wanted
to
update
you
on
what
I'd
been
working
on.
D
Well,
I
have
just
one
question
yeah
for
the
spacey
models
that
once
you
wrote
like,
are
we
running
the
script
for
testing
that
you
mean
in
the
ci
or
like?
Are
there
multiple
tests
that
are
dependent
on
shell
scripts.
A
A
D
I'll
just
see
where
the
file
is
spacey,
it's
running
examples
ner
and
train
data,
so
I
guess
we'll
have
to
skip.
We
don't
have
this
for
windows,
because
we
are
using
this
in
the
documentation
right.
A
Okay,
yeah-
and
this
is
okay,
so
this
is
another
thing:
okay,
yeah
and
then
let
me
just
let
me
just
go
grab
for
this
because
now
I'm
curious.
A
A
Ar
integration,
yeah,
yeah,
okay,
okay,
yeah,
so
that
that
makes
sense,
and
so
okay,
so
here's
the
thing
is
that
this
type
of
stuff?
This
is
the
next
that
sort
of
feeds
in
okay.
Let
me
just
write
this
down
and
then
we'll
feed
into
what
I
was
going
to
talk
about
here.
So.
A
Okay,
spacey
spacey
scappy,
there's
some
other
one,
there's
like
a
bunch
of
these
that
are
like
very
variations
on
those
five
letters:
a
bunch
of
python
packages,
okay,
so
bugs
found
all
right.
Okay,
so,
let's
face
it.
A
A
All
right,
okay,
so
here's
here's
the
thing.
So
let
me
go
and
so
finish,
testing.
Okay,
so
I
finished
testing
the
tutorials
using
the
new
console
test.
Sphynx
extension
that
I
wrote
in
docs,
ext
console
test
dot
py.
A
Now
this
thing
is
a
bit
of
a
convoluted
mess
right
now,
a
bit
of
a
mess
right
now,
but
oops
wrong
wrong
kind
of,
but
let's
see,
but
I
have
an
issue
to
track
that
I
want
to
get
done
before
we
finish.
The
0.4
release
is
tracking
cleanup
and
documentation
of
it.
A
That
when
you
guys
write
when
it
when
we
all
write
more
stuff
it,
it's
we're
all
on
the
same
page
with
how
we
can
use
the
the
console
test,
plugin
that
we
now
have
to
test
the
documentation
and,
let's
see,
let
me
think
of,
I
can
pull
out
some
of
these
guys.
A
Some
of
these
commits
so
test
with
console
test,
and
this
one
may
not
be
very
elegant
here.
So,
let's
see
yeah
this
one's
a
bit
of
a
mess
transition
wise
okay.
So,
let's
see,
I
think
there
was
a
better
one
somewhere
something
test
with
console
test.
Okay,
so
this
is
in
general.
A
A
You
add
tests-
and
I
know
I
explained
this
last
time,
but
just
sort
of
as
a
refresher
and
then
there's
a
few
other
things
like
you
can
do
replacements
of
stuff
and
there's,
like
oh
it'll,
start
things
as
as
demons,
so
it'll
run
the
http
server
in
the
background
and
assign
it
a
random
port
specifically
for
the
http
server,
and
then
you
can
use
the
random
port
and
be
like
okay,
wait,
replace
this
port
number
here
in
this
command,
with
whatever
random
port
was
chosen
and
sort
of
just
you
know
more
more
more
stuff
like
that.
A
Basically
right
because
we're
just
we're
just
the
whole
point
here
is
anything
that's
displayed
in
the
documentation.
We
want
to
make
sure
that
we're
testing
and
we
want
to
try
to
get
the
test
code
basically
as
close
to
the
documentation
code
as
possible.
So
the
way
we
do
that
is,
we
wrote
the
sphinx
thing
to
keep
it.
You
know
basically
to
to
make
to
make.
Is
that
we
we?
A
A
Up
with
all
these
things
in
random,
you
know
all
these
random
test
files
that
are
now
running
bash
scripts,
which
is
a
mess
so
we're
trying
to
get
away
from
that
right,
so
that
feeds
into
just
as
a
refresher
that
feeds
into
this
issue
with
the
train
data.sh
in
the
ner
examples.
So
the
thing
about
this
is
that
the
reason
we're
running
into
this
is
because
we
were
testing.
A
We
were
testing
the
models,
their
doc
strings
are
also
restructured
text,
and
so
we
were
doing
the
same
thing
here
where
we
were
right
at
first,
we
weren't
testing
them
now
we're
testing
them
and
now,
when
we
test
or
currently
the
way
that
they're
being
tested
is
that
we
had
to
make
those
other
separate
test
scripts
so
that
we
go
and
run
all
of
these
things
that
we
would
see
here,
which
we
had
to
literal
include,
so
that
we
didn't
have
to
parse
it
out
and
then
run
it.
A
And
then
we
went
to
moving
just
run
the
thing,
and
now
we
have
the
plugin
which
can
extract
it
from
the
dock
strings
and
run
then
run
the
tests.
So
that's
the
next
step
here,
so
basically
that
so
I've
done
it
for
the
main
tutorials.
A
A
So
all
the
contributing
stuff
has
not
been
tested
at
this
point
and
that's
partially
because
partially
because
it's
kind
of
specific
and
the
test
framework,
I'm
not
really
sure
if
it's
a
good
fit
for
what's
going
on
in
there
yet,
but
it
will
be
important
to
go
back
through
and
make
sure
that
all
that
stuff
is
correct
right,
but
sort
of
top
priority
was
user
facing
documentation.
We
can
always
roll
a
four
point,
0.4.1
release
to
fix
this
stuff,
and
so
basically
now
what
needs
to
happen?
A
Is
that
just
kind
of
like
with
those
modifications
that
we
saw
there?
Ideally,
what
we
do
is
is
we
go
through
and
we
start
modifying
these
model
docs
strings
like
and
actually
the
auto
sklearn
one
was
the
one
that
I
was
going
to
do,
because
we
don't
have
examples
for
that
and
that
was
sort
of
what
I
was
going
to
start
with
here
and
and
we
make
it,
we
follow
the
same
format
right
so
and
just
as
an
example
here
we
can
go.
A
I
can
go
sort
of
do
this
one
right
now,
but
this
is
basically
what
what
we
end
up
doing
here.
Is
you
go
through
and
anytime?
You
see,
a
literal
include,
you
go
with
bim.
This
becomes
pretty
easy,
oops
read
file
and
all
of
a
sudden,
okay
and
here's
the
other
thing,
so
it
didn't
implement
support
for
this
eof
situation.
So
there's
this.
A
It's
not
you're
not
actually
running
most
of
these
commands
in
bash,
because
we
need
the
ability
to
modify
the
commands
and
do
some
other
things
that
mean
that
reconstructing
the
command
as
a
bash
command
can
be
tricky.
So
we
ended
up
just
running.
We
end
up
just
running
it
with
subprocess,
which
also
means
that
we
should
be
able
to
run
these
things
on
windows.
Eventually,
we
just
need
to
figure
out
like
what
the
various
quirks
are.
A
Quarks
are
probably
related
to
piping
between
commands
in
that
situation,
but
we'll
we'll
need
to
figure
that
out
when
we
get
there,
but
it
also
means
that
we're
not
tied
to
actually
just
bash.
So
when
we
see
something
like
this,
where
we're
creating
files,
we're
actually
just
gonna,
say:
hey,
you
know:
here's
here's,
the
file
data
and
we're
gonna
we're
gonna
sort
of
just
display
it
as
a
file
right
and
say:
hey,
here's,
your
file
and
you
use
whatever
the
hell
editor.
You
want
to
put
it
in
this.
A
You
know
in
this
file
because
I
think
that
was
also
a
thing
that's
happening
is
I
noticed
that
some
people
see
this
and
they
get
very
confused
about
what
the
hell
is
going
on
here.
Right,
they're,
like
I
don't
I
don't
know
what
that
syntax
is,
but
they
do
know
that
when
they
see
a
file
on
a
file,
name
they're
supposed
to
put
it
in
a
file
right.
A
And
yeah,
so
we
basically
go
through
and
we
do
this.
Let
me
just
make
sure
that
was
the
correct
file
name,
yes
and
then
we'll
train
the
model.
A
A
A
But
okay,
is
it?
Is
there
any
any
questions
on
this?
Okay,
cool,
awesome,
yay,
so
yeah?
This
is
basically
you
know
this
is
this
is
the
gist
of
it
and
now
we're
pretty
much
right
now.
This
is
this
is
pretty
much
done
here.
So,
okay
and
then
the
other
thing
is
that
looking
at
these
backslashes
here,
it
looks
like
this.
Isn't
one
of
the
code
blocks
that
we
we've
got
to
do?
We've
got
to
put
an
r
there
you
get.
A
If
you,
I
think
you
both
know
what
this
is,
but
for
the
sake
of
the
recording
the
when
you
put
the
r
there
this
means
don't
like
don't
don't
treat,
don't
treat
a
backslash
as
escaping
the
next
character,
treat
this
whole
thing
as
if
it's
like
a
regex
and
it's,
I
think
it's
called
a
raw
literal.
This
is
what
the
real
term
for
it
is.
So,
let's
say
python
come
on:
what's
a
real
good
string,
little
rules,
this
is
old
documentation,
but
still.
A
D
Page,
so,
have
you
already
added
the
support
for
like
what
this
will
run
on
what
os.
A
No,
I
have
so
yeah,
so
I
haven't
so
that's
one
of
the
things
with
the
tabs
doc,
the
tabs
plug-in
that
you
found
yeah,
so
the
the
okay.
So
basically,
one
of
the
next
things
that
needs
to
happen
is
yeah.
We're
gonna
need
to
run
them
on
the
other
os's
right,
we're
gonna
right
now,
so
right
now
it
runs
only
on
linux
for
all
of
the
examples
and
everything
that's
in
the
docs
right
now.
Actually,
let
me
just
finish
going
through
this
real
quick
and
do
this
last
one.
A
Doesn't
yeah
it
doesn't
run
those
those
so
right
now
those
code
blocks.
Don't
have
this
little
right.
There's
there's
code
blocks,
but
they
don't
have
that
test
guy
next
to
them,
so
they're
not
actually
being
run,
and
that's
basically
that's
one
of
the
next
things
we
need
to
do
with
that
console
test.
Plugin
is
to
go
in
there
and
look
at
it
and
say
and
say
so
so
in
this
docs
ext
console
test
in
here,
there's
a
there's
a
whole.
A
In
here-
and
this
is
what
I
need
to
document-
because
it's
a
bit
it's
a
bit
of
a
convoluted
mess
right
now-
and
it's
not
documented.
So
this
is
I'm
gonna
document
this
and
then
it'll
be
more
clear,
but
essentially
the
flow
of
this
is
that
all
right,
so
there's
this
thing
called
with
sphinx
you'll
have
things
there's
this
thing
called
a
builder
and
a
builder
is
like
you
know,
there's
the
html
builder
right
that
builds,
so
they
have
the
html
builder,
which
builds
the
html
version
of
the
docs.
A
They
also
have
a
version
of
the
builder
that
builds
man
pages
and
they
have
like
one
that
built
a
pdf.
So
basically
they
it's
split
into
two
pieces.
There's
the
parsing
of
the
there's,
the
parsing
of
the
documentation
files
and
then
there's
the
into
like
data
structures,
python
data
structures,
dictionaries
specifically
that
are
serializable
with
pickle,
so
anything
that
you
parse
in
you
have
to
be
able
to
serialize
out
with
pickle,
and
then
there
is
the
the
writing.
A
Out
of
you
know
basically
take
these
dictionaries
and
make
some
you
know
make
make
me
my
specified
format
right.
That's
what
the
builder
does
and
let's
see
so,
and
so
basically,
what
we
do
here
is
we
implement
this
builder
okay,
so
we
do
usually
what
you
would
do
is
you
would
implement
a
new
directive
right
and
a
directive.
So
a
directive
is
something
like
like
code
block
or
literal
include
anything
with
the
dot
dot
and
then
to
colons.
A
Is
a
directive
in
sphinx
speak
and
so
at
first,
what
I
did
was
I
implemented
ones
that
were
like
console
test,
hyphen
code
block
and
then
console
test
type
in
literal
include,
but
then
I
realized
the
syntax.
Highlighting
and
bim
doesn't
work
for
you.
A
Let
me
actually
use
func
tools,
wrap
and
make,
like
you
know,
make
a
new
run
function
like
so
assign
this
the
run
run
function
to
to
be
this
new
function,
which
is
going
to
call
the
original
function
as
if
it's
sort
of
like
super
dot
run,
and
that's
what's
happening
here
with
funk.
So,
basically,
you
can
think
of
this
sort
of
like
we
created
a
subclass,
but
we
can't
really
create
a
subclass
because
we
can't
up
like
we
can't
replace
the
class
within
sphinx.
A
We
can
only
modify
the
class
that
exists
there,
so
this
is
sort
of
you
know
it's
a
little
bit.
It's
yeah,
it's
it's
it's
a
little
bit
hacky,
but
it
the
interface
ends
up
being
way
nicer,
because
all
you
have
to
do
is
add
colon
test
colon
and
then
all
of
a
sudden,
your
stuff
gets
tested,
and
so
this
is
this
is
the
meat
of
what's
going
on
here
with
the
console
test
stuff
and
basically
it
it
parses
some
stuff
right,
and
that
happens
in
funk,
and
it's
like
okay.
A
Well,
you
do
your
regular
parsing
right
in
in
your
funk,
which
is
basically,
you
can
think
of
that
as
like
super
dot
run.
If
this
was
in
a
subclass,
because
this
is
the
run
function
and
then
we
go
through
and
we
we
start
looking
for
you
know
we
we
add
our.
This
is
our
custom
logic
right.
So
if
we
see
this
little
test
right
colon
test,
if
there's
a
colon
test
under
something,
then
it
ends
up
as
without
the
colons
in
self.options
right.
A
So
we
say:
okay,
if
I'm
supposed
to
test
this,
then
parse
all
the
commands
in
the
content
and
then
get
all
the
rest
of
these
little
options
that
we
have
and
there's
a
few
other
ones
that
I've
added.
So
basically,
like
the
you
know,
the
daemon
one
that
I
was
talking
about,
you
can
specify
the
standard
in
you
can
give
it
a
little
function
to
compare
the
output.
You
know
and
only
pass
the
code
block.
A
You
know
maybe
say
the
command
exits
successfully,
but
you
don't
want
it
to
you,
don't
want
it
to
continue
running
it
until
or
you
don't
want
it
to
continue
running
unless
it
exits
successfully
and
there
was
a
specific
stuff
in
the
output.
So
that's
the
compare
output.
So
basically,
all
of
this
logic
ends
up
here
right.
This
is
the
meat
of
our
our
our
logic
that
we're
adding
so,
and
this
is-
and
this
is
how
we
this
is,
how
we,
you
know,
get
access
to
the
parsing
functionality
of
this.
A
Functionality
of
sphinx
so
that
we
can,
you
know,
add
our
custom
logic
in
here
right
so,
and
this
is
just
like
I
was
saying
so.
This
is
all
about
reading
the
stuff
when
you're,
when
you're
in
these
directives
right
and
the
directives
or
the
code
block
and
the
little
literal
include
in
our
case
and
then
here
we
have
the
one
about
file
path.
A
So,
basically,
if
you're,
if
you're
doing
an
inline
file,
then
you
can
specify
the
file
path
and
it's
basically
going
to
take
everything
in
that
code,
block
and
write
it
out
to
a
file
right,
which
is
what
we
just
saw.
That
is
what
we
just
saw
here
with
the
train.json
and
the
test.json
right.
A
So
we
put
a
little
json
code
block
and
we
said
hey
put
that
in
test.json,
and
so
all
this
stuff
is
going
to
end
up
in
temporary
files
or
in
a
temporary
directory,
because
for
each
document
that
we're
testing
we
end
up
taking,
we
end
up
creating
a
temporary
directory
for
it
right
to
basically
say:
okay,
you
know,
for
example,
I'm
I'm
some
user.
I'm
like.
Let
me
run
this
tutorial
all
right.
A
Let
me
just
create
some
directory
that
I'm
going
to
run
it
in
and
then
let
me
start
running
my
current
commands
right
so
now
that
we've
parsed
all
the
stuff.
The
next
step
is
okay,
like
how
do
we
run
this
stuff
right?
So
that's
where
the
builder
comes
in
and
the
builder
what
the
builder
is
going
to
do.
A
A
That's
that
ends
up
being
called
a
node,
and
so
then
it
just
goes
through
the
nodes
and
it
it
so
it
builds
these
little
classes
called
commands
and
then
it
says
command
dot
run
for
each
one
of
them,
and-
and
so
I
mean
that's
the
gist
of
how
this
whole
thing
works,
then
there's
some
special
magic
that
happens
within
the
test
docs.
So
the
way
it
actually
gets
run
usually
the
way
you'd
run.
This
is,
if
you
look
at
scripts.
A
Docs.Sh,
so
usually
the
way
we
run
this-
and
this
was
some
funky
stuff
that
we
did
to
avoid
path
issues,
but
usually
it's
like
swing
spilled.
I
think
sphinx
build
what
is
it
sphinx
build
oh
dash,
b,
html
docs
pages,
so
this
was
this
giant.
Mess
here
is
just
to
find
swing
spilled,
so
that's
usually
how
we
do
this
right
and
then
that
says:
okay
use
the
html
builder,
build
it
from
the
docs
directory
and
put
the
output
in
the
pages
directory
now.
A
So
if
we
were
to
do
this
with
console
test
right,
we'd
say:
console
test,
we'd
say:
build,
console
test
and
use
the
docs
directory
as
your
input
and
use
the
console
direct
console
test
output
directory
as
your
output,
and
that
I
believe
I
mean,
let's
see
what
the
hell
happens
here,
but
okay.
Well,
it
didn't
like
that.
Oh
and
that's
why
I
think
we
were
doing
the
whole
giant.
I
think
that's
something
that
something.
A
I
think
that
error
has
something
to
do
with
this
situation,
which
is
it
wasn't
on
some
machines,
loading,
the
correct
versions
of
things.
But
the
point
is
that
if
you
run
that
it's
going
to
run
it's
going
to
run
that
builder
aux
ext
console
test
which
is
going
to
build
all
of
the
which
is
going
to
run
all
of
the
dock
tests
basically
or
run
all
of
these
console
tests.
Now.
A
Okay,
so
there's
a
bunch
of
code
in
here
right
now
that
I'm
going
to
take
out
that
was
hoping
to
get
full
coverage
information
out
of
all
of
this
because
right,
the
thing
is
that
these
are
test
cases
and
it
would
be
great
if
we
got
the
line
coverage
for
the
code
cup
plug-in
out
of
this
right.
A
But
unfortunately
it
doesn't.
It
doesn't
persist
through
sub
process
invocations.
So
if
you
start,
if
you
have
a
process
running
python
right,
if
you're
running
python,
using
code,
cov
and
or
using
coverage,
and
then
you
call
subprocess
exec,
another
python
file
coverage
is
not
going
to
persist
through
that
exec.
A
So
you're
going
to
not
get
any
coverage
information
there.
So
we're
not
getting
any
coverage
information
out
of
this,
unfortunately,
which
is
really
a
bummer,
but
eventually
we'll
we'll
try
to.
Basically
if
they
can
fix
that
bug
they.
Maybe
I've
basically
settled
on
it's
easier
to
fix
that
bug
in
coverage
than
to
do
it
in
here.
I
tried
doing
it
in
here
and
what
happened?
I
can't
exactly
remember
what
happened,
something
something
think
bad
things
happened,
but
yeah
it's
so
so.
A
A
It
doesn't
pick
up
the
changes
to
the
system
path
appropriately
and
so
basically
like
if
you
change
sys.path,
which
is
where
python
looks
for
all
of
the
modules
that
it
should
load
import
lib
has
a
function
called
invalidate
caches
which
actually
will
make
it
pick
up.
Those
changes,
however,
package
resources.
A
It
doesn't
appropriately
pick
up
those
package,
those
changes
and
everything
I
did
to
try
to
do
that
ended
up
in
fiery
doom.
Basically,
what
happens
is-
and
I
ran
into
this
problem
before
and
you
guys
may
run
into
this,
but
you
can
end
up
in
a
situation
where,
where
python
loads,
two
different
versions
of
a
module
for
two
different
modules,
and
that
I
don't
know
that
doesn't
quite
make
sense.
But
basically
what
happens
is
say
I
load.
A
Okay,
so
say
I
load,
and
this
is
what
happened
is
that
we
have
util
or
we
have
util
cli
command
right.
We
have
that
stuff
with
the
util
commands
right
now,
then
we
have
this
stuff
with
in
made
the
main,
slash,
cli
dot,
the
cli,
slash
cli.py
right.
We
have
okay,
so
when
in
dfml
cli
cli.py
right,
we
have
this
stuff
and
then
we
have
dffml
util
cli
command,
dot,
py,
so
okay
yeah
in
here,
so
it
starts
it.
This
is
this
gets
so
weird,
but
basically,
okay.
Let's
take
this.
A
For
instance,
when
we
load
these
you,
you
invalidate
all
the
caches,
you
say,
reload
all
the
modules
and
then
you
try
to
patch
package
resources
inter
entry
points
so
that
it
picks
up
the
changes
as
well
as
all
the
import
lib
stuff.
Well,
what
ends
up
happening
is
that
the
the
version
of
the
command
class
here,
that's
loaded,
is
actually
different.
There's
two
instances
of
this
class
in
memory.
A
The
definition
of
the
class
in
memory,
which
means
that
is
subclass
now
returns
false
here,
and
this
happens
all
over
the
place,
and
so
it
creates
a
giant
mess,
which
is
why
we're
not
getting
coverage
information,
but
I'm
sorry,
okay,
I'm
sorry!
I
took
you
guys
down
a
long
road
here.
A
But
now
you
know
how
everything
works,
sort
of
you
pretty
much
got
the
gist
of
it.
There
do
you
guys
have
oh
and
then
the
last
thing
I
wanted
to
say
is
basically
the
that
the
builder
stuff
there's
tech
stocks,
okay.
A
So
this
is
some
of
the
stuff
that
happens
with
the
builder,
where,
basically,
I
looked
in
the
sphinx
code
and
figured
out
how
they
actually
run
one
of
these
builders,
and
this
is
actually
how
sphynx
ends
up
running
one
of
those
builders,
and
so
what
we
did
is
we
just
made
a
just
basically
figured
out.
A
How
does
one
run
a
builder
and
then
run
it
program
program
programmatically,
so
that
we
can
only
run
the
build
on
one
file
because
or
else
sphinx
will
read
the
entire
source
tree
of
all
of
our
docs
every
time
which
takes
way
too
long
to
parse
all
that
stuff.
We
only
care
about
parsing
one
file.
So
that's
what
hap
that's
what
this
code
is
in
here
this
mk
test
case
or
whatever
it's
called
yeah
mk
test
case.
A
So
this
basically
does
sets
up
the
sphynx
spilled
stuff
and
says:
hey
only
read
only
read
that
one
file
that
we
care
about
building
and
then
only
test
that
one
file,
and
so
what
we
need
to
do
is
there's
the
last
little
bit
of
work
here.
A
Docstring
put
it
in
a
temporary
file,
now
run
the
test
on
the
temporary
file
right,
so
that
and
then
we
need
to
go
through
all
of
the
all
of
the
modules
or
all
of
the
yeah
all
of
the
plugins
and
make
it
so
that
they're
they're
doing
this
right,
and
this
is
how
we
end
up
testing
the
docs
for
that
class,
anyways.
Okay,
so
that
was
really
long.
Is
there
any
questions
on
that?
A
D
A
A
A
A
A
All
right-
and
so
basically
this
comes
down
to
well,
I'm
not
exactly
sure
what
it
comes
down
to,
but
some
people
tried
to
solve
it.
It's
been
open
since
2010,
so
that's
not
necessarily
optimistic
on
this
thing
getting
solved
that
it
hasn't
been
fixed
in
10
years,
but
I
don't
think
I
mean
everything
can
be
solved.
There's
like
you
know,
everything
is
possible,
so
it's
yeah,
it's
gonna,
be
there's.
There's
got
to
be
a
way
to
solve
this.
A
Actually,
I'm
kind
of
thinking
that
and
you
guys,
I
don't
know
if
you
guys-
you
guys,
probably
didn't
see
this,
but
this
is
a
random.
I
think
this
is
a
random.
A
August
distributed
orchestrator
stuff
here
I
can
pull
up
the
patch.
Actually
it
might
sort
of
make
sense
what
I'm
talking
about
here,
but,
okay,
where
is
it
so?
I
think
that
the
solution
may
lie
in
something
like
this.
Basically,
when
we
start
okay,
so
have
you,
I
don't.
Have
you
guys
seen
the
multi-processing
manager
stuff
within
python.
A
Okay,
let's
see
okay,
let's
see,
does
this
one?
Have
it
too
or
okay?
No,
all
right
and
let
me
pull
up
the
docs
yeah.
Let's
pull
up
the
docs
for
this
okay,
so
this
is
fun
stuff.
This
is
fun
stuff
so
and
it's
fun
to
play
around
with.
If
you
haven't
played
around
with
what's
going
on
here,
it's
it's
fun,
it's
worth.
If,
if
you're
bored
bored
one
day
and
like,
let
me
just
go,
try
something
out!
A
This
is
fun,
so
definitely
definitely
something
to
check
out
and
see
what
the
hell
is
going
on
here.
So
basically,
what
happens
is
that
so
the
multi-processing
module
is
used.
It's
it's
a
good
way.
It's
it's
the
best
way
to
actually
run
things
in
separate
processes
rather
than
threads
within
python,
and
so
it
and
and
and
the
main
reason
why
you
like
to
do
one
likes
to
do
this.
A
Is
that,
like
it
says
it
sidesteps
the
global
interpreter
lock,
which
is
basically
this
per
process
thing
that
it's
a
it's
a
it's
a
lock!
It's
like
you
know
what
is
it
called
like
a
well
okay,
you
guys!
Well
you
guys.
Probably,
you
probably
know
the
concept
of
locking
right
like
blanking,
on
what
the
hell,
the
right
term
that
it
goes
under
is
but
like.
A
Yeah,
so
so,
basically
they're
yeah
to
keep
everything
yeah.
So
so
python
has
this
this
global
interpreter
lock
and-
and
you
can
you
can
you
can
re
okay,
there's
this
guy?
That
gives
this
talk.
He
gives
this
talk
like
every
year
about
something
that
he
calls
the
gila
ectomy
gillette
me
larry
larry.
What
is
this
guy's
name
yeah
larry,
hastings,
okay
yeah,
so
this
guy.
This
is.
E
A
Okay,
so
basically,
if
you
want
to
learn
more
about
this,
this
is
interesting.
I
recommend
I
would
recommend
just.
A
Maybe
a
bit
of
this
okay.
A
Your
students,
okay,
okay,
so
basically
this
lets
us
run
multiple
processes
and
as
a
part
of
that,
some
of
them
you,
you
might
want
to
share
memory
with
some
of
them
like
some
you
you
could
have
you
may
have
you
may
start
them
all
from
the
same
place,
in
which
case
they
might
be
able
to
share
share
memory.
Now
you
could
start
them
on
different
actual
computers
and
then
obviously
they
can't
share
the
same
memory.
A
So
in
that
case
they
may
want
to
work
through
like
a
networked
situation,
and
sometimes
it's
just
easier
to
make
them
all
work
through
a
network
situation,
because
that's
yeah
well
you'll
find
out
what
you
mess
with
this,
and
so
basically,
what
we
did
here
in
this
patch
set
in
this
patch
was
basically
use.
A
This
use
the
idea
of
of
the
fact
that
so,
basically
you
can
take,
you
can
take
various
python
objects
like
in
classes
and
expose
them
through
this,
this
this
manager
class
and
then
basically
and
then
what
you
can
do
is
you'll
you'll
end
up
being
able
to
modify
them
in
a
different
process
or
on
it
from
a
different
machine
as
if
they
are
on
the
machine
that
you're
at
and
it
does
it
does
the
appropriate,
locking
and
handles
things
for
you
and
yeah.
A
So
I
think
this
may
be
part
of
the
key
to
to
you
know
it
could
be
helpful
in
this
situation
because,
basically,
you
could
exec,
you
could
exact
sub
process
and
then,
when
you
execute
a
sub
process,
you
say:
hey,
you
know,
is
there
a
manager
running
around
here
and
if
there
is
then
hey?
I
want
to
you
know,
connect
to
that
manager
and
you
know
grab
that
coverage
object
and
keep
adding
to
that
coverage.
Object
right.
A
A
It
would
be
great
if
we
could
get
the
full
coverage
information
right,
because,
right
now
our
coverage
stats
are
really
just
on
the
main
package
and
we
don't
cover
the.
We
don't
have
coverage
on
the
the
the
plugins
right.
So
it
would
be
good
if
we
could
get
that
information
and
running
figuring
out
how
to
orchestrate
all
of
that
grabbing.
A
All
that
data
and
putting
it
together
in
the
same
place
is
also
going
to
be
tricky
because
things
run
in
separate
jobs
and
if
you
guys
know,
this
is
something
that
I've
been
wanting
to
deal
with
recently.
But
the
installation
of
the
dependencies
is
becoming
a
really
long
thing
for
all
of
these
jobs,
especially
for
the
ones
that
need
all
of
the
dependencies
installed
like
the
docs
in
the
main
package
and
some
of
these
console
test
entries.
A
So
if
you
guys
know
in
git
lab,
I
know
you
guys
may
have
seen
that,
but
in
gitlab
there's
this
concept
of
a
pipeline
and
it's
kind
of
like
with
the
mermaid
js
diagrams,
where
you
have
so
the
first
thing
you
would
do
would
be
build
container
image
and
then
you
would
push
image
somewhere
and
then
you
would
let's
see,
then
you
would.
A
A
It
happens
in
every
single
one
of
our
checks
here,
which
now
there
are
71
of
or
not
every
single
one,
but
it
happens
in
in
a
substantial
enough
number
of
them
right
for
all
of
these.
A
They
just
download
whatever
ones
they
require,
but
then
for
the
ones
where
they're
integration
tests,
like
I'm
split,
I
also
split
all
these
out,
so
all
the
tests
will
run
in
their
own
thing
because
it
ended
up
running
them
in
parallel
and
it
runs
like
I
think,
like
10
minutes
faster
or
something
than
it
did
without
it
yeah
21
minutes
and
we're
down
from
like
29
minutes,
which
was
as
we
added
more
tests
it
kept
running.
You
know
in
series,
and
that
was
going
slower.
A
A
Please
please,
let
me
know-
or
or
you
know
add
that
here,
because
that
that
would
that
would
reduce
the
amount
of
random
network
errors
that
we're
hitting
and
and
speed
up
the
builds.
I
think
so.
Can
you
repeat
your
last
line
so.
A
So
if
you
find
a
way,
basically,
if
you,
if
you
find
a
way
to
if
you,
okay,
so
github
so
it'll,
be
some
kind
of
modification
to
this
workflow
file
right
and
these
workflow
files
right,
they
say-
okay
runs
on
you
know,
whatever
the
os
is
right.
Well,
there's
also
a
way
to
say
it
runs
on
a
container,
and
so,
if
you
find
a
way
to
have
one
of
these
jobs
build
a
container
and
then
the
next
one
say
that
it
runs
on
that
container
and
for
the
ones
that's
it.
A
What
I
want
to
understand
is:
is
there
a
way
to
make
the
jobs
that
run
on
a
container
wait
for
the
job
that
builds
the
container?
I
can't
I
haven't
been
able
to
figure
that
out.
So
if
you
figure
that
out,
please
speak
up
and
because
that
would
be
cool
so
yeah,
that's
basically
what
I'm
trying
to
say
there
lab
already.
A
Lab
has
that
yeah
they
have
this
concept
of
pipelines
and
so
yeah,
if
you
check
out
their
their
pipeline
support,
so
gitlab
pipelines
yeah.
They
I
mean
github
is
ahead
of
the
game
with
the
some
of
the
stuff.
Let's
see
they
should
have
a
nice
diagram
here
now
come
on
where's,
your
nice
diagrams.
A
They
have
real
nice
diagrams,
okay,
yeah
here
so
yeah
they'll
do
like
build
and
then
test
and
then
deploy
and
then
deploy
to
production,
and
so
you
can
gate
things
and
you
can
yeah
it's
more
fully
featured
why
github
has
become
so
slow
on
adopting
features
I
got
well.
I
guess
I
was
about
to
answer
my
own
question:
it's
because
they
have
a
much
larger
user
base
and
things
break
more
can't
do
new
features
all
the
time
when
you're
always
fixing
old
stuff.
A
So
I
guess
that
that
answers
the
question
there
of
why
they
may
not
have
this
yet
but
yeah.
So
that's
what
I'm
that's
what
I'm
hoping
to
find.
So
if
you
guys
find
it,
please
please
let
me
know
or
just
go
implement
it
because
that'd
be
great.
So.
C
A
A
A
The
other
thing
is
that
that
it
caches
some
of
the
pip
packages
right
now,
but
we
can't
cache
other
things
right
and
there's
other
things
that
we're
doing
like
installing
stuff
with
apt,
sometimes
and-
and
all
of
that
would
be
nice
to
be
able
to
cache
in
one
place
right,
because,
especially
if
this
this
is
the
other
thing
is
that
if
this
container
image,
so
the
container
image
is
only
going
to
change
whenever
the
dependencies
change
right.
A
So
if
any
of
and
I'm
about
to
basically
take
this
is
the
next
thing
before
we
do
the
releases.
We
need
to
be
able
to
pin
the
version
numbers
on
all
of
these
packages
right
all
of
our
dependencies,
because,
as
we
found
all
of
a
sudden,
somebody
updates
something,
and
then
it's
like.
Oh
well,
crap
like
something
doesn't
work
right.
So
when
we
we
want,
when
users
to
install
the
packages,
it
always
gives
them
a
known
good
configuration.
A
Even
if
it
may
be
not
the
latest
version
of
some
dependency
that
you
know
they
don't
really
care
they're.
The
whole
point
of
this
is
that
they
don't
have
to
care
how
they're
using
tensorflow
they're
using
tensorflow
right
or
pie
torch
or
whatever
right
like
we,
we
abstract
that
away
from
them
right,
so
we
need
to
be
able
to
pin
the
versions
so
that
only
when
we
do
a
release
we
talked
about
this,
but
only
when
we
do
a
like.
D
A
I'm
not
I'm
not
exactly
sure.
I
think
that
me,
I
did
see
some
stuff
about
that.
I
saw
some
of
the
azure
pipeline
stuff
and
intel
I'm
actually
I'm
part
of
this.
I'm
part
of
you.
A
You
guys
may
think
this
is
very
unsurprising,
but
I'm
part
of
the
I
forced
my
way
into
the
the
group
of
of
people
trying
to
figure
out
what
we're
doing
with
the
ci
situation
at
intel,
because,
obviously
I
have
lots
of
interest
in
making
sure
that
we
have
good
ci
systems
because
we
have
now
70
or
68
ci
jobs
so
I'll
be
damned
if
I'm
not
figuring
out
how
we
get
the
right
ci
system,
but
it
sounds
like
we
may
end
up
with
support
for
that.
A
But
I'm
also
not
sure
I
think
it's
a
paid
situation
as
far
as
I
can
tell,
but
if
it
ends
up
being
something
that's
free,
then
it
can
be
free
like
as
in
paid
as
in
it
won't
work
on
open
source
projects,
which
is
this
is
an
open
source
project.
A
E
Use
gitlab
pipeline.
A
A
If
we
can
figure
out
any
ways,
yeah,
it's
kind
of
just
it's
sort
of-
let's
see,
monitor
no
yeah.
So
basically,
if
we
can
figure
this
out,
because
I
mean
I'm-
I'm
not
able
to
see
it
right
now,
but
it
may
exist
right.
So
if
you,
if
you
see,
if
you
see
okay,
you
found
it
great.
A
A
Yeah,
so
maybe
let's
keep
this
around
because
this
may
we
never
know
right.
This,
maybe
will
work
for
us.
It's
not.
This
is
not
a
huge
deal
right.
The
other
thing
that
I
was
thinking
is,
we
can
also
maybe
make
it
so
that
we
have
this
okay,
so
here's
another
alternative
to
this.
A
We
make
it
so
that
we
have
so
so
this
can-
and
this
is
the
other
thing
which
I
was
talking
about,
which
is
why
this
is
kind
of
nice
is
because
that
container
image
is
also
a
container
image.
It
would
also
be
the
container
image
that
we
used
to
run
ci
locally
right,
so
we
would
know
we'd
have
a
uniform
environment
when
we
want
to
run
ci
tests
locally,
and
then
it
also
we
could
make
it
so
that
it
it
does.
A
Caching
correctly
where-
and
this
is
where
it
kind
of
gets
tricky
with
the
the
have-
has
there
been
an
updated
package,
but
I'm
sure
we
can
figure
it
out,
but.
A
We
can
make
it,
we
should
be
able
to
make
it
so
that
if
none
of
the
dependencies
change,
then
the
docker
image
rebuild
doesn't
happen,
and
if
the
docker
image
rebuilt
doesn't
happen,
then
we
that
that
six
or
seven
minutes
that's
happening
on
the
end
front.
End
of
every
single
test
now
just
goes
to
zero
right,
because
we
didn't
even
have
to
rebuild
that
whole
docker
image
right.
So
we
should
this.
Basically,
the
whole
point
of
this
is
I
want
to.
A
It
would
be
great
if
we
could
speed
up
the
ci
test
right,
because
the
the
slower
build
and
test
cycle,
we
have
the
better
right.
It's
it's
yeah,
but
this
is
all
sort
of
you
know
it's
minimally,
it's
minimally
important
compared
to
some
of
the
other
stuff
that
we
have
to
do,
but
just
I
just
wanted
to
put
that
out
there
in
case.
You
guys
see
something
like
that.
A
A
D
When
you
mentioned
the
core
logic
of
what
console
test
plugin,
you
wrote:
isn't
it
really
easy
to
like
include
tests
for
windows
and
mac
os
separately
like
it's?
It's.
A
Oh,
that's
what
I
was
yeah,
that's
what
I
was
supposed
to
be
talking
about
there
when
I
talked
about
the
nodes.
So
thank
you
so
basically,
yes,
it
should
be
really
easy,
because
the.
A
D
D
A
Exactly
yeah,
that
would
be
exactly
something
that
we
would
do,
and
actually
you
know
what
even
in
what
I'm
thinking
what
I
meant,
what
I
was
trying
to
say
when
I
was
talking
about
all
these
nodes
that
it
builds
here.
You
have
access
to
the
parent
node.
I
can't
remember
how,
but
you
have
access
to
the
parent
node,
which
means
that
when
you're
in
something
like
docs
installation,
where
we
have
these
group
tabs,
you
can
tell
if
you're
in
the
linux
one
or,
if
you're
in
the
windows
one.
A
So
that
may
be
also
a
way
we
can
do
that.
But
from
the
perspective,
I
think
I
think,
there's
two
takes
on
this
from,
because
I
think
the
console
test
as
a
thing
it
might
be
worth
splitting
into
its
own
separate,
open
source
project.
If
this
ends
up
being
you
know,
if
we,
if
we
find
long-term
value
out
of
this,
I
I
hadn't
seen
anything
else
out
there.
That
was
particularly
like
it.
A
There
was
a
few
other
things,
but
they
didn't
quite
give
the
same
functionality
like
the
ability
to
start
the
demons
and
stuff
and
manage
the
docker
containers
was
something
that
it
was.
I
need
we
need
that
right,
so
I
think
it
might
be
worth
splitting
into
its
own
open
source
project,
and
from
that
perspective
people
may
not
want
the
group
tab
thing.
They
may
want
it
as
an
option,
but
I
think
we
could
also
look
at
the
group
tab.
So,
yes,
I
think
you're
spot
on.
D
A
Well,
you
know
okay,
so
here's
the
thing
is
that
all
of
the
pipes
and
stuff
is
there's.
You
know
the
way.
I
have
written
a
lot
of
things
and
we've
all
written.
A
lot
of
things
is
especially
when
I
went
through
and
you
know
redid
some
of
these
it's.
I
definitely
use
pipes
a
lot
right
like
pipes
and
tea
and
right
so
there's
a
lot
of
piping
and
there's
a
lot
of
usage
of
like
tea,
the
t
command,
which
is
like
okay,
it's
basically
it'll
come
down
to
like
okay.
A
Some
of
these
things
use
grep.
They
use
t
they
use.
Curl
right,
like
is
this
stuff
installed
on
windows
right,
does
windows
have
pipes
and
is
it
installed
on
windows?
And
what
I
heard
from
someone
at
one
point
was:
basically
windows
doesn't
have
good
pipe
support,
or
something
like
that.
I
can't
remember
I
think
powershell
does
so.
We
could
just
tell
people
use
it
in
powershell
if
you
need
to
pipe
from
one
command.
D
A
Yeah,
so
I
think
the
thing
is
that
what
we'd
want
to
do
is
we'd
want
to
do
it.
In
the
builder
here,
we'd
say:
we'd,
look
at
this,
so
we'd
build
the
so
those
the
that
that
patching
of
the
run
code
block
run
stuff.
That
just
does
the
parsing
right
and
then
we
go
in
here
and
this
and
then
it
builds
all
the
commands
right
and
then
you'd
basically
say
hey
before
you
run
before
I
run
this
command.
A
You
know
you
do
something
like.
If
system
dot
platform
equals
equals
command
dot
platform
right.
If,
if
command
is
not
none
and
system
platform
equals
command
our
platform,
then
you
know
you
run
the
command.
I
think
that's
pretty
much
yeah
right
and
then
you're,
basically
going
to
say,
command.platform
equals
it's
going
to
be
something
like
that
form
equals
yeah.
That's
I
mean
I
think,
that's
pretty
much
gonna
be
and
then
there's
one
more
place
where
you
add
it,
which
is
down
here,
plot
form.
It's
it's
gonna
be
roughly.
A
That
is
going
to
be
pretty
much
your
set
of
changes
there,
but
yeah,
which
is
basically
check
the
platform.
And
then
you
know
yeah,
so
I
think
I
think
that'll
be
pretty
much
it
right,
we'll
need
to
go
through
and
we'll
need
to
put
a
platform
on
every
single
code
block
right
or
we
could
just
say
you
know
it's
probably
something
like
the
reverse,
where
it's
like.
Okay,
if
it
doesn't
have
a
platform,
then
run
it
if
it
does
have
a
platform
assigned
to
it
only
run
it
on
that
platform
right.
A
So
this
logic
is
probably
not
correct
here,
but
yeah
so,
but
I
think
you're
on
the
right
track.
There.
D
A
All
right
so
yeah,
I
guess
that's
that's
just
sort
of
from
what
I'm
doing,
I'm
gonna
basically
go
through
and
I'm
gonna,
I'm
gonna
I'll,
probably
change
a
couple
of
these
like
I'll.
Probably
do
like
you
know.
I
change
this
one
and
I'll
make
it
so
console
test
can
do
this
stuff
and
then
I'm
just
going
to
create
issue,
because
we
have
tests
for
some
of
this
stuff
right
now.
A
So
then
I'm
going
to
create
issues
and
or
I'm
going
to
update
the
documentation
and
say
that
this
is
how
we
should
do
this
going
forward.
Right
like
this
is
how
you
test
your
docs
string
is
by
you
know:
do
you
do
this
import
statement
in
your
tests
and
you
say:
console
test
my
class
right
and
then
it'll
test
your
classes
stock
string,
because
I
think
we
wanted
to
add
that
to
the
new
model
tutorial
under
the
packaging
or
whatever,
or
something
like
that.
A
So
I'll
do
that
and
I'll
document
the
console
test,
I'm
gonna
pin
the
release
numbers
and
then
I'm
gonna
roll
the
release
when
that's
done.
If
there's
anything
else
that
you
guys
wanna
see
in
the
for
the
0.4.0
release,
let
me
know
I
like,
because
we
can
wait
for
it.
Otherwise,
I'm
planning
on
I'm
planning
on
not
leaving
a
gap
in
between,
because
we
haven't
released
since
april
right
and
I
want
to
get
back
to
that
goal
of
releasing.
A
You
know
even
two,
every
two
weeks
right
things
just
got
way
too
way
too
out
of
control
and
and
it
needs
to
get
back
into
lots
of
stuff
in
you
know
like
less
stuff
in
every
release
and
more
often
releases.
So
that's
my
plans
right
now,
and
so
just
you
know,
let
me
know
if
you,
if,
if
there's
anything
that
we
need
to
change
before
then
because
you
know,
like
god,
forbid,
we
end
up
with
another
six
months:
release.
A
Then,
let's
make
sure
that
we
got
the
right
stuff
in
this
last
release.
So
you
know,
let
me
know
if
things
don't
look
right
and
and
we
need
to
do
more
stuff-
and
you
know
if
you
get
the
wind.
Obviously
it
would
be
really
great
to
see
you
know
to
know
exactly
what
is
supported
and
what's
not
supported
for
windows,
so
that
if
we
do
the
release,
we
can
say
you
know
to
update
that
installation
page
and
say:
hey
windows.
Is
these
things
are
supported
on
windows
right
because
it
already
seems.
B
A
A
It'll
be
good
if
we,
if
we
can
figure
out
you
know
what
is
the
list,
if
we
can
tell
people
hey,
you
know
this
is
what's
supported
and
what's
not,
that
would
be
really
great,
but
you
know
I.
I
would
also
say
we
we
got
probably
another
week
or
two,
maybe
before
I
can
nail
all
of
this
stuff
down
that
I
was
talking
about
there.
So
if
you,
if
you
completed
by
then
that
would
be
awesome.
A
If
not,
then
like
no
worries,
hopefully
we
can
get
the
next
release
out
after
that
soon
enough
that
then
we
put
it
on
the
docks.
So
all
right,
sorry,
I
know
there's
a
lot
of
information.
Sorry,
any
sweet
questions
comments.
Anything
else
you
guys
want
to
talk
about
today.
A
B
A
All
right
well,
thank
you.
Both.
It
was
great
to
talk
to
bill
today
and
hope.
Everything
goes
well
and
I'll
talk
to
you
guys
later.