►
From YouTube: Weekly Sync 2020-10-20
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.rpxtfoy7kvrb
B
Yeah,
okay,
yeah
windows
is
definitely
going
to
be
an
issue
at
the
moment
here,
so
so
yeah,
we'll
probably
we'll,
probably
want
to
get
you
we'll
probably
want
to
get
you
working
on
linux.
To
start
with
here,
because
I
think
we've
got
yash
is
working
on
getting
windows,
support
working
and
now
we've
got.
Let's
see,
we've
got
saksham
and
sudhanshu.
B
So,
let's
see
all
right,
okay
and
then
I
need
to
share
my
screen.
C
B
And
I
think
we
are
all
good
to
go
here
all
right
great,
so
so
we've
got,
we've
got
a
we've
got
a
newcomer
here
and
I
do
want
to
introduce
yourself
to
saksham
and
sudhanshu
they're
been
around
for
a
little
bit.
They
can
tell
you
what
they're
they're
up
to
you
can
tell
them
sort
of
your
interests.
A
Yeah
sure,
hey
guys,
I'm
sure
I'm
like
a
sophomore
in
new
new
delhi,
india
and
yeah.
I'm
excited
to
be
here.
E
D
C
B
Yeah
cool,
so
saksham
did
some
work
on
on
on
image.
Processing
models
over
the
over
the
summer
here
and
sudhanshu
has
been
work,
doing
some
work
on
accuracy
and
accuracy
scoring,
and
then
we
will
all
sort
of
do
those
are
sort
of
the
main
things
that
have
been
going
on
and
and
then
we
all
do
things
here
and
there
and
such
I'm
also
did
some
neural
network
stuff
with
pi
torch
recently
to
get
pi
torch
models.
B
B
All
right
accidentally
closed
my
tab.
That
was
the
meeting
all
right
so
yeah,
so
I'm
sort
of
the
main
maintainer
of
the
project,
and
so
we've
got
a
bunch
of
different.
We've
got
a
bunch
of
different
areas
right
so
we've
got.
Let
me
know
you
haven't
shared
your
screen
again.
Thank
you
and
we've
got
plenty
of
technical
difficulties
on
our
weekly
sinkholes.
B
So
we've
got
three
main
areas:
we've
got
and
let's
see,
look
we
and
then
we're
gonna
have
sort
of
four
once
sudhanshu
is
done
with
the
the
accuracy
scores.
Where
is
the
about
page
yeah?
Okay,
so
basically
yeah
there's
the
models,
oh
and
here's
yash.
B
So
we've
got
models
which
abstract
the
usage
of
machine
learning
models
right
in
the
creation
of
those
models
right,
so
you
basically
give
the
model
some
data
or
the
way
that
a
model
works
in
dffml
is.
Is
that
you
know
you,
you,
you've
got
this
class
and
the
class
will
create
a
trained
model
for
you
or
it
will
give
you
predictions
based
on
the
train
model
and
the
accuracy
stuff
is
sort
of
in
flux
at
the
moment,
and
then
we
abstract
the
data
sources
as
well.
B
So
that
that
way
you
don't
have
to
you
know.
A
specific
model
is
not
the
code
that
a
model,
the
code
that
we
use
to
program
the
model
is
not
tied
to
a
specific
data
source,
and
that
way
you
can
swap
out-
and
you
can
use
you
know
some
kind
of
a
database
or
a
csv
file
without
changing
the
code
of
the
models.
B
And
then
we've
got
this
data
flow
stuff,
which
is
basically
like
how
we
do
data
set
modification
and
generation,
and
so
you
said,
that's
just
sort
of
at
a
high
level.
What's
going
on
here
right
and
you
said
that
you're
you're
you're
interested
in
doing
some
stuff
with
the
models
right.
So
I
guess
my
first
question
here
is:
were
you
using?
Did
you?
Did
you
use
the
the
version
of
the
the
documentation
here?
Were
you
on
this
page?
Are
we
on
the
one
that
was
for
the
for
the
master
branch.
A
I
think
I
was
there
for
the
one
on
the
master
branch.
I
have
no
clue
where
this
one
is.
B
Okay,
okay,
yeah,
so
that's
that's!
This
is
basically
this
is
for
for
all
of
us
doing,
like
you
know,
development
work
and
we've
all
got
our
our
our
get
trees
pulling
down
from
the
master
branch
then
usually
we're
we're
referencing,
the
basically
it's
on
it's
on
the
bottom
of
this
front
page,
but
you
just
add
slash
master
to
this
url
and
that
sort
of
gives
you
the
the
stuff
that
hasn't
been
released.
B
Yet
so
the
working
copy
that
that
most
of
us
work
like
when
we're
working
everybody
who's
working
on
the
project
will
have
your
get
tree
pulled
down
right.
So
we
get
clone
the
project
and
then
we
work
we
work
on
it
and
and
then,
as
we
merge
things
into
the
master
branch
on
the
slash,
intel,
dfml
repo,
then
they'll
show
up
here
in
this
the
documentation,
the
changes
to
the
documentation
will
be
reflected
once
it's
finished.
Building
on
this
slash
master
url,
so
we
can
always
sort
of
see.
B
You
know
like
what
the
next
version
of
the
docs
is
going
to
looks
like
so
this.
This
is
the
this
is
the
master
branch
right
now
and
then
this
is
the
latest
released
version,
and
so
you
can
see
there's
like
some
differences
in
the
in
the
use
cases,
the
use
case,
documentation
and
stuff
yeah,
oh
yeah.
I
just
changed
this
so
anyway,
so
you
said
you
were
going
through
the
the
model.
B
And
so
what
were
you
and
you're
you're
doing
this
on
windows?
Right,
so
did
you
so
you
probably
ran
into
some
some
hiccups
like
we
talked
about,
but
so
what
what?
What
was
it
that
you
ran
into
and-
and
you
know
what
how
how'd
it
go.
A
I'm
sorry,
your
voice
is
like
really
breaking
a
lot.
Okay,.
B
A
F
A
Beginner
because
so
yeah
I
went
over
the
tutorials
page
and,
to
be
honest,
it
seemed
clear
what
I
needed
to
do
so
I
guess
I
have
installed
the
required
packages
and
I'll
get
started
as
soon
as
possible.
B
Okay,
cool
and
then
you
may
want
to
do
so.
So
let
me
just
write
down
where
we
try
to
take
meeting
minutes
in
this.
Let's
see
this
is
not
the
right
date.
Okay,
so
we
got
the
medium,
so
we've
got
medium
minutes.
We
hold
these
meetings
once
a
week
and
then
the
recordings
are
on
youtube
and
then
I
take
some
notes
in
this
dock
here
and
then
we
try
to
make
github
issues
out
of
things,
so
you
were
trying
to
go
through.
B
Is
to
go
through
the
new
model
tutorial
and
so
we'll
use
linux
or
use
ubuntu
machine
for
now,
since
we
are
still
in
progress,
making
windows
work.
A
D
G
A
G
For
windows
that
that's
exactly
what
we
are
working
on
right
now,
because
during
the
during
the
installation,
we
face
a
lot
of
errors
in
windows
specifically
and
we
haven't,
have,
we
haven't,
had
tests
for
windows
yet
so
it
would
be
difficult
for
you
to
develop
on
windows.
B
All
right
so,
okay,
yeah
so,
and-
and
so
basically
this
is
and
like
yes
was
saying,
you
know
you
got
to
get
everything
set
up
first,
and
so
this
is
like
sort
of
the
installation
page
for
if
you're,
not,
you
know
going
to
be
working
on
the
project
itself
right.
You
know
you
download
it
from
pi
pi
and
you
know
you
maybe
import
it
and
use
it.
B
But
if
you're
actually
going
to
use
you
know
you're
going
to
work
on
the
project
itself,
then
you're
going
to
want
to
go
to
the
you
know,
you're
going
to
want
to
go
to
the
contributing
pages
and
you're
going
to
want
to
look
at.
You
know
how
do
I,
how
do
I
we've
got
separate
docs
for
that
here,
because
you
need
to
install
things
in
development
mode
and
it'll
talk
about
virtual
environments
and
docker
containers.
B
If
you
want
to
use
that
so
and
then
the
last
thing
is
that
you
will
want
to
be
looking
at
the
master
branch
version
of
these
stocks
because
because
you're
you're
going
to
be
working
you're
going
to
be
pulling
down
the
master
branch.
So
you
want
the
docs
that
are
applicable
to
that.
So,
let's
see,
let
me
paste
this
in
here
so
now,
I'll
we'll
make
sure
you
have
so
get
set
up.
B
With
the
contributor
so
get
set
up
by
following
the
contributing
docs
from
master
of
branch,
so
we'll
put
the
link
here.
B
C
B
What
the
hell
so
you're
gonna
want
to,
you
know,
get
set
up
like
it's
talking
about
here.
Right
it'll,
basically
tell
you,
you
know
clone
it
install
it,
and
then
you
know
if
you
want
to
go,
and
you
want
to
look
at
the
models
and
adding
new
models.
I
think
yeah.
I
just
sort
of
changed
this
a
little
bit,
so
we
got
the
using
a
model,
then
writing
a
model
and
then
packaging
a
model.
So
it's
split
up
a
little
bit
more.
B
Try
running
through
the
new
model,
tutorials,
so
yeah
there
you
go
and
then
so
yeah
after
that,
basically
so,
and
and
then
the
last
one
here
when
you
go
to
packaging,
the
model
you're
gonna,
so
this
one
is
maybe
not
gonna
be
so
important.
It
depends
on.
We
can
talk
about
this
after
you've
implemented
a
model,
but
essentially
what
we
do
is
is
we
have.
The
whole
thing
is
a
bunch
of
plugins,
and
so
we've
got.
If
you
go
to
these
plugins
pages
here,
you'll
see
that
we've
got.
B
You
know
a
bunch
of
a
bunch
of
different
model,
plugins
right
in
these
wrap
different
machine
learning
libraries.
So
this
lets
people
you
know
switch
which
underline
implementation
that
they're
training
with.
B
So
if
you
want
to
use
one
of
these
specific
libraries
right,
if
you
want
to
so,
if
you
want
to
just
use
some
stuff
from
numpy,
then
you're
going
to
end
up
putting
it
in
the
in
this
model,
scratch
plug-in
and
if
you
want
to
use
because
that's
sort
of
like
numpy's,
basically
like
you
know
from
scratch,
without
some
kind
of
backbone
library,
because,
basically
that's
what
people
consider
numpy
to
be
so
so,
if
you
want
to
use
an
existing,
you
know
framework
and
you're
wrapping
you
you're
doing
a
kind
of
like
an
implementation
of
a
model
using
pi
torture
using
tensorflow
or
something
like
you've
got
some
specific
architecture
that
you've
you've
created
that
you
want
to.
B
You
know,
create
you
want
to
write,
write
an
abstraction
layer
around
that
specific
model
architecture
using
that
underlying
library.
Then
you're
going
to
be.
You
know
adding
to
one
of
these
one
of
these
plugins.
That's
like
the
dfml
model,
pi
torch
or
dfm
model
tensorflow
plugins,
because
the
the
way
that
it's
structured
is
that
you
install
whatever
one
of
these
plugins
that
you
you
installed
the
plugins
that
you,
you
install
only
the
stuff
that
you
need
right.
B
So
if
I
only
want
to
use
some
tensorflow
stuff,
I'm
only
going
to
install
tensorflow
right,
and
this
is
because
these
libraries
are
huge
right.
So
we
try
to
split
it
up
so
that
people
don't
have
to
install
everything
if
they
don't
want
to
so
basically,
you'll
go
through
the
tutorial
and
then
you'll
start
adding
code
to
one
of
these
specific
plugins
depending
on
what
what
dependencies
you
want
to
use
right.
B
So
if
you
want
to
use
tensorflow
you'll
come
in
here
to
dfml
model,
tensorflow
and
you'll
start
adding
code
under
here
and
you'll
we'll
get
to
that
we'll.
We
can
talk
about
this
as
as
you
get
to
it,
but
that's
just
to
give
you
an
idea
of
sort
of
how
things
work,
but
yeah
run
through
the
tutorials
and
then
just
sort
of.
Let
us
know
how
things
are
going
and
we're
on
gitter
pretty
much.
Usually.
There's
somebody
on
there
and
then
we've
got
the
meetings.
F
B
So
yeah
pretty
every
time,
there's
a
distinct
set
of
dependencies
right.
So
every
time
we've
got
a
new
distinct
set
of
dependencies,
we'll
create
a
new
plugin
and
that's
and
that's
because
right
we
don't
want
to
make
people
install
more
than
what
they
what
they
need
right.
B
So
if
you
want
to
use,
for
example,
if
you
were
using
yeah,
if
you
were
using
tensorflow
you'd
come
in
here
and
add
to
tensorflow
right
if
you're
using
you
know
these,
and
these
are
basically
the
existing
libraries
that
we
have
right
now
that
we're
wrapping
or
using
in
some
way
right.
So,
if
you're,
using
something,
that's
not
in
this
in
this
list,
then
you're
probably
going
to
end
up
creating
a
new
plugin.
B
Yeah,
so
it
depends.
You
know
what
kind
of
underlying
like,
if
you
were
to
use
some
other
library
to
create
that
convolutional
neural
network
right.
So
if
you
used
pytorch
to
do
that,
you
would
add
to
the
pi
torch
plug-in
here.
If
you
use,
you
know,
if
you
wrote
it
all
from
scratch,
using
just
numpy,
you
might
add
it
to
this
scratch
plug-in
right
and
if
you
that
do
you
see
where
I'm
going
with
this.
B
We
can
sort
of
figure
it
out
more
as
as
you
get
there
and
and
what
will
usually
happen.
Is
you
build
whatever
you
need
to
build,
and
then
we
figure
out
where
it
goes?
That's
usually
how
this
works
so
anything
else.
You
have
any
any
other
questions
here
or
otherwise
you
can
sort
of
you
can
you
can
hang
out
and
just
kind
of
see
see
what
else
is
going
on
so
far.
A
No,
nothing
that
I
can
think
of
right
now.
How
so,
basically,
once
I
get
set
up
and
go
through
the
tools
and
be
able
to
start
contributing
and
writing
code
right.
B
Yep
yeah
so
going
through
the
tutorials
is
a
good
place
to
start
and,
like
I
said,
I'm
in
the
progress
of
making
sure
I
remember
I'm
in
the
process
of
making
sure
all
of
these
tutorials
are
actually
tested
within
the
ci
right
now.
So
so,
if
you
find
something
that
doesn't
work,
let
me
know,
but
pretty
soon
here
we
should
be
sure
that
they
all
work
all
the
time
so
yeah.
Let
us
let
us
know
if
something's
not
working.
A
B
Thanks
cool
yeah,
thank
you
all
right
just
and
then
we
can
circle
back
at
the
end
here
and
and
see
if
you
have
any
other
other
things
that
you
wanted
to
talk
about.
Anything
gets
brought
up
that
that
you
want
to
think
about.
So
all
right.
So,
let's
see
we've
got
yes.
We've
got
sudhashu.
B
All
right,
yes,
did
you
have
any
so
so,
usually
the
way
that
we
do
this
also
is
we
go
through
and
we
say
sorry:
what
are
the
agenda
items
for
the
day?
So,
yes,
what
did
you
want
to
talk
about
today?
Anything
in
particular.
F
G
That
I
would
so
I'll
just
run
the
installation
tomorrow
for
sure
again
and
whatever
issues
I
face,
I'll
post
them
as
an
issue
with
the
screenshots
and
the
log.
Awesome
I'll
definitely
do
that
tomorrow
and
I
I
just
have
been
very
busy
last
week
so
no
ways
I
actually
have
a
meeting
at,
and
I
just
wanted
to
mention
this
cool.
B
G
I'll
start
working
on
this
from
next
week.
B
Nice,
okay,
yeah
that'll,
be
great
that'll,
be
great
yeah.
I
think
I'm
thinking
that
we
will
be
able
to
you
know
I'm
trying,
basically,
my
my
just
at
a
high
level
here.
You
know
my
like
I
keep
saying
we
want
to
get
the
release
out,
but
you
know
now.
I
I
saw
somebody
raise
an
issue
with
the
docs
recently,
and
so
I
want
to
make
sure
that
all
the
docs
are
tested
before
we
do
the
release.
B
So,
basically,
there's
two
steps
here:
there's
make
sure
all
the
documentation
is
all
the
tutorials
get
tested
within
the
ci
and
then
there's
pinning
all
the
version
num
all
the
version
numbers
of
all
the
packages,
and
so
those
are
the
two
things
I'm
doing
before.
I
do
a
release
and
then
we'll
try
to
get
the
our
release
cadence
back
on
track,
and
so
hopefully
we
can.
We
can
get
back
to
you.
B
It's
actually
yeah,
I
can.
We
can
I'll
I'll
go
over
that
here
after
we
get
all
the
agenda
items
and
we
can
start
talking
about
it
because
then
you
guys
will
want
to
start.
B
Using
information
I
created
a
sphinx
plugin
and
I'm
actually
going
to
try
to
I'm
going
to
publish
that
as
a
separate
repo
once
we've
got
sort
of
the
kinks
ironed
out
here
so
yeah,
because
I
looked
around
and
I
couldn't
find
anything
that
did
what
I
wanted
so
now
we
have
something
so
suit
honshu
anything.
You
want
to
talk
about
accuracy,
stuff
anything
else.
E
E
And
and
the
the
status
of
the
status
right
now
is
like
all
of
the
models
have,
they
are
fixed.
Okay,.
F
E
E
B
So
all
the
models
are
fixed,
except
so
transformers
and
psychic
strains.
Okay,
great.
C
E
B
Right
yeah,
we'll
we'll
check
it
out,
then
cool
anything
else.
There.
D
B
D
Yeah,
I'm
working
on
that
and
also
I
was
going
through
the
documentation
and
I
think
we
need
to
fix
the
python
documentation.
Okay
and
that's
so.
B
D
B
Oh
yes,
we
may
need
to
do
something
more
like
the
scikit.
You
can
check
the.
B
C
A
D
A
D
On
that,
because
the
image
killerization
stuff
I'm
working
on,
if
we
can
integrate
the
stuff
like
using
pre-trend
models
with
custom,
neural
networks,.
F
D
B
I'm
going
through
sweet,
sweet,
okay,
so,
let's
just
put
in
let's
put
in
a
oops,
let's
put
an
issue
for
this,
because
I
can
see
this
getting
lost
plugins
models.
Let's
see
oh
yeah,.
E
I
D
B
D
For
now
right,
yeah
yeah,
so
I
just
wanted
to
bring
that
up.
It
almost
got
off
my
data.
B
Okay
and
what
was
that
again,
the
this
is
the
loss,
functions
and
everything.
B
D
Pointing
for
the
loss
functions
right
now,
but
we
are
using
get
adder
for
the
optimizer
optimizers,
because
the
document
we
don't
have
annotations
yet
in
the
latest
version
of
touch
vision.
Oh.
B
D
We
are
doing
this
to
to
do
the
entry
point
stuff
we're
not
getting
rid
of
it.
Oh.
D
I
B
C
B
In
the
setup
pyi
right,
yeah,
okay,
great,
okay,
yeah
and
right
now,
basically,
we
couldn't
choose
the
optimizer
right.
D
I
B
Okay,
so
once
we
have
the
next
release,
we'll
do
the
exact
same
thing
we
did
for
the
loss
functions.
The
and
you'll
also
be
working
on
getting
the
pre-trained
networks.
B
To
work
with
the
custom
defined
ones
right
to
link
them
together
is
that
we
were
saying.
F
B
B
All
right,
great
okay,
so,
let's
see
all
right.
So
let's
talk
about.
Let's
talk
about,
okay,
let's
talk
about
console
tests
first
in
case
anybody
has
to
go
because
I
want
you
guys
all
to
hear
this
and
then
I'm
sure
I
can
talk
about
accuracy.
B
Okay,
so
let
me
take
you
all
to
the
recent
changes
and
I
was
working
working
on
this
and
then
I
hit
issues
with
issues
with
conda,
because
the
ci
uses
conda
and
so
the
virtual
buy
stuff
doesn't
work
right.
The
same
all
right,
let's
see,
where
is
this.
C
B
Okay,
this
might
be
okay,
all
right,
so
this
is
an
example
of
of
I
just
did
this
one
recently,
but
this
is
how
I
modified
the
the
file
source
tutorial
to
be
tested
using
the
console
test
plugin.
So
essentially,
what
you'll
see
is
that
there's,
it's
really
just
the
addition
of
these
little
test
options
to
the
two
sphinx.
B
So
basically,
you
put
your
code
block
right,
just
like,
as
you
normally
would,
and
then
you
put
test-
and
that
tells
it
that
this
is
a
part
of
the
the
tested
stuff
in
this
file
so
and
the
way
that
it
works.
Is
that
so
there's
some
there's
some.
So
I
did
some.
I
did
some
digging
into
the
sphinx
code
and
yeah.
You
can
see
that
over
here,
but
so
I
got
it
so
that
I
got.
B
We've
integrated
it
with
the
unit
test,
so
each
each
file
becomes
its
own
test
function
within
the
unit
tests
and
so
you'll
you'll
you'll
be
able
to
run
them
and
I
haven't
updated
the
document,
so
I'm
going
to
update
the
documentation,
but
this
is
just
me
telling
you
guys
what's
going
on
that?
Basically
you
can.
You
can
run
them
using.
C
B
Bigger
all
right
so
yeah,
the
gist
of
it,
is
that
it's
test
console
test
test,
docs,
test
tutorials,
and
so
then
you
do
test
underscore
and
then
the
path
to
the
to
the
file,
but
with
the
under
the
slashes
replaced
with
underscores.
So
if
you
want
to
test
the
docs
slash
tutorial
sources,
file,
dot,
rst,
you
do
test
tutorial
sources
file
and
it's
broken
right
now
with
the
non-conda
setup.
B
But
basically
what
you'll
see
is
that
it
goes
through
and
it
runs
each
of
those
commands
and
see
it's
not
supposed
to
be
doing
conda
right
now,
but
it
is,
and
so
that's
and
basically
all
that
all
you
need
to
do
is.
Is
you
add
the
so
you
can
you
not
render
this
okay
so
yeah?
B
You
just
add
this
test
wherever
you
wanted
to
run
these
blocks
here,
and
so,
if
you
have
like
a
file
so
then
the
next
thing
have
the
the
next
part
of
this
is
basically
the
you
know.
B
If
you
need
to
write
specific,
like
we
always
try
to
keep
python
code
in
python
files
for
the
most
part,
so
that
we
can
have
it
so
that
it's
formatted
with
black
and
actually
I
think
this
is
probably
an
instance
where
this
needs
to
happen
here
too,
but
you
can
it
works
for
python
or
csv
or
whatever.
So,
basically,
if
you
have
a
little
block
of
code
that
you
of
something
that
you
want
to
actually
maybe
at
the
end,
here's
a
better
example
right.
So
here's
this
any
file
right.
B
So
we've
got
this
code
block
any
and
we
want
to
have
this
be
in
a
data.any
file
within
within
within
the
testing
environment
and
so
the
the
testing
environment.
The
way
it
works
is
it
creates
itself
its
own
directory
right.
B
So
we
get
a
fresh
temporary
directory
where
all
this
stuff
is
running
and
we
get
a
fresh
virtual
environment
that
we
can
install
things
into
right,
and
so,
when
you
create
files,
it'll
get
created-
and
you
know
whatever
the
the
current
working
right,
whatever
that
temporary
directory
is
right
and
if
you
do
cd,
then
it'll
change
the
directory
right,
just
sort
of
like
normal
and
obviously
not
so,
there's
a
there's,
a
basically
a
sort
of
an
emulated
shell
environment
right.
That
gives
us
control
more
enough
control
that
we
can
actually
do
the
testing
correctly.
B
So
it's
not
really
it's
sometimes
you're
running
bash,
but
a
lot
of
the
times.
You
know
unless
you
say
bash,
and
then
some
bash
file
you're
going
to
be
running
in
this
in
this
environment.
That's
basically
it's
within
the
docslash
est.
Ext
consoletest.py
is
basically
a
bunch
of
sub
process
run
commands
so
that
we
can
manage
the
virtual
environment
and
the
just
various
other
things
where
we
have
to
change
the
environment,
variables
and
stuff
so
that
we
can
have
full
control
over
this
testing
environment
all
right.
So
you
basically
yeah.
B
You
just
add
tests
to
these
code
blocks
and
then,
if
you
want
to
write
out
files,
you
put
tests
there
and
then
you
put
the
file
path
you
want
to
write
out
to
and
then
this
you
know
whatever's
in
this
block,
it's
written
out
to
that
file,
and
so,
for
example,
here
right
we
did
all
right.
Here's
this
data
dot,
any
write
it
out
to
this
data,
dot
any
file
and
then
now
run
this
code
block
here
and
so,
which
means
we're
going
to
run
this
command
right.
B
B
The
the
output
contents
kind
of
like
the
doctest
plugin
does
or
then
the
doc
tests
do,
but
it's
just
going
to
check
the
status
code
and
make
sure
it's
zero
right
so
that
the
command
exited
zero,
which
is
you
know,
the
return
success
code
for
for
unix,
commands
right
and,
let's
see,
and
then
oh
yeah.
The
other
thing
is
the
last
thing
would
be.
The
literal
includes
right,
so
it's
all
the
same
sphinx
stuff
that
we've
been
using.
B
Basically,
we
just
added
a
few
directives
or
a
few
a
few
options
to
these
directives
and
the
directive
is
something
like
little
literal
include
or
code
block.
B
So
if
you
want
to
have
this
file,
so
if
you
have
a
little
include-
and
you
want
to
make
the
file
that
you're-
including
appear
in
this-
you
want
to
copy
it
over
to
the
this
temporary
directory
that
we're
using
as
our
current
working
directory
for
the
tests.
You
need
to
have
you
put
tests,
you
put
tests
there
and
then
you,
you
can
put
the
file
path.
If
you
don't
put
the
file
path,
it's
just
going
to
copy
the
it's
going
to
just
going
to
use
the
whatever
the
last.
B
The
file
name
is
and
put
it
in.
There
is
the
file
name,
and
this
is
just
diff
stuff,
so
you
can
also
that's
just
styling,
like
usual,
oh
yeah,
and
then,
if
you
use
the
lines,
if
you
use
the
lines
option,
basically
it's
going
to
look
in
just
just
like
it
does
usually
right.
So
what's
a
this,
is
this
is
kind
of
good?
B
Okay,
here's
a
better
example
right,
so
this
tutorial
we
go
through
and
we
write
this
dfml
source
underscore
any
slash,
misc,
dot,
py,
and
we
start
with
this
little
block
here
right,
and
so
it
writes
it
out
to
this
file
right
and
then
we
want
to
so
the
and
the
reason
why
we
do
this
is
because
the
reason
why
we're
having
this
specific
header
here
is
because
the
rest
of
this
tutorial
is
actually
pulling
from
the
any
source,
that's
included
in
within
the
main
code
base
here
right,
and
so
it's
we.
B
B
B
Now,
basically
we're
telling
the
user
write
this
stuff
in
this
file
right,
that's
what
we're
asking
them
to
do,
and
then
we
ask
them
to
please
write
this
stuff
in
this
file
and
then
this
stuff
right
and
so
every
time
we
come
down
or
every
time
we
see.
This
little
include
with
these
lines
we're
going
to
write
those
specific
lines
right.
So
we're
only
going
to
write
these
lines
to
that
file
right
and
we're
going
to
do
an
append.
B
It
assumes
that
you're
appending
to
the
file
and
if
you
want
to
change
it,
so
if
you
have
a
tutorial
that
doesn't
work
exactly
like
this
right,
so
if
you
have
a
tutorial
that
needs
to
do
something
else,
we've
got
the
basically.
The
plugin
is
within
the
code
base.
Right
now
and
that's
what
I
was
sort
of
saying
by
you
know
we'll
we'll.
Eventually
we
might
make
this
console
test
plugin
as
or
console
test,
sphinx
extension
its
own
project
right.
It
doesn't
really
need
to
be
a
part
of
dfo
from
all.
B
Other
projects
might
use
this
right,
but
what
we
need
to
do
is
we
need
to
figure
out
how
to
test
all
our
tutorials
right,
and
so
some
of
these
require.
You
know
we're
going
to
run
into
interesting
cases
here,
but
this
is
sort
of
the
basics
and
you'll
see
stuff
where,
like
let's
see,
tutorials
models,
slr,
okay,
yeah,
so
with
the
slr
model,
we
end
up
yeah,
so
little
literal
includes
no
file
path,
means
just
make
it
myslr.py.
B
Whatever
the
last
thing
is:
here's
some
csv
examples.
It
doesn't
know
how
to
highlight
csv
for
some
well,
there's.
No
real
highlighting
to
be
done
there,
so
I
didn't
put
the
the
highlight
type
and
let's
see
where
does
this
get
interesting,
yeah,
so
the
http
server
so
with
the
http
server.
So
when
you
that
this
is
one
of
the
things
where
we
need
sort
of
custom
control
over
stuff
right?
B
So
if
we
start
the
http
server,
we
need
to
make
it
bind
to
a
random
port
right
because
you
might
be
using
port
80
for
8080
for
something
right,
so
we
gotta
basically
you'll.
If
you
look
in
the
console
test
plug-in
or
the
extension
code
in
docs
underscore
ext
console
test
dot,
py
you'll
see
that
there's
some
there's
some
code
to
handle
the
the
having
this
actually
bind
to
a
random
port.
B
So
it
basically
replaces
this
8080
with
zero,
and
you
guys,
if
you
guys,
have
done
some
networking
code.
You
might
know
that
when
you
ask
the
system
bind
to
port
zero,
what
it
actually
does
is
it
assigns
you
a
random
port,
and
so
what
we
do
is
we
grab
that
random
port
number
and
we
put
it
in
this
sort
of
con.
This
context,
that's
present
for
the
lifetime
of
the
test,
and
this
context,
that's
present
for
the
lifetime
of
the
test
holds
things
like.
B
What's
the
current
working
directory
for
this
test,
and
it
also
holds
things
like
you
know
what
http
servers
am
I
running
and
on
what
ports
did?
I
start
them.
So
I
told
this
one
to
start
on
port
8080,
but
within
this
context
this
this
object.
We
actually
have
what
port
it
actually
started
on,
which
is
the
randomly
assigned
port.
B
So
you
know
you
can
add
things,
and
this
is
where
you
know
as
we
go
through
and
we
write
tutorials
we're
going
to
find
that
we
have
specific
use,
use
cases
that
aren't
that
that
this
doesn't
cover
yet
right.
So
we
need
to
go
change.
The
code
within
the
console
test
extension
to
make
sure
that
we
can
test
everything
that
we
need
to
test
right,
and
so
one
of
these
things
was
basically
okay.
Well,
I
know
we
know
we
need
to
be
able
to
replace
the
port
number
with
the
one
that
was
randomly
assigned.
B
So
we've
got
this
little
replace
option
here
and
basically
what
this
does
is.
It
runs
this
python
code
and
lets
us
edit
the
command.
That's
going
to
be
run
to
replace
to
replace
the
port.
This
8080
right
it
goes
and
replaces
8080,
with
whatever
the
randomly
assigned
port
was
right,
and
so
that's
the
kind
of
stuff
where
and
there's
one
more
example
of
this.
But
but
this
is
the
kind
of
stuff
where
there's
specific
use
cases
right
because
of
the
you
know,
because.
B
These
tutorials,
like
there's,
gonna,
be
very
specific
things
that
we
need
to
have.
You
know
you
know
python
code
to
help
us
do
right
and
that
python
code
goes
in
the
console.
Test.Py
file
and
figuring
out
and
you'll
see
how
it's
structured.
Basically,
there's
a
couple
classes
in
there
and
stuff,
and
you
you
just
you,
you
add
code
there
and
you
make
it
you
make
it
work
right.
B
The
point
is,
we
want
to
be
running
these
commands
right,
but
these
commands
may
need
a
little
help
to
get
set
up
and
torn
down
right
around
around
what
they
need
right.
So,
let's
see
examples,
docs
examples,
integration.
C
B
Oops
well,
the
other
one
is
the
the
integration
example
and
on
this
one.
B
Apparently,
I
can't
spell
it,
but
on
this
one
here,
we've
got:
let's
see
this
one
starts
docker,
containers
and
stuff,
so
there's
another
one,
you'll
you'll
notice.
If
you
go
in
that
file,
you'll
notice
that
there's
yeah
the
daemon
option
basically
says
this:
the
last,
the
whatever
the
last
command
in
this
series
will
stay
running
for
the
lifetime
of
the
test
rate,
so
the
cgi
server
runs
for
lifetime
of
the
test.
B
Oh,
and
I
think
this
might
need
to
be
or
no.
This
is
an
old
version,
oh
yeah,
and
then
there's
basically.
So
we
started
my
sql
server.
This
is
old,
so
I
changed
this
obviously,
but-
and
so
you
can
say,
like
basically
run
this
command
until
you
see,
you
know,
pull
this
command
so
run
whatever
the
command
is
in
this
block.
B
Until
you
see
some
output
right-
and
so
this
this,
this
command
we're
telling
the
user
to
wait,
keep
running
this
command
until
the
mysql
server
comes
up
and
we're
checking
for
it
by
grepping,
and
so
we're
basically
replicating
that
here
and
ignore
errors
means
don't
fail
the
whole
test
if
grep
returns
non-zero,
which
it's
going
to
do
until
it
it
finds
ready
for
twice,
and
I
think
there
was
one
more
thing:
yeah
the
docker
stuff.
Basically
it
it
calls
docker
rm
on
the
running
container
when
it's
done,
but
I
think
that's
pretty
much.
B
It's
pretty
much
how
it
works.
I'm
gonna
go
document.
I
know
that
was
a
lot
of
information,
but
I'm
gonna
go
document
it.
You
know
thoroughly
with
a
whole
documentation,
page
and
and
so
then
so
I'm
gonna
go
through
and
I'm
gonna
convert
the
docs
as
they
exist.
You
know
at
the
moment,
but
then,
as
you
guys,
add
more
stuff
and
we
all
add
more
stuff,
then
obviously
you're
going
to
you're
going
to
do
this
yourself.
B
So
I
just
want
to
make
sure
that
we're
I'm
covering
all
the
use
cases
that
exist
first
before
I,
you
know
throw
something
that
you
know
so
that
I
can
throw
something
that
works
at
you
guys,
instead
of
something
that
doesn't
work
right
because
yeah,
that's
that
would
be
no
fun.
So
does
that?
Does
anybody
have
any
questions
on
that
or.
C
B
Okay,
all
right:
okay,
because
that
was
a
bit
of
a
got
a
little
rambly
there.
So
all
right
I'll
make
sure
that
it's
more
clear
when
we
do
the
docs
so
gabe
overview.
C
B
Docs
yeah,
let's
see
well,
if
I
remember
docs
correctly,.
F
E
I
E
So
actually,
this
is
so
in
the
like
for
the
pip
like
in
the
and
the
code
of
the
people.
They
also
use
this
same
frame.
B
B
Yeah
this
is,
I
mean
this
is.
I
know,
talks
is
good.
I
pref
I
hadn't.
We
haven't
used
talks
for
this
because
it
sort
of
replicates
the
stuff
that
github
ci
ends
up
doing
where
it
creates
different
environments
right
for
different
versions
of
python
and
stuff
and
that's
kind
of
then
we
lose
the.
Then
we
lose
the
different
line
items
and
the
ci.
If
we
do
that.
B
B
Instance,
yeah,
I
would
love,
I
would
I
mean
I
would
love
to
use
something,
that's
more
robust,
let's
see
if
it
exists,
let's
see
because
the
real
I
mean
so
the
the
main
thing
here
is
that
we
want
to
be
able
to
test
okay
generate
documentation.
We
want
to
be
able
to
test
the
restructured
test
that
restructured
text
files
right
because
we
have
the
unit
test
stuff.
Well
we're
going
to
have
to
re-figure
out
the
unit
text
stuff.
B
I
B
B
What
so,
just
just
to
be
clear.
The
reason
for
this
was
because
I
we
we
want
to
make
sure
that
dox.
B
Literal
include
the
sh
file
and
then,
for
example,
I
mean
like,
with
the
model
with
the
model,
docs
plugins
models,
different
model
right,
so
right,
we've
got
all
this
type
of
flow
here
right
we
have
all
these
literal
includes,
and
then
we
write
a
separate
python
test
script
to
go
test.
Those
you
guys.
B
Is
to
to
avoid
that
right
is
to
is
to
make
that
you
know
and
right
now
right
now
it
we
don't.
I
mean
I
guess
we
could
put
it
all
in
here.
We
could
put
it
all
in
here.
I
guess
I
just
I
haven't
I'm
doing
one
by
one,
and
so
I
haven't
I
haven't
done
that
yet,
but
so
yeah.
The
idea
is
basically
to
take
this
and
turn
it
into
into
this.
C
B
Block
console
test,
and
that
way
you
know
it's
just
a
little
more
it's
it's,
we
can.
We
did
the
the
document
we
we
reduce
with
the
boilerplate
right,
because
essentially,
every
time
we
write
documentation,
we
end
up
having
to
write
some
test
code
right
to
go
with
it,
and
so
with
this
approach
we
don't
we
don't
do
that
and
so,
like
I
said,
I'm
open,
I'm
open
to
other
plugins.
That
may
be
more
robust
and
more
well
established.
I
couldn't
find
anything
that
that
did
this.
B
When
I
looked,
I
saw
something
that
was
about
testing
tutorials
with
as
a
sphynx
plug-in,
but
it
wasn't
clear
to
me
that
it,
you
know,
had
the
support
for
the
the
background
processes
as
demons
and
stuff
like
that,
which
is
a
pretty
common
thing
that
we
have
to
deal
with
with
docker
containers
or
at
least
within
the
fewer
the
tutorials,
where
we
have
docker
containers
and
http
servers,
and
so
you
know
I
need
it.
We
need
a
solution
that
can
handle
that.
B
So,
if
you
guys,
if
you
find
anything,
let
me
know
I'm
not
I'm
not.
I
don't.
I
don't
think
talks.
Actually
I
mean
if
you,
if
you
know
if
or
if
you
know
more
about
toxin
it
can
do
that,
then
let
me
know,
but
I'm
kind
of
skeptical.
B
C
B
Oh
yeah
and
then
basically
this
right,
so
these
are
the
this
is
sort
of
the
main
one
of
the
main
pain
points
here
is
that
when
we're
writing
the
when
we're
writing
the
tests
for
the
plugins
themselves
right
and
we
we
want
to
write
the
little
blurbs
of
you
know,
what
is
what
is
the
command
line
usage
of
this
plugin?
And
what
is
the
and
I'll
open
it
up
here,
so
that
we
can
actually
see
the.
I
Main
code
model
tensorflow.
B
So
right
when
we're,
when
we're
writing
these
doc
strings
here
right
now
this
now,
this
will
start
looking
like
this
and
that's
it's
going
to
be
a
little
more,
a
little
more
more
easy
to
parse,
because
you
don't
have
to
then
go
open
that
file
and
go
open
that
file
and
go
open,
app
file
to
see
what
the
hell
hat
is
going
on
right.
So
the
next
step
here
is
really
I'm.
I'm
kind
of
thinking
that
we'd
like
to
get
this.
B
B
That
is
that
it's
a
unit
tested
we
get
this
stuff
gets
run
within
our
unit
tests
and
so
we'll
basically
add
this
to
the
unit
tests
of
every
plugin
so
that
the
plugins,
so
they
they
run
their
tests
within
their
line,
item
and
github
actions,
and
we
don't
sort
of
have
everything
end
up
running
within
the
main
test
suite
because
that
would
be.
You
know.
B
That
would
be
a
little
bit
much
to
run
every
single
model's
command
line
tests
within
the
main
test
suite
that
would
make
that
that
ci
test
really
long
but
anyways
all
right.
So
let's
talk
about
any
any
any
sort
of
follow-up
questions
to
that
or.
B
Cool
so
let's
see
yeah
so,
okay,
so
all
right!
So
let's
take
a
look
at
the
the
pr.
E
Okay:
okay,
okay,
so
in
the
transformers
tests,
what
we
are
actually
doing
is
we
are
taking
the
accuracy
scorer,
but
we
are
not
using
it.
B
E
Yes,
so
I
have
just
implemented
that
that.
B
E
B
E
E
Yes,
okay,
so
is
actually
failing
because,
like
there's,
some
dependency
dependency
issues.
B
C
B
E
F
E
Let's
see
or
maybe
like,
we
need
to
like
use
some
lower
version
of
transformers.
B
Yes,
so
oh
yeah
look
at
that
10.15.,
so
that
was
just
the
other
day,
so
I
guess
well.
The
question
really
is
like
how
much
has
changed
right
so
because
yeah,
you
see
what
I'm
saying
so
the
question
is:
if,
if
it's
passing
in
master
then-
and
this
is
obviously
not
passing
because
of
that
stupid,
conda
thing-
but
it
looks
like
it's
passing
over
here,
so
the
question
is,
then
what
this
is,
what
I
usually
do
is
when
we
run
into
a
situation
like
this.
B
B
B
B
Okay,
so
basically
we're
gonna
diff
the
setup.py
file
right
between
your
branch
and
master
branch
and
see
what
the
requirements
are
right
and
if
the
requirements
are
the
same,
then
it
doesn't
really
it's
it's
it.
It
doesn't
really
make
much
sense
that
it
would
be
well.
This
is
the
master
branch,
so
it
doesn't
really
make
much
sense
that
it
would
be
pulling
a
different
version.
B
Oh
okay,
and
maybe
this
is
why
okay,
so
good
fetch
right.
So
we
basically,
I
think
it
looks
like
that
first
commit
there's,
probably
what's
going
on.
So
let's
do.
B
B
I
made
failed,
commit
to
exclude
that
and
then
let's
see,
okay,
so
there's
this
3.2
from
kamacho
on
july
10th
we've
got
the
numpy
thing
that
happened
and
which
must
just
work
now
and
then
I
screwed
up
on
excluding
that
version,
and
then
I
actually
did
it
correctly.
So
I
I
said
only
only
installable
blow
version
3.1.0,
so
it
looks
like
you
maybe
need
to
cherry
pick.
This
commit.
B
Okay
yeah,
so
I
just
all
I
did
was
you
know
I
went,
I
pulled
master.
I
cherry
picked
that
commit
so
so
hopefully
I
mean
that
right.
So
this
is
this
should
pin
it
below
that
version
and
I
think
basically
our
answer
is
we
have
an
open
issue
that
tracks
this.
B
It
looks
like
so
and
once
again,
yeah.
I
think
we
need
to
talk
to
himanshu,
so
I
think
he's
been
busy.
He
got
that
job
so
yeah,
okay,
so
that
I
mean
at
least
at
least
at
least.
We
know
that
this
has
been
an
issue
for
a
month
now,
so
it
wasn't
anything
that
you
did,
which
is
good,
so
well,
we've
gotta
fix
right.
Is
there
okay,
so
and
then
so,
let's
see
let's
reference
this,
so
I
guess
it
was.
B
Yeah,
so
you
can
see
if
that
fixes
that
okay,
and
so
we
still
need
to,
is
there
so
so
from
from
looking
at
it
right
now
or
reviewing
it
right
now
standpoint,
what
let's
see?
What
should
we
be
looking
at
here.
B
C
B
B
B
Yeah,
I
don't,
I
don't
think
that's
that's
sort
of
what
I'm
saying
is.
I
don't
think
we
really
need
to
worry
about
it
right
now,
because
I
think
that
was
phase
yes,
six
now
we
pushed
phase,
we
added
a
phase,
five
so
yeah.
So
that
was
don't
worry
about
that
yet
and
then
okay,
so
let's
see
psyched,
we
need
to
still
we
probably
want
to
check
with
himacha,
and
I
don't
think
I
need
let's
see
what's
going
on
here.
B
B
Into
all
right
anything,
so
anything
else,
you
want
to
talk
about
right
now,.
B
So
just
yeah
just
keep
me
posted
on
anything
and
I'm
trying
to.
I
think
I've.
I
C
Okay,
yeah
there's
something
to
look
at
here:
set
up
tests,
just
youtube's
core.
B
Oh
god,
so
have
you
guys
seen
this
one
before
okay?
So
basically,
what
happens
here
is
sometimes
sometimes
unit
test
decides
that
it
should
call
functions.
So
basically
it
goes
through
and
it
loads
it.
It
reads
all
our
python
files
and
it
tries
to
find
things
that
are
classes
that
are
subclasses
of
unit
test.test
case.
Well,
sometimes
it
screws
up
and
it
starts
instantiating
python
functions
as
if
they
were
classes,
and
so
basically
it
starts
trying
to
call
functions,
and
that's
what's
going
on
here.
Is
it
thinks
accuracy
see?
B
The
unit
test
loader
is
is
is
trying
to
create
a
test
case
by
creating
you
know
by
instantiating
this
object,
but
this
object
is
the
accuracy
function,
probably
from
high
level,
and
it's
saying
well
that
you
can't
do
that
because
that's
not
how
you
call
this
function.
C
D
Because
of
name
space
pollution
right.
D
I
think
it's
somewhere,
you,
you
must
have
done
import
everything
from
a
file.
Oh.
B
B
F
B
Yeah,
okay,
so
issue
with
namespace
solution.
Okay,
so
we'll
fix
this
later
because
this
is
just
like
this
is
just
yeah.
Okay,
so
we'll
fix
this
later.
C
B
Name:
okay,
great!
Thank
you
section,
all
right
cool,
so
I
think
yeah.
So
I
think
your
your
path
forward
here
is
basically
you'll
cherry
pick
some
stuff
and
then.
B
F
B
Was
great
it
was,
it
was
shaw
right.
B
All
right
so
so
yeah,
it's
great
having
you
shaw
and
we'll
we'll
hope
to
see
you
again
soon
and
just
let
us
know
how
things
go
so
all
right!
Well,
thanks
everyone
and
have
a
great
rest
your
week,
and
you
know
you
know
where
to
find
me
all
right
have
a
good
one.