►
From YouTube: Weekly Sync 2021-03-16
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.czotw51mhbql
Great meeting all new people! Thanks for joining!
A
All
right,
so
this
is
our
regular
weekly
sync
meeting
for
all
those
who
are
new.
I
assume
you're,
probably
here
because
of
gsoc,
is
anybody
randomly
happen
to
find
us
at
the
same
time
as
gsoc,
but
not
because
of
gsoc?
A
I
would
be
surprised,
okay,
so
the
way
that
we
run
this
is-
and
I
don't
know
if
you
guys
have
you
might
have
watched
the
recordings,
but
so
there's-
and
let
me
just
fill
everybody
in
here,
but
so
if
you
go
to
the
website
the
main
documentation
website,
we
have
this
contact
page,
and
you
know
this
is
this
is
all
the
ways
you
can
contact
us
which
this
needs
to
be
removed.
A
Now
I'm
realizing-
and
we
have
recordings
posted
of
all
of
these
meetings
here
on
youtube,
and
so
you
can
go
through
and
you
can
see
you
know
all
of
all
of
every
week
in
case
you
in
case
you
missed
something
or
in
case
you.
You
know
something
right
and
then
this
is
the
meeting
minute
stock,
which
is
this
guy
so
and
then
every
and
this
is
so
time
zones
change.
A
If
you
can
add
this,
google
calendar
to
your
your
to
your
to
your
calendar
and
then
and
then
you
will
be-
you
know
slightly
yeah
you
will.
We
will
maybe
hit
the
time
zone
change.
Hopefully
we
can
get
rid
of
that
stupid
time
zone
change,
but
that's
that's
not
within
my
scope.
So
all
right!
So
what
we
usually
do
is
we
go
through
everyone's
is:
did
yeah?
Okay,
so
we
usually
go
through
everyone's.
A
What
what
everybody
needs
to
work
on
or
what
everybody
is
working
on
right
and
and
what
problems
you're
stuck
on
and
then
so
we'll
list
those
out
and
then
we'll
dive
into
them
depending
on
you
know
what
time
wise?
What's
going
to
take
the
longest
so
something's
going
to
take
a
really
long
time,
we
usually
try
to
put
it
at
the
end,
so
this
is
sort
of
a
drop
in
drop
out
situation.
A
You
know
if
you,
if
you
end
up
at
the
into
the
meeting
here
like
if,
if
we
end
up
with
somebody,
who's
got
something
that's
going
to
take
an
hour,
then
we're
going
to
try
to
push
that
towards
the
end
of
the
meeting.
And
then
you
know
everybody
everybody
who
doesn't
have
to
listen
to
that
can
leave
right.
So
don't
feel
pressure
to
stay
if,
if
you're
getting
towards
the
end-
and
nothing
is
relevant
to
you
anymore,
okay,
but
it
is
obviously
a
lot
of
people
stay
for
the
whole
time.
A
If
you're
curious
about
what's
going
on,
that's
that's
fine
too.
Just
you
know,
don't
feel
obligated
all
right,
so
I
want
to
go
around
and
do
some
introductions,
because
we've
got
a
lot
of
new
people
here.
So
I
will
start
with
myself.
I'm
john.
I
work
for
intel
and
so
I'm
the
lead
maintainer
on
this
project,
and
so
basically
I
you
know,
I
run
these
meetings.
A
I
do
code
review
my
work
on
the
project
a
little
bit
and
you
know
we're
we're
driving
things
forward.
So
the
the
project
itself,
as
you
guys,
are
all
probably
familiar
with
you-
know,
we're
a
machine
learning
focus
project.
You
know,
there's
also
this
data
flow
concept
in
there.
That's
really
about
you,
know,
sort
of
feature
engineering.
It's
got
sort
of
broader
applications
as
well.
A
The
main
strengths
of
the
project
lie
in
the
the
machine
learning
aspects
and-
and
you
know
one
of
one
of
those-
that's
that's
really
going
to
round
us
out
here
is
this-
I
don't
think
sudanshi
is
here,
but
but
the
accuracy
part
of
it,
and
so
once
we
have
that
accuracy
part
we'll
be
really
really
well
well-rounded
out
there,
and
then
we,
you
know
we
need
to
think
about
things
like
auto
ml
and-
and
you
know,
our
data
set
clean
up,
and
so
that's
sort
of
you
know
our
picture.
A
Looking
forward
on
on
the
on
the
on
the
ml
side
of
things
also
josh-
and
I
have
been
talking
recently
about
doing.
I
can't
like
cache
data
sets
so
in
providing
sources
that
do
those
and
and
there's
issues
that
describe
sort
of
all
the
things
that
we're
planning
on
doing
and
and
there's
also
within
the
documentation.
A
There
was
a
a
news
announcement
here
of
of
the
4.0
release
or
0.4
release,
and
this
covers
the
you
know
the
stuff
that
we've
had
in
here
and
then
these
are
sort
of
the
the
main
things
that
were
we're
looking
to
hit
going
forward
before
we
hit
beta,
which
is
our
next,
our
next
big
milestone
here,
and
so,
if
you're
wondering
you
know
what
what
are
the
plans
for
the
project?
This
is
what
they
are
as
well.
A
As
you
know
the
various
issues,
but
this
is
sort
of
really
what
we're
trying
to
hit
here
these
these
things
all
right.
So,
let's
go
around
and
we
can
start
with
shaw
since
we'll
introduce
everybody
who's
already
here
to
everybody
who
you
know
maybe
knew
and
shaw.
You
are
to
the
to
the
top
right
on
the
screen
right
now
that
I'm
looking
at
so
you
get
to
go
first.
B
All
right,
hey
everyone,
I'm
shaurya,
most
people
call
me
shaw:
I'm
a
computer
science
sophomore.
I
live
in
new
delhi
and
just
trying
to
do
my
best
and
contribute
to
this
project.
A
B
So
so
far,
I've
been
working
on
adding
the
pandas
data
frame
source
and
in
the
future
I
think
I'll
be
working
on
implementing
video
support
for
dfml
and
hopefully
adding
a
few
models
like
yolo,
I
think,
and
some
other
stuff,
hopefully
yeah
cool.
C
Yeah
hi,
I'm
ciuko,
I'm
a
current
student
from
vietnam.
I
think
I
haven't
seen
enough
time,
so
I
have
been
wondering
in
open
source
community
to
contribute
to
it,
so
you
have
any
with
one
of
myself
there.
A
Cool
yeah
we're
good,
it's
good
to
have
you
back
all
right,
let's
see
so
we're
gonna
go
and
I
see
well.
My
screen
configuration
here
is
different,
but
let's
just
keep
going
down
to
the
right.
A
So
is
it
sahil?
D
Yes,
it's
sahil,
I'm
from
india,
I'm
currently
pursuing
dual
degree
in
iit
from
indoor
in
india
and
I'm
looking
forward
to
contribute
to
the
organization,
and
I
have
like
picked
up
a
couple
of
things,
and
I
would
be
really
grateful
if
you
can
like
show
me
more
about
like
what
I
should
discover
because
dfml
I
checked
the
source
code
on.
It
is
really
a
lot
of
things
to
wrap
my
head
around.
So
it
would
be
really
great
if
I
get
some
guidance.
A
Cool
yeah,
that
is
that's
the
purpose
of
this
meeting.
We
we
go
through
and
and
we
we
try
to
you
know,
address
things
that
are
our
blockers
and
things.
We
want
to
understand
more
about
cool.
A
A
Next,
we
have,
is
it
sanjapon.
E
E
So
yeah
hello,
I'm
sanjivan
singh
gupta,
I'm
also
a
computer
engineering
undergrad
from
india.
I
mean
recently
started
contributing
to
dffml
and
I've
been
looking
into
some
issues
and
trying
to
contribute
to
them,
and
I'm
currently
like
studying
and
learning
about
more
about
the
data
flow
architecture
and
trying
to
add
some
more
models
and
stuffs.
Yeah
cool.
F
Hi
everyone-
this
is
harsham,
I'm
a
senior
year
undergrad
student
from
pakistan,
I'm
generally
interested
in
machine
learning,
so
contributing
to
dffml
has
been
quite
good
for
me,
so
I'm
currently
working
on
separating
the
confidences
from
predictions
on
that
issue.
So
yeah,
that's
all
about
me.
G
Yeah
hi
everyone
anita
out
of
this
side,
I'm
I'm
from
india
and
I'm
finally,
student
of
msc
computer
science
and
I
have
comp
contributed
a
couple
of
models
to
a
dfml
and
also
fix
the
issues
and
right
now,
I'm
working
on
a
issue
about
the
http
service
right.
A
And
sudhashu,
I
I
had
already
briefly
mentioned
your
your
accuracy
work,
but
you
want
to
give
yourself
a
little
introduction
to
everyone.
H
So
hello,
hi
everyone,
my
name
is
sudhanshu.
I'm
a
final
year
undergraduate
computer
engineering
student
in
mumbai,
yep.
I
Oh
hi,
everyone-
I
am
pratik,
I
am
from
I'm
from
jadapur
university
kolkata.
I
recently
started
contributing
in
dffml.
I
A
A
And
then
sahil
right.
A
All
right,
great,
all
right
cool,
let's
see
so
you
know
okay,
so
the
first
thing
that
we
do
meeting
wise
is
you
know,
like
I
said
this
meeting's
open
to
everybody.
We
want
everybody
to
join.
If
you
got
any
questions,
you
know
jump
in
here
and
and
we'll
throw
them
on
the
agenda
right.
So
we
go
through
put
everybody's
name,
and
you
know
what
you
want
to
talk
about.
So
we
are
so
shaw
already
said
today.
A
You
know
he
would
like
to
talk
about
the
the
the
data
data
frame
stuff.
Do
we
have
any
previous
notes
on
that.
A
Okay,
yeah
yeah
this
okay,
the
data
frame,
stuff
issue,
nine
foot;
two;
okay,
great
so,
and
let's.
A
A
B
Yeah,
I
want
to
talk
a
little
bit
about
video
support
and
how
we
might
implement
that.
Okay,
okay,
great.
A
All
right
great
and
let's
see
so
coco
you're
next
on
my
screen
here.
So
what
what
did
you
want
to
talk
about
today?
Did
you
did
you
have
anything
you
wanted
to
talk
about.
A
Getting
back
in
the
swing
of
things,
good,
good
stuff,
thanks
for
thanks
for
being
here,
all
right!
So
and
then,
let's
see
so
so
sahil
do!
Do
you
have
anything
you
had
questions
on
today's
yet
or
you
just
just
wanna.
D
I
I
wanted
to
know
like
she
was
working
on
orion
model
before
this,
and
I
I
I'd
have
to
watch
that
last
week.
But
if
there's
something
like,
I
have
picked
up
that
issue
and
if
there's
something
like
he
could
tell
me
what
he
was
trying
to
just
share
through
those
links.
So
it
would
save
time
and
get
me
up
and
running
on
that
video.
Okay,
go
faster.
A
Let's
see
and
and
you
know
I'm
gonna
so,
let's
see
I
think
I
haven't-
I
haven't
yet
looked
at
that
one
yeah
and
I'm
guessing.
Let's
see
well
we'll
go
so
I'm
getting
ahead
of
myself
here
again.
Well,
I'm
outlining
this,
but
but
so
yeah
we'll
we'll
go
through
and
we'll
we'll
get
to
that.
That's
probably
a
quick
one.
So.
D
About
that
download
progress,
one,
it
was
oh.
A
Or
you
are
you,
let's
see?
Yes,
you
are
programmer
long
string
of
numbers,
that's
what
that
is.
Okay,
great!
It's
like
okay,
all
right!
So
great!
Okay,
let's
see
so
natash.
What?
What
do
we
got
going
with
you
today.
A
A
Let's
see
these
guys,
I
think
we
just
need
to
rerun
this.
Let's,
let's
rerun
this
and
and
then
like
make
sure,
because
I
I
think
you
know
and
actually
wait
a
minute
do
we
need
to
add
this
to
changelog.
A
I
think
we
need
to
add
this
to
changelog
yeah.
I
think
we
we
had
issues
with
the
ci.
I
just
want
to
get
to
this
now
because
you
know,
obviously,
then
we
can
merge
it
later,
while
the
ci
runs
in
the
meeting
so
yeah.
Let's
just
add
this
to
the
change.
Can
you
go
add
this
to
the
change
log
no
touch,
and
then
we
can
merge
it.
You
know,
as
soon
as
the
ci
goes
great,
we'll
get
that
done
now.
Finally,.
A
Okay,
all
right
great
anything
else
in
the
test
sure
is
that.
A
All
right,
hey
so
himanshu
we
went
around
and
did
we
got
a
lot
of
new
people
here,
so
we
wanted
to
do.
Do
a
little
round
of
introductions.
Could
you
give
a
introduction
of
yourself.
A
C
A
I'm
no,
let's
see
all
right,
okay!
Well,
I'm
not
here
in
himanshu,
I'm
not
sure
if
you,
if
you
just
just
let
us
know
in
the
chat
and
we'll
we'll,
because
I
want
to
get
everybody
we're
trying
to
get
everybody
introduced
all
right.
Let's
see
all
right!
Okay!
So,
let's
see
all
right
looks
like
next.
We've
got
sutonshu,
hello,
okay,
we
got
him
how's
it
going.
J
Hey
I
just
yeah,
sorry,
I
wasn't
getting
the
audio
and
yeah
that's
good.
A
Cool
yeah
can
you
I
see
a
lot
of
people,
yeah
yeah.
We
got
a
lot
of
people
here.
Can
you
give
us
a
little
update
on
on?
You
know
just
introduction
yourself
and
and
what
you're
doing
these
days.
J
Okay
sure
so
guys
I
was
last
summer
yeah,
so
yeah
I've
been
working
with
a
person.
Currently
I'm
working
as
a
software
engineer
at
smartwatch.
So
microsoft
is
a
company
that
works
with
matlab
and
assembling
this
yeah,
so
I'm
working
there
in
python
interface.
J
A
Pretty
much
it
great,
very
cool,
very
cool
yeah,
it's
good
to
have
you
back
all
right,
so
all
right,
okay,
so
suit
honcho!
So
let's
see
did
you.
I
think.
A
All
right,
perfect
yeah,
because
I
just
there
was
a
couple
that
were
there
were
duplicates
all
right.
Okay,
great
thank
you
is
there
anything
else
you
wanted
to
address
today.
A
A
good
one,
all
right,
I'm
excited
about
that.
Oh
my
gosh.
We've
wanted
to
see
that
get
done
forever.
All
right,
I
pretty
commits
an
accurate
score
and
I
think
that
accuracy
stuff
is
getting
pretty
dang
close
here.
We're
going
to
want
to
do
that.
I
saw
somebody
threw
up
a
master
domain
commit
and
we
definitely
want
to
get
accuracy
merged
before
we
do
that.
A
So,
let's
see
all
right.
Okay,
working
on.
A
Working
all
right,
okay,
so,
let's
see
who's
next,
let's
see
well,
who
somebody
speak
up
who's?
Who
wants
to
go
next
cause?
Now
I've
got
all
out
of
order
here.
F
So
I
just
wanted
to
remind
you
that
I'm
done
with
that
pr,
the
one
that
separates
confidence
from
okay.
F
F
A
Yeah
yeah
yeah,
yeah,
okay,
we
can
discuss
yeah.
Let's
we
can
discuss
that.
That
sounds
good.
Let's
see
examples
on
okay,
yeah.
F
Yeah,
because
I
was
thinking
that
after
this
vr,
I
could
start
working
on
that.
A
All
right,
great
yeah,
we'll
we'll
touch
on
that
all
right.
Anything
else
on
your
side.
F
No,
that's
it.
Thank
you
all
right,
cool.
A
A
Yeah,
all
right
sounds
great
all
right.
So
what
what
are
you
so
so
what
we
do
is
yeah,
just
just
what
you're
working
on
or
or
what
you
were
here
were
you
here
last
week,
yeah.
E
Yeah,
the
two
cool
requests
that
I've
made
the
creator,
the
creating
the
hash
functions.
A
All
right,
so,
let's
see
I
I
got,
I
was
too
slow,
so
hash
functions.
I'm.
A
E
Yeah,
so
I
was
also
studying
about
the
data
flow
architecture
and
working
on
for
the
for
another
issue
like
it
was
for
adding
section
to
identify
how
the
data
flow
should
be
executed.
So
I
was
studying
that.
E
A
I
Basically,
I
was
once
working
with
the
pythos
models
right
when
we
are
actually
passing
the
dictionary
object
to
the
pythos
network
class.
It
was
actually
throwing
an
error
and
the
dict
was
not
converted
to
a
neural
network
sequential
model
right
for
that.
I
actually
created
a
pr,
and
I
think
I
need
to
share
my
screen
after
some
time
to
actually
show
what
that
error
is
actually
coming
up.
Is.
J
I
I
A
Okay,
okay
and
I
think,
okay
great,
let's
see.
I
And
also,
and
also
I
created
a
peer
regarding
the
excellent
assets,
all
right
yeah.
Regarding
that,
I
was
actually
facing
an
error
in
the
plug-in
section,
because
I
actually
don't
get
that
error.
I'm
just
saying
attaching
the
link
here.
Yes,
it's
actually
showing
the
new
plugin
is
not
being
identified
or
so.
A
All
right
so
and
right
now,
I'm
going
to
say
you
know
we're
already
at
at
40
minutes
past
the
hour.
So
I'm
going
to
say
there's
some
of
these
that
are
definitely
going
to
go
offline
and
offline,
meaning
we're
not
going
to
do
them
in
this
meeting.
So,
let's
see
and
not
not
just
yours
every
once
so
I
will.
I
will
let
everyone
know
what
those
are.
So
anything
else
critique.
I
Most
and
another
thing
in
the
in
the
recent
issue
means
we
were
where
we
were
supposed
to
means
validate
the
commit
messages
yeah.
I
Actually,
in
the
docs
section
we
are
having
come,
we
are
having
those
formats
which
actually
specified
the
means
files
which
we
are
changing
right
now,
if
suppose,
the
user
is
having
10
comments,
then
having
the
same
type
of
commitment
may
be
means
may
be
ineffective
in
telling
means.
Actually
this
is
the
syntax
of
the
commit
message
that
is
source,
csv
and
parser
due
to
the
separator.
A
Yeah,
so
so
so
the
the
rationale
behind
this
is
is,
and
so
so
sub
module
files
for
description.
So
so
the
idea
and-
and
so
this
is
like
ci
lint
commits
so
and
the
way
that
this
is
structured.
I
realize
we
haven't
really
touched
on
this
enough,
but
the
the
way
that
the
commit
message
and
issue
formatting
works
is
meant
to
be
sort
of
like
your
your
guide,
a
guide
to
the
directory
tree
here
right.
A
A
A
Here
and
then,
and
then
you'll
run
one
of
these
guys,
it's
like
you
know
does
so
the
idea
would
be,
you
know,
add
a
new
one.
This,
the
the
way
that
this
commit
this
issue
title
is
formatted
implies
that
you
should
add
a
new
one.
Here
called
you
know,
run
commits,
and
that
would
you
know,
link
the
commit
messages.
A
Yeah
you
know
that
might
be
a
good
idea.
So
let's
see
well,
let's
see,
let's
see
we're
getting
derailed
here
I
got
I
got
derailed.
So
let's
just
see,
let
me
put
this
down
and
then
make
sure
that
we've
got
everything
and
then
we'll
we'll
come
back
to
this.
So
let's
see
commit
message:
formatting
ci
job.
A
All
right,
let's
see
so,
let's
make
sure,
did
we
get
everybody.
Let's
see,
I
think
my
list
has
gotten
collapsed
here.
Did
we
get
everybody
all
right?
I
think
we
got.
Everybody
did
speak
up
if,
if
I,
if
I,
if
we
didn't
get
you
down,
so
let's
see
all
right
so
then
this
is
what
happens,
copy
paste
and
we
start
going
through
them
all
right,
and
this
will
be
this.
This
gets
faster.
Obviously,
when
we,
when
we
aren't
doing
introductions.
F
Excuse
me,
john
yeah,
do
you
mind
if
I
get
back
in
a
few
minutes?
Something
came
up.
A
Oh
yeah
no
worries
so,
let's
just
say
I
think,
keep
on
looking
for
confidence,
use
cases.
Examples
we
might
want
to
just
talk
about
this
offline,
so
you
and
I
can
sync
on
gitter
or
something
later
all
right.
Cool
all
right
have
a
good
one.
Hashim.
F
A
F
He
just
wants
to
know
about
the
video
that
we
discussed
this
issue
and.
A
Okay,
like
which
video
is
that
you
mean
like
go
yeah.
F
Yeah,
I
actually
gave
the
link,
but
he
probably
doesn't
know
what
he's
looking
for.
A
F
Cool
yeah,
it's
on
the
it's
on
the
issue,
the
link
to
the
video,
okay,
great
and
yeah.
A
Cool,
so
so
he'll
do
you
have
any
any?
Did
you
look
through
the
video
or
do
you
have
further
questions
or.
F
A
A
Right,
great
cool
all
right,
well,
yeah,
we'll
let
you
go
hashem,
have
a
good
one!.
A
A
Right
and
then
this
is
offline.
So
a
lot
of
these
pr
reviews
become
offline
topics
because
well
you
know
they
they
take.
They
take
a
while.
So
unless
there's
any
specific
things
that
that
you
know
so,
we
usually
do.
We
usually
take
pr
reviews
offline
unless
there's
stuff
that
we've
been
reviewing
several
times
and
and
we're
we're
having
trouble.
You
know
communicating
in
a
text-based
format
right
then
we
start
we
start
doing
those
those
in
in
the
meeting.
A
So
let's
go
through
and
see
what's
going
on
here,
so
I
think
you
know.
The
first
thing
I
want
to
cover
here
is:
is
this
let's
just
see
if
anything,
so
that
light
gbm
model?
Did
we
rerun
the
ci
on
that
touch?.
I
A
Okay,
did
it
or
did
you
push
that,
let's
see
didn't
you
were
going
to
push
the
change
log
right.
A
A
Okay,
oh
and
it
looks
like
this
is
still
yeah.
We
need
to
fix
this
too,
all
right.
Okay,
I'm
glad
we
didn't
merge
that,
so
this
is
the
the
plug-ins
test.
So
what
is
this.
G
I
think
I
think
there
was
an
issue
in
s2o
model.
That's
where
the
light
gbm
is
still
failing.
Something
like
that.
A
Let's
see,
I
think
I
think
we
now
are
to
the
point
where
they're
separate
enough.
Let's
see,
I
think
it
was.
You
know
we
had
some
issues
with
the
the
pin
depth
stuff.
So,
let's
see
what's
going
on
here
yeah,
I
think
this
might
be
just
a
rerun
case.
Let's
fight
her,
let's
figure
this
out,
so
this
oh
no!
This
is
the
same
thing.
So
what
is
this
all
plugins
appear
in
dfml
plugins
make
sure
that
any
setup
to
appear
in
okay?
A
A
So
get
grep,
if
you
guys
don't
know
about
get
grep.
This
is
this.
Is
your
new
best
friend?
This
is
great,
and
I
can't
I
can't
tell
you
how
great
gig
grep
is.
Okay,
so
make
sure
that
all
setup,
py
files
associated
with
the
plugin
appear
in
dfmo
plugins.py.
So
I
believe
that
is
the
issue
with
both
those
pr's
at
the
moment
is
that
they
both
need
to
show
up
here.
A
So
we
just
need
to
add
them
to
this
list
in
both
those
pr's
and
then
add
to
the
change
log
as
well,
and
then
I
think
I
think
we
will
be
good
to
go
other
than
that
yeah
great,
so,
oh
and
then
also
okay.
So
another
thing
here
is
that
if
these
fail,
then
maybe
the
the
last
place
that
you
have
to
add
them
is
you
have
to
make
sure
that
they're
added
in
github
workflows,
testing,
and
so
they
need
to
be.
A
It
will
start
to
fail
the
other
jobs
if
they
are
not
in
this
list
as
well.
So
you
add
it
to
the
plugins
list,
and
so
so
try
running
that
c.
Try
running
that
test
locally,
that
had
a
set
of
tests
that
test
dot,
test,
ci
and
locally,
and
and
that
that
will
hopefully
you
know,
bring
us
to
bring
us
to
a
good
point
here.
So.
I
A
A
A
Unit
test
v
tests,
test.test
ci
and
to
validate.
A
All
right
and
let's
see
what
else
do
we
need
to
decide
whether
we're
taking
offline
or
not
so
pr
for
cash
progress,
download.
A
We
got
a
lot
of
pr's
right
now,
so
cash
progress
where'd,
you
go,
it
said
yeah
there
we
go.
A
Oh
yeah,
okay.
I
remember
I
just
found
this
okay
great.
A
Did
you,
let's
see
a
couple
of
cis
surveillance,
yeah
yeah,
whatever
lgtm?
Okay,
this
one
is
just
like
there's,
there's
lgtm
is
just
a
mess
right
now
we
need
to
go
clean
it
all
up
at
some
point.
I
can't
figure
out
how
to
turn
it
off.
So,
unfortunately,
it
keeps
failing
jobs.
A
A
It
is
because,
because
yeah
so
heel,
I
know
I
know
you
you,
you
ran
this
under
conda
and
it
blew
up
right
is
anybody's.
Is
anybody
still
running
under
con
under
anaconda
anymore,
because
I
know
most
people
have
have
moved
to
most
packages
have
moved
to
releasing
on
pie
pie.
So
I'm
wondering
you
know
is:
is
this?
Is
this
anaconda
environment?
Anything
is.
Is
this
something
that
we
need
to
be
running
and
testing
against
and
fixing
right
now.
H
A
A
A
It's
probably
just
one
environment
out
there,
no
all
right
so
so
this
the
issue
that
that
was
ran
into
here
was
the
the
what
was
it
the
yeah?
Oh,
this
is
a
console
test
issue
because
console
test
attempts
to
go,
create
a
new
virtual
environment
to
run
some
of
these
tests
and
that
that
blows
up
so
and
console
test
needs
to
be
split
out
and
refactored
at
some
point,
but
but
for
now
it'll
stay,
but
so,
let's
see
yeah
I'll,
just
rerun
ci
test.
Okay.
A
So,
regarding
condo
support,
I'm
making
changes,
so
I
think
what
we
need
to
do
is
we
really
need
to
that
console
test
plug-in
probably
just
needs
to
get
split
out
of
this
project
and
have
its
own
ci
in
its
own.
It
probably
just
needs
to
become
its
own
thing,
and
this
is
so.
This
is
another
thing
that
everybody
probably
needs
to
know
about.
A
So
with
regards
to
the
fact
that
we
have
a
million
ci
jobs,
so
this
is,
I
mean
this
is
all
well
and
good
and
nice
and
it
allows
us
to
cross
validate
all
the
changes
from
the
main
repo
into
other
repos
or
sorry.
It
allows
us
to
to
cross
validate
all
the
it
allows
us
to
cross-validate
all
the
plug-ins
against
the
main
against
the
main
domain
domain
package
right,
because
we
have
this
sort
of
tiered
architecture
where
we
have
the
main
project
and
then
we've
got
all
the
plugins
right.
A
So
there's
this
concept.
We
want
to
have
this
concept
of
core,
or
you
know
first
first
party
plugins
right
and
those
are
the
things
that
are
maintained
within
the
dffml
source
code
right,
and
so
these
are
things
that
this
is
the
main
package
here
right.
So
this
is
the
main
package,
dfml
dfml
and-
and
this
is
where
you
know
all
our
stuff-
that
when
you
say
pip
install
dfml,
everything
in
here
gets
installed
right.
So
that's
the
main
package
and
now
everything
else
in
here.
A
So
if
you
go
into
like
model
right,
these
are
all
plugins
right
and
right
now,
they're
all
maintained
as
a
part
of
the
the
the
code
base.
That
is
this.
You
know
intel
dffml
and
we're
thinking.
You
know
this
obviously
puts
a
lot
of
code
in
the
code
base,
so
we
may
try
to
go
and
and
and
have
this
approach
where
we
have
you
know
we
want.
We
want
to
also
facilitate
people
hosting
their
own
plugins
right.
You
know
maybe
under
their
own
github
usernames.
A
So
we
want
to
create
this
space
for
second
party
plugins,
which
is
sort
of
plugins
that
are
still
maintained
by
you
know
us
as
the
community
and
we're
going
to
try
to
go
fix,
but
you
know,
aren't
necessarily
maintained
as
a
part
of
the
main
package
right.
So
so,
for
example,
we
split
out
the
model
transformers
because
oh
the
ci
passed
well,
we'll
all
be.
We
pl.
We
split
out
model
transformers
because,
oh
I
know
why
it
passed
because
it
has
a
conflict.
A
We
had
to
upgrade
tensorflow
and
then
we
had
to
upgrade
the
the
the
apis
of
of
spacey
and
model
transformer
and
we
didn't
get
around
the
upgrading
model
transformer
before
the
4.0
release
the
the
0.4
release.
So
we
we
threw
it
in
this.
This
secondary
sort
of
second
party
modules
right
so
with
the
hope
that
we'll
get
it
back
and
as
soon
as
we
have
time
to
upgrade
those
apis
or
update
our
usage
of
their
apis
right
with
their
new
new
version,
and
so
the
idea
is
right.
A
We
test
against
the
various
the
various
versions
of
dfm,
also
we'll
test
against
the
master
branch
and
we'll
test
against
you
know,
whatever
the
latest
release
was,
and
that
way
we
can,
you
know,
maintain
all
these
plugins
in
a
different
space
and
and
so
we'll
move
to
this
architecture
soon.
A
So
just
something
to
be
aware
of
that's
upcoming,
and
why
did
we
start
talking
about
that?
Oh
well,
I
can't
remember
cash
download.
Oh
because
of
of
oh
because
of
console
test
and
splitting
out
console
test,
so
console
test
will
probably
go
there
too.
So
let's
just
put
a
note
that
that
we're
going
to
we'll
be
splitting
out
related
plugins
along
so
we'll
be
splitting
out
plugins
along
with
console
test
test
into
dfml
github
org.
A
A
Okay,
so
the
pr
on
cash
download.
A
And,
let's
see
anything
else,
we
need
to
really
talk
about
here:
double
cis
yeah.
So
don't
worry
about
that
lgtm
we
haven't.
We
need
to
go
and
fix
a
bunch
of
stuff
to
get
that
working.
Okay
and
then
it
looks
like
you
said
this
passed.
Wakanda
support.
Don't
worry
about
that
at
the
moment,
but
yeah
go
go
for
it
if
you
want
to
fix
it
great,
but
you
know
we're
not
gonna
test
for
it
in
the
ci
until
you
know
maybe
a
month
or
two
from
now.
A
So
regarding
tensorflow
model
accuracy
issue,
ci
tests
fail
sometimes
yeah.
That
would
be
great
if
we
could
see
the
test.
Well,
actually
wait.
No
we've
gone
over
this
multiple
times.
We
we
have
decided
not
to
seed
those
tests
because
well
you
know
we
want
to
make
sure
that
that
that,
if
per
chance,
someone
changes
something
it
only
happens
to
work
with.
You
know
the
seed
in
that
way.
Then
then
it
doesn't.
We
aren't
getting
false
false
passes
in
our
ci
right.
A
So
so
so
the
the
risk
we
run
of
of
just
having
to
rerun
the
ci
jobs
is,
is
probably
it's
it's
a
bit
of
an
inconvenience,
but
it
doesn't
happen
that
often
it
definitely
used
to
happen
more
often.
I
thought
I
think
at
one
point
we
added
some
retry
in
there
but
yeah.
We
we
we're
gonna,
leave
that
so
logging
module
implemented
helpful
function.
Okay,
great
so,
let's
see
I'll
do
the
rest
of
this
review
offline.
Then,
unless
do
you
have
anything
you
want
to
say
about
that.
D
A
All
right,
so
you
follow
up
with
questions
as
necessary,
all
right
great.
So
then
so
these
pr's
okay,
we're
probably
waiting
for
the
cia.
I
assume
on
this
to
pass
strip
spaces
csv
source
pr,
so
strip
spaces.
Okay,.
A
A
A
A
A
Let's
just
just-
and
this
is
just
a
minor
thing-
but
let's
not
do
unless,
let's
not,
let's
not
add,
let's
not
add
white
space
where
there
doesn't
need
to
be
white
space
just
because
it
results
in
you
know
it
results
in
more
more
more
more
lines
changed
and
and
when
we
look
through
things
in
in
this
in
this
format
here
you
know
we
want
to
see
the
least
slides
changed,
all
right,
cool
and
that's
you
know
not
so
much
for
unitash,
because
I
think
you
know
you've
probably
heard
that,
but
for
everybody
else,
let's
see
so
all
right
great
so
and
then
I
think
actually,
let's
just
let
me
just
do
that
right
now,
I'll
just
hit
commit
on
this.
K
A
So-
and
this
is
what
I'm
talking
about
with
the
pull
request
or
the
the
naming,
so
we
do,
you
know
sort
of
dig
down
into
the
the
directory
structure
here
and
we
ignore
the
top
level
module
at
the
top
level
modules,
the
main
one,
if
it's
not
the
main
one,
when
we
prefix
with
what
kind
of
plug-in
it
is
so
sort
of
csv,
remove
white
space
changes
all
right.
A
A
All
right,
great
we're
working
through
this,
pretty
quick
all
right,
so
these
pr's
on
the
in-n-out
channels.
A
A
All
right,
okay,
so
I'm
thinking
here
that
this
is
something
that
predict
object,
has
no
attribute
to
okay
and
so
saksham
says
it's
working
cleaning.
The
branch
20
required
changes
might
resolve
it.
So.
I
A
So
I
think
that
you
know,
I
think
that
that
it
might
be
good
to
just
create
a
new
example
for
this.
You
know,
if
you,
if
you
don't
want
to
lose
these
changes,
then
creating
a
new
example
file
would
be
a
good
approach
here
and
this
this
type
of
thing
with
the
you
know
this
this.
This
is
the
type
of
thing
where
sort
of
yeah
yeah.
So
sorry,
one
thing
at
a
time,
I
would
create
a
new
example
file
that
contains
these
changes,
and
that
way,
you
don't
lose
your
work.
A
You
know,
and
and
do
that
as
a
separate
pull
request.
A
A
I
A
More
examples
is
great.
We
want
that,
especially
since
you've
already
done
it.
If
you
could
just
throw
it
in
another
file
and
and
then
make
sure
that
there's
something
that's
testing
it,
then
we
we
definitely
want
that,
but
yeah,
let's
try
to
keep
it.
You
know
we'll
try
to
keep
keep
it
as
it
exists
right
now,
for
the
purposes
of
just
doing
this
change.
I
Right,
okay
and
the
main
function
which
I
created
to
convert
the
dict
into
the
sequential
model
was
that
create
network
function
just
like,
I
think.
I
I
A
All
right
so,
let's
go
into
code,
cub
and
and
just
double
check
what's
going
on
here!
No,
no!
No,
because
this
this
will
give
us
a
for
sure
on
on
whether
this
is
being
called
or
not.
This
is
a
good
good
resource
to
to
go
over
here.
So.
A
Okay,
what
we
were
at
90
when
the
hell
was
that
oh,
oh
april,
2019.,
okay,
let's
go
back!
No,
let's
not
go
back!
Okay,
all
right!
So,
let's
see
polls.
A
A
The
plug-ins,
and
that
will
also
be
something
that's
splitting
out
the
splitting
out
those
those
projects
into
into
their
own
projects,
will
make
that
a
lot
easier
to
measure.
Okay,
all
right.
A
You
know:
okay,
is
there
anything
specific,
because
if
there's
nothing
sort
of
specific
here,
I'm
going
to
let
saksham
continue
being
your
reviewer
on
this?
Is
there
anything
specific
that
you
want
to
call
out
right
now
or.
I
Especially
in
the
sense,
the
previous,
the
actually
there
was
an
issue
regarding
this
fear
means.
One
issue
was
there
in
the
convolutional
models,
the
input
layers
of
a
con
configuration
layer
will
take
the
would
be
same
as
the
output
of
the
previous
convolutional.
I
Where
we
are
actually
missing
out
some
of
the
input
layers
of
the
input
parameters
of
the,
I
think
yes
in
this
in
this
test
cases.
I
If
you
look
in
the
convolutional
layer,
two
of
the
in
channels,
it
is
none,
okay,
yeah,
so
it
it
should
take
the
parameters
from
the
convolutional
output
features
of
the
convolutional
layer
in
the
make
in
the
just
which
is
just
above
it.
A
I
Basically,
the
function
which
I
created
of
create
network
was
basically
handling
all
these
stuffs,
okay
and
also
creating
the
indian
model
from
the
yml
file
like
that.
A
A
I
think
yeah-
okay,
oh
and
that's
what
suck
trump
said
here
is,
I
guess
so,
I'm
not
really
up
to
date
on
this
and
I'm
kind
of
having
trouble
understanding.
What
what
if,
if
there's
anything
that
we
really
need
do?
Is
there
anything
that
we
really
need
to
discuss
here
in
this
meeting,
or
do
you
think
you
can
you
can
continue
this
with
saksham.
I
A
You
know
his
implementation
here
so
yeah,
so
he's
he's
got
a
lot
more.
You
know
he's
already
reviewing
this
and,
and
you
know
I
I
would
have
to
go-
get
up
to
speed
on
on
what
this
review
is.
You
know,
and
I
can't
I.
A
Do
that
right
now
right?
So
so,
let's
just
let
him
take
that
and
then,
if,
if
you
know
you
guys,
if
you
guys
need
some
more,
you
know
if
you
need
some
more
input,
then
then
pull
me
in
okay
but
yeah.
I
don't
think
saksham
is
not
here
today,
so
we'll
just
leave
that
as
an
offline,
so
offline
saksham
will
continue
this
review
so
yeah,
just
just
that
sounds
good
all
right!
Okay!
So
in
this
xls
source
this
this
one
is
definitely
going
offline.
K
A
C
A
Just
say
one
thing
about
this:
first,
so
this
load
load,
fd
and
dump
fd
is
not
the
appropriate
implementation
for
this.
So
because
you
don't
know,
let's
see
we
don't.
Okay,
this
load
workbook
function.
A
A
Oh
well,
okay,
so
so
the
usage
that
you
have
you
know
what
what
you've
done
here,
where
you've
you
know
this,
this
stuff
looks,
you
know
it
it
looks.
I
haven't.
I
haven't
my
cursory
glance
that
it
looks
like
this.
This
makes
sense
right,
because
what
you're
doing
is
you're
trying
to
create
record
classes
where
the
features
are.
You
know.
A
Map
the
columns-
obviously
this
I
mean
it
looks
it
looks
correct,
but
I'm
looking
at
it
for
half
a
second.
So
if
this
passes
yeah
this
I
mean
this
is
what
you're
looking
for
here
right.
So
the
thing
is,
the
main
thing
is:
let's
see,
load
fd.
What
do
we
have
here?
We
have
it
again,
okay,
so
we
got
to
remove
this
file
right
so
because
we
have
a
duplicate
file
here.
A
Let's
see
and
then
this
guy
great
you
put
that
in
there
so
yeah.
I
think
the
main
thing
here
is
you
want
to.
If
you're
going
to
do
this,
if
you're
going
to
do
this
load,
fd
dump
fd
approach,
you
need
to
read
the
data
out
of
the
file
itself
rather
than.
I
I
And
then
we
could
further
call
that
dump
every
kind
of
function
inside
the
xls
class,
so
that
could
also
perform
means.
Do
the
job.
A
Let's
see,
are
you
talking
about
like
dumping
csv
to
to
taking
csv
files
and
dumping
them
to
excel
files?
Is
that
what
you're
talking
about.
I
A
A
Where's
that
merge
command.
No,
no,
I
guess
it
got.
Oh.
We
got
rid
of
the
docs
for
that
at
some
point
because
we
had
to
change
the
syntax,
but
we
do
have
a
command
to
do
that,
so
dfml,
merge
or
yeah.
A
So
this
command
here
will
merge,
two
data
sets,
and-
and
so
basically
you
can,
you
can
take
one,
that's
one,
that's
empty.
You
can
take
the
source
as
the
csv
file
and
the
destination
as
this
empty.
A
Excel
file
and
you'll
take
a
csv
file
and
you
would
convert
it
to
a
excel
file
effectively,
because
you
would
do
load
fd
would
happen
from
the
csv
file
and
dump
fd
would
would
happen
to
the
you
know,
you'd
end
up,
calling
that
sequence,
like
you
just
talked
about
so
the
main
thing
here
is
that
we
need
to
figure
out.
How
do
we
save?
How
do
we?
How
do
we?
You
need
to
figure
out?
How
do
we
use
this
excel
library
to
write
out?
A
You
know
a
bunch
of
bytes
instead
of
instead
of
the
actual
you
know
you
make
making
it
work
with
the
file
the
file
name,
because
these
load,
fd
and
dump
fd
are
supposed
to
work
on
file
descriptors
and
the
file
descriptor.
So
where
is
that
test
test
excellence
records?
A
Okay,
so
I
believe
we
have
a.
What
is
it
called.
A
File
source
test
where's
that
file
search
test
yeah
all
right,
so
these
guys,
so
the
file
source
also
does
some
stuff
where
it
will
support
compressed
files,
and
so,
if
we,
let's
see
test
key
test,
config
set
okay.
Now
this
is
all
csv
specific
csv
source,
okay
from
versus
all
right.
So
this
guy.
A
A
You
know
if,
if
it's
reading
from
one
of
these
compressed
files,
that's
getting
unpacked
or
repacked
by
the
by
the
abstraction
provided
by
the
file
source
and
what
you'll
notice,
when
that,
when
you
apply
this
this
file
source
test
and
we
need
to
get
rid
of
that
test
tag,
but
you'll
notice
that
it'll
blow
up
when
you
have
that
that
fd
dot
name
because
the
name
property
won't
exist
on
these
compressed
versions
of
the
files.
A
Yeah,
I
know,
and
that's
what
I'm
saying
is
that
we
don't
want
to
do
that,
because
we
don't
want
to
end
up
in
the
situation
where,
where
you,
where
you
have
a
compressed
file
and
then
because
when
you
so
when
you
have
this
compressed
file
here,
so
if
you're
subclassing
from
the
file
source
you'll
end
up
in
a
situation
where
this
is
the
open
that
happens
to
open
the
file,
so
it
looks
for
the
the
file
right,
and
this
is
the
allow
where
allow
empty
comes
in
and
then
it
says:
okay
is
this
compressed
file,
then,
if
it
is
compressed,
then
I'm
gonna
do
this.
A
You
know
read
mode
compressed
and
then
you're
gonna
end
up
with
you
know
this
opener
and
the
opener
is
so,
for
example,
yeah
it's
it's
it's.
You
know
gzip
compressed
right.
So
this
when
you
call
gzip.open
on
this
file
name
that
may
not
have
a
dot
name
attribute.
A
So
if
you
have
xls
dot
gz,
then
you're
going
to
end
up
with
opener
you're,
going
to
put
your
effectively
have.
Opener
is
fd
because
you're
doing
with
opener
as
fd
and
then
you're
doing
fd.name,
and
it's
going
to
tell
you
well
opener
has
no
name
attribute,
and
so
for
that
reason
we
really
want
to
make
sure
that
we're
were
were
able
to
read,
read
and
write
bytes.
You
know
bytes
of
that
file.
Does
that
make
sense.
A
A
I
A
Yeah
we
can't
get,
we
aren't
guaranteed
the
presence
of
that
dot,
name
property
on
that
that
file
file
object.
So
we,
what
we'd
like
to
do
is
we
would
like
to
understand
you
know,
is
there
a
way
that
we
can
have
this
workbook
so
workbook.save?
A
Where
is
that
let's
say
workbook?
So
if
we
look
at
the
api
documentation
here
and
we
looked
at
workbook
dot
save
and
it
says:
okay,
so
save
the
workbook
to
a
given
file.
Name
use
this
function
so.
I
I
Means
we
are
not.
Oh,
I
think
the
lord
every
and
dumb
people
are
overriding
from
the
base
classes,
right,
yeah
so
so
means
if
we
are
having
a
csv
file,
so
then
the
overrided
function
will
run.
That
is
in
that
cases
we
will
run
the
csv
classes
fd,
and
I
don't
think
there
will
be
an
issue
in
that
case.
A
Well,
so
I
yeah,
but
we're
only
talking
about
the
excel
source
at
this
time,
right
we're
talking
about
the
excel
source
right
now
and
the
excel
source.
You
know
it,
it
will
so
so
so
there's
a
so
there's
an
easy
okay!
You
you
see
what
I'm
saying
about.
If
you
open
in
a
compressed
file,
it
won't
have
a
name
attribute
right.
A
Yeah,
okay.
So
what
what
and-
and
I
think
part
of
this
is
you
know-
it'd
be
good
if
we,
let's
see
so,
if
we
subclass,
if
you
had
subclass
from
file
source
test,
you
would
have.
You
would
have
seen
this
immediately,
but
I
don't
think,
are
we
we're
not
mentioning
fossil
tests
here?
Yeah,
no,
we're
not
all
right!
So
that's,
which
is
why
you
didn't,
which
is
why
you
didn't
see
that
so
makes
sense.
But
it's
just
something
something
to
to
consider
here,
because
we
ran
into
this.
A
We
ran
into
this
at
some
point
before
and
that's
why
I'm
trying
to
head
it
off
so,
which
is
how
we
ended
up
with
that
test
case.
So,
if
there's
a
work
around
to
this,
which
is
you
can
create
a
temporary
file
so
or
we
can
just
basically
exclude
the
usage
of
x
of
file
extensions.
A
So
let's,
let's
see,
let's
see
what
what's
going
on
here.
So
so,
and
you
see
what
I'm
saying
right
so
you
know
we
can
either
say
you
know,
don't
allow
file
extensions
opening
with
with
this
excel
file
with
the
excel
source
right,
so
we
could
say
we'd.
Maybe.
A
Well,
so
I'm
saying
that
we
need
this
code
here
not
to
run
you
know
unless
unless
so
we
need
some
sort
of
like
allow
list
here
to
say
you
know
compression
allowed
or
something.
A
A
You
know
this
is
the
type
of
thing
where,
where
we'd
say
you
know,
if,
if
you
know
we'd
add
this
here
right
and
we'd
say
you
know:
if
the
suffix
is
this
and
the
suffix
is
in
you
know
in
self
dot
compressed
compression
allowed.
Then
then
we
open
it
right.
Otherwise,
we
just
try
to
open
it
and
it'll
explode
or
I
guess
maybe
we
need
to
do
some
validation
there.
Maybe.
B
A
That's
probably
beyond
the
scope
of
this:
let's
add
you
know
what
we
should
do
right
now
is.
We
should
probably
just
take
this
and
make
an
issue
out
of
it,
because
if
somebody,
if
somebody
blows
up
on
the,
if
somebody
tries
to
do
a
compressed
file
and
it
blows
up,
then
they'll
they'll
end
up
with
an
issue
anyways.
A
So
why
don't
you?
Why
don't
you
go
through?
And
why
don't
you
just
make
an
issue
for
this
and
to
track
this
right,
and
then
you
can
put
it
to
do
here
and
say
you
know
I
I
you
can
put
it
to
do
in
the
code
and
reference
the
issue,
and
that
way
you
know
it's
at
least
documented
that
that
that's
not
gonna
work
right.
A
Okay,
so
let's
see
we'll
create
an
issue
to
either
support
or.
A
For
file
name,
okay,.
A
Reference
issue
in
to
do
comment
within
the
code
above
usage
of
ft.name
anyway.
That
way
that
way,
at
least
we
know
we
know,
we
know
that
things
are
going
to
blow
up
right,
cool
all
right.
So,
let's
see
so
we'll
do
this
offline
with
sorry.
This
is
saksham.
This
is
offline
so
and
then,
let's
see
so
pr
frame,
secure
hash
functions.
A
Okay,
so
all
right!
Unfortunately,
we
have
a
lot
of
people
this
meeting.
So
is
this
an
hour
earlier
or
later
than
everybody
else
is
used.
A
B
G
K
A
Hour
later,
if
that
we
can
start
sort
of,
you
know
what
would
be
30
minutes
ago
going
forward,
if
that,
if
that's
better
for
people.
G
Okay,
for
now
it's
it's
okay,
because
it's
9139
right.
A
Yeah,
just
let
me
just
so
so
let
me
put
that
as
an
open
too.
So
let
me
know
on
the
median
timing
if
this.
If
because
I
can,
I
can
easily
move
this
meeting
later.
I
can't
move
it
earlier
because
I
have
another
meeting
before
it,
but
we
can
move
it
an
hour
later,
if
that,
if
that
fixes,
whatever
time
zone
issue
happened,
there
so
meeting.
A
Or
move
an
hour
later,
and
we
may
go
back
to
our
two
meetings
a
week
situation
if
we
continue
to
have
this
many
people,
because
obviously
this
is
not.
A
A
A
lot
going
on
here,
yeah,
I
don't
want
to
burn
you
guys
out.
I'm
sure
this
is
a
long
time
to
be
in
a
meeting
all
right,
okay.
So,
let's,
let's
get
to
okay,
I'm
looking
at
the
rest
of
these
and
we're
going
to
be
here
for
a
while.
I
think
if
we,
if
we
do
all
of
these,
so
we're
at
the
time
for
this
meeting,
so
let's,
let's
pick
up,
is
there
anything
that
anybody
is
blocked
on
immediately
and
and
needs
to
be
dressed
immediately.
A
Otherwise,
we're
gonna
schedule
a
meeting
for
later
in
the
week
and
and
and
and
we'll
figure
it
out
from
there.
So
is
anybody
have
anything
that
that
they
are
immediately
blocked
on
right
now
that
that
prevents
them
from
from
progressing
within
the
next
few
days.
That's
on
this.
G
A
Yeah
yeah
yeah
no
worries
no
worries
the
hp
service,
okay,
so
and
and
that's
something
that
sue
tancho
and
I
had
talked
about
recently.
We
want
to
try
to
refactor
that
that
that
service
code,
so
all
right.
Okay,
so
let's
see
service
http,
all
right
so
and
and
what
what
so?
Let's
let's
go
to
the
documentation?
A
Okay,
so
and
let's
go
how
to
run
http
service-
and
I
think
somebody
else
was
looking
at
this
too:
okay,
so
the
docs
for
the
http
api,
okay,
yeah,
okay,
so
this
is
the
this
is
the
main
example
here
and
so,
which
are
you
talking
about?
The
python
example
here
is
this.
G
G
Oh,
I
have.
I
have
made
the
changes
accordingly
right,
so
for
me,
it's
it's
running,
but
how
to
run
the
example
or
the
training
part
on
the
http
api.
That's
what
I'm
asking.
A
Is
not
gonna,
this
code
needs
to
have
this
stuff
taken
out
of
it.
I
can't
remember
how
we
got
that
in
there
probably
weird
formatting
thing:
okay,
so,
let's
see
okay.
So
how
do
we?
How
do
we
use
a
model
from
the
http
api?
Okay?
That
was
your
question
you
want
to.
You
want
to
use
a
model.
I
J
K
A
Okay,
so
the
configuration
so
essentially
the
flow
is,
you
know,
yeah,
okay,
actually
it's
listed
in
the
formula
flow
great,
so
the
the
flow
is
configure
a
model,
create
a
context
and
then
you
know
use
the
model
right.
So
the
first
step
here
is
we
configure
mine.
Let
me
just
pull
these
up
side
by
side,
we'll
just
hack
it
hack
this
together.
A
Okay,
I
hope
we're
recording
yeah,
okay,
great
all
right.
So
let's
do
examples.
A
Where's
that
python
example
coming
from
source
dot.
A
A
Oh
that's:
what's
going
on,
we
have
this
giant
code
block.
Okay,
yeah!
No
wonder
this
is
formatted
funkyly!
Okay.
I
think
we
probably
I
can't
remember
what
we
did
all
right.
So,
let's
just
take
this
and
let's
just
write
this.
A
A
So
this
is
this
tool
that
I
use,
if
you
guys,
haven't
seen
this
nodemon.
This
is
great.
It
runs
things
on
file
change.
You
probably
all
most
of
you
probably
heard
me
harp
about
how
great
it
is
a
million
times.
A
So
you
get
this
rapid,
rapid,
rapid,
build
and
test
cycle,
so
we're
going
to
run
the
different
util
console
test,
our
testing
console
test
and
we're
going
to
run
it
on
this
example.
File
set
docs
python
rst,
I
believe,
okay
and
what
files
is
okay,
yeah,
all
right,
so
so
we're
gonna
read
in
this.
This
is.
A
A
Okay,
so
this
is
also
a
short
demo
of
how
to
write
console
test
stuff.
So
this
is
basically
so
with
console
test
you.
Basically
you
put
your
regular
documentation
here
with
code
block
console
and
you
put
the
little
test
option
on
it,
and
so
now
it's
going
to
start
this
and
it's
going
to
start
it.
You
know
as
in
test
mode,
and
so
when
we're
gonna
say
it's
a
damon.
So
damon
has
a
running
background
process
and
it's
running
on
port
8080
all
right.
A
So
then,
let's
see,
then
we
need
this
code.
Okay,
so
and
we
don't
really
need
okay,
so
yeah,
linear
regression
model.
Let
me
indent
all
this
one,
because
I
think
it
was
not
indented
all
right.
So
then
we
can
also
put
a
file
path
to
write
the
stuff
out
too,
and
the
file
path
would
be.
You
know
client.py.
A
So
this
is
our
client
file
and
then-
and
we
usually
try
to
prefix
when
we're
writing
documentation.
Examples
we
try
to
put
in
bold
what
the
file
name.
We're.
Writing
is
so
we'll
say,
first
start
the
server
and
then
we're
going
to
configure
the
model
and
then
we'll
create
the
model
context.
A
So
should
we
configure
the
model?
Actually,
let's
make
these
separate
files.
A
And
we're
just
going
to
use
url
lib
because
that's
built
in
okay,
so
we'll
configure
the
model
and
the
problem
with
this.
Is
we
can't
format
with
black?
Oh
well,
all
right,
so
8080
do
we
have
replacement
functionality
in
here?
Maybe
we
don't
all
right.
We
will
just
run
it.
Okay,.
A
A
So
endpoint
is
a
way
that
you
know
we
might
commonly
refer
to
a
a
server
that
we're
running
against.
So
the
url
or
the
url
of
the
server.
A
So
this
is
the
url
that
the
http
service
is
running
on,
and
so
that
would
be
you
know.
For
example,
you
know
just
the
base
part
so
http
in
this
example.
So
this
is
the
local
url
and
then
8080,
you
know,
but
the
the
thing
that's
going
to
happen
here
is
when
we
run
this
example.
It's
going
to
bind
this
the
service
to
run
on
a
random
port
because
you
know
we
don't
want.
We
want
our
examples
to
always
choose
a
random
port
in
case
that
port
is
already
in
use.
A
So
we'll
just
say:
okay
grab
the
endpoint
from
the
the
argument
there
and
configure
the
model.
So
so,
let's
see
url
okay,
so
the
url
would
be
and
we'll
use
the
this
new
style
of
formatting.
A
It's
f
string,
f,
string,
stuff
is
fun
all
right
so
and
we
can
say
url
lib
request,
dot,
url
open,
and
there
is
the
that's
that
and
I
think
we
actually
need
to
create
a
request
object
here.
Lib
url,
lib.request.request.
A
And
we'll
give
it
the
url
to
go
to
as
the
first
parameter
and
I
think
we
need
headers
and
we
can
make
that
a
dictionary
and
don't
stop
indenting
two
spaces.
So
yeah
we'll
leave
this
open
for
the
docs
pie.
We
want
url
lib.
So
this
is,
I
use
duck
duckduckgo
it
lets
you
do
this
bang.
You
can
do
exclamation
point
pi
and
then
it'll
search
the
python
documentation
for
you
huge
time.
Saver.
A
A
Let's
see
oh
yeah
this
guy,
so
basically-
and
this
is
this-
is
basically
gotten
from
this
configure
documentation.
So
so
we
said
configure.
This
is
how
we're
reading
these
docs
right.
So
we
say:
okay,
here's
the
model
right
that
we're
wanting
to
configure
here
and
okay.
It
looks
like
this
is
taking
it
directly
from
that
example.
So
it
should
be
good,
so
we
will
grab
this
as
the
data,
and
so
this
is.
This
is
directly
from
that
example.
There,
okay.
A
Okay,
and
is
there
another
level
there
needed
or
no
okay,
we're
just
using
multiple
levels
of
indentation,
great,
all
right,
okay,
yeah,
so
we
grab
the
endpoint.
We
make
the
request
with
url
open.
So
if
you
look
at
the
documentation
for
request,
it
takes
this
url
or
it
says
url,
which
can
be
either
a
string
or
a
request,
object.
Okay,
I
guess
we
can
specify
data
there.
So
we
can
just
do
this.
A
Data
equals
stude
model
config
all
right.
Let's
just
make
sure
that
that
is
correct.
There
yeah
and
we
post
that
all
right.
So
we
should
be
able
to
post
this
here
and
we
will
have
configured
a
model.
A
Oh
we're
supposed
to
that's,
why
all
right,
so
the
url
the
data
and
then
we
want
to
just
basically
read
this
response.
So
and
we'll
just
do
this
because
you
know
we
don't
really
need
to
do
anything
with
this
response
here,
we're
just
going
to
read
it
and
display
it
for
our
own
edumacation.
A
A
We'll
just
real
quick,
get
rid
of
anything
that
might
be
laying
around
there
and
reinstall.
A
A
A
A
Model
is
not
confined:
okay,
yeah,
we
all
right
great.
So
it's
running
the
test
now
so
you
see
over
here,
it's
we
said
you
know
run
run
the
console
test
plugin
against
this
restructured
text
file.
So
it
started
that
daemon
and
it
started
it
on
this
random
port.
Here
and
then
you
know
we
wrote
out
this
file
this
configure.py
and
this
is
the
contents
and
then
you
know
we
ran
the
file
and
I
forgot
to
pass
the
url,
so
we're
gonna
do
a
let's
see.
A
Where
is
that
you
know
we
want
to
pass?
Let's
see.
A
And
this
2.127.0.01
is,
is
the
local
address
and
we're
going
to
go
ahead
and
we're
actually
going
to
do
a.
J
K
A
So
I
need
to
we
need
to
go
ahead
and
replace
the
random
port.
That's
been
allocated,
so
we
say,
okay
on
the
list
of
commands
that
I'm
about
to
run
replace
the
last
argument,
which
is
negative,
one
of
the
first
commands
and
replace
the
port
that
says
or
replace
the
string
8080
with
the
with
the
port
of
the
running
server,
which
is
you
know,
the
server
that
we
started
up
here.
All
of
this
is
in
that
console
test
documentation
and
it's
in
the
rest
of
the
documentation
so
that
that
will
actually
run.
A
Point
negative
three,
and
so
then
the
model
type
would
be.
Let's
just
do
slr,
because
that
one's
built
in
and
then
we'll
say
you
know
my.
A
A
On
hash
will
type
dict,
oh
yeah,
we
have
to
encode
this,
so
we
need
to
do
a
json.dumps
and
then
we
gotta
do.
You
know
we
gotta
encode,
that
so
in
code,
because
it
it
takes
a
raw
raw
bytes.
A
403
forbidden,
oh,
this
is
because
of
my
stupid
machine
here.
For
some
reason
they
have
a
okay.
Let
me
put
it
as
localhost
for
some
reason
on
this
machine.
I
don't
know
what's
going
on
it
doesn't
like
it
when
you
do
that,
404
not
found
all
right
compose,
configure
okay,
it
doesn't
like
the
double
slash.
I
think.
A
So
then
the
next
thing
that
we
want
to
do
is
we
can
add
logging
to
this.
So
we
can
see
you
know,
I'm
just
going
to
go
through
the
configure
one
because
we're
you
know
we're
always
over
time
here.
So
there
there's
the
log
okay,
so
logging
is
on
and
it
said,
400
bad
request.
A
Okay,
and
here
we
get
a
message
from
the
server
that
says
missing,
predict
oh
yeah,
and
so
we
are.
A
Bad
status
line
yeah,
that's
an
interesting
one,
bad
status
line.
What
how
are
you
making
a
request?
A
I
guess:
well,
you
can
change
it
to
make
it
to
make
the
request
like
this,
because
that's
that's
yeah,
that's
not
good!
Oh
and
this
hasn't
been
updated
or
it
has
so.
Let's
see
my
model,
okay,
so
predict,
and
we
want
you
to
predict
okay,
this
is
slr,
so
this
should
just
be
y.
Let's
just
have
this
be
x
and
y.
A
Bad
status
line
means
there's
something
wrong
with
your
http
client
bad
request.
A
Ssl
later,
okay,
yeah
and
that
well
that's
because
you're
requesting
that's
because
you
have
mismatched,
insecure
and
https,
so
you'll
notice.
This
is
http
right.
You
need
to
be
doing
http
if
you
started
the
server
with
insecure
and
if
you
didn't
start
the
in
server
with
insecure,
then
you
need
to
make
sure
that
you
well,
you
won't
be
able
to
start
it
because
it'll
give
you
a
bunch
of
errors,
but
if
you
started
it
with
insecure,
you
read
the
security
docs.
If
you
want
to
start
it
securely,
you
start
it
within
secure.
A
Then
you
need
to
make
sure
that
you're
making
an
http
request
rather
than
https
request,
because
that's
gonna,
what's
gonna,
be
giving
you
the
bad
status
line,
because
it's
sending
a
bunch
of
garbled
encrypted
stuff.
And
it's
then
it's
that
you
know
it
doesn't
read
that.
As
you
know,
you
know
a
post
whatever
right,
which
is
what
it
should
be
so,
and
this
is
getting
mad
about
what
again
and
what's
the
predict.
A
Yeah,
okay,
this
stuff
is
well
okay.
Where
is
so?
We
can
also
reference
the
tests
here.
So,
let's
reference
the
test,
because
let
me
see
what's
going
on
all
right
so
test
model,
and
this
is
configure
all
right.
So
when
we're
doing
this
here
we
say:
okay,
I
guess
we're
just
posting
that
okay,
json
equals
config
model
predict
model
directory
models,
features.
A
Oh,
it's
because
this
is
not
formatted
in
the
yeah
we
need
to
just
so
it
has
to
be
formatted
in
this,
and
this
is
that
unified
config
thing
that
we're
always
talking
about
is
this
format.
We've
got
this
config
format
that
allows
us
to
that.
Allows
us
to
internal
server
of
wonderful
that
that
allows
us
to
configure
things
on.
The
object.
Has
no
hp
account?
A
Okay,
so
it's
not
doing
that
properly.
So
it's
not
okay,
yeah,
that's
not
good
either!
So
well,
we
found
a
bug.
So
this
is
the
unique,
unified
config
stuff,
like
the
main
problems
that
we're
having
with
this
is
that
this
we
need
to.
We
need
to.
We
need
to
go
through
and
and
make
it
so
that
that
all
of
this
stuff
gets
gets
configured
the
same
way
right,
and
so
we
have.
A
Then
it
ends
up
pulling
from
this
config
and
then
it
pulls
the
information
from
the
prop
the
plug-in
property
of
like
you
know
what
to
put
there
and
that's
directly
because
of
it's
directly
because
of
the
fact
that
you
know,
when
we
have
this
hyphen
model
hyphen
directory,
you
might
have
hyphen
model
right
and
then
you
would
say
you
know
what
the
model
is
and
then
you
need
to
be
able
to
go
in
underneath
that
right
and
if
you
just
had
model
and
what
the
model
was,
then
you
can't
obviously
use
that
as
like
a
sub
key.
A
So
we've
got
that.
That's
why
that
structure
exists.
Why
this
thing
is
not
being
cooperative?
I
don't,
I
don't
know,
that's
another,
that's
a
mystery
to
me,
but
we
need
to.
We
need
to
hunt
down
this
bug
too,
because
I
don't
know
what's
going
on
here,
that
the
tests
are
obviously
passing
so
there's
something
else
happening.
That's
weird!
It's
trying
to
instantiate.
A
A
You
know
we
need
to
go
figure
out
this
unified
config,
stuff
and
saksham,
and
I
have
been
working
on
that
so
that
that'll
that'll
get
done
eventually,
but
you
know
until
then
we
gotta
work
around
some
bugs,
but
so
yeah,
that's
how
you
configure
and
then
you
know
it'll,
be
it's
pretty
much
a
similar
process
from
there.
A
A
A
Okay,
so
you
might
want
to
reference
the
javascript
example
because
that'll
have
that'll.
Have
you
know
like
how
you
you
know
that
that
that
has
examples
on
that?
I
think
the
requests
are
correct
for
that
one,
so
yeah,
and
then
I
think
this
is
just
let's
say
so
my
model
and
then
my.
A
Context
and
find
label
so.
A
And
this
will
create
the
context
and
then
the
other
thing
is
that
you
need
to
figure
out
how
to
do
the
instantiate,
the
source
and
stuff.
The
problem
is,
the
thing
is
like
we
need
to
do
a
refactor
of
this.
I
guess,
if
you're
getting
into
this,
you
could
do
the
refactor.
That
would
be
helpful,
but
we
needed
sutonshu
stuff
to
go
in
first,
because
there's
just
too
many
changes
that
happen.
Otherwise,
at
the
same
time
resulting
in
giant,
merge
conflicts,
so
it's
probably
best
to
do
one
and
then
the
other.
A
So,
but
if
you
wanted
to
get
in
here
and
get
familiar
with
that,
then
that
would
be
good.
Okay,
now
we've
created
the
context.
So,
okay,
great
I
mean
it
got
a
temporary
directory.
So
you
know,
and
then
then
the
final
step
here
is,
you
know,
use
the
model
right.
A
So
where
is
model
context
train
okay
and
then
you
pass
it
the
the
data
set
itself,
and
so
this
is
where
it
gets
annoying,
because
you
have
to
actually
go
instantiate
a
data
set
and
that's
the
stuff
that
that
javascript
example
might
be
also
helpful.
For
so
you
can
copy
the
syntax
there.
So,
let's
see
train
the
model.
My
training
data
say
yeah,
okay,
and
this
didn't.
A
This
didn't
actually
include
the
the
data
set
itself,
so
you
got
to
go,
do
the
same
thing
for
the
source,
and
so
we
instantiated
a
model,
and
then
we
created
a
context,
and
now
you
have
to
go,
create
a
source
and
create
a
source
context
before
you
can
do
train
and
where
you
post.
This
is
the
request
body
where
you
reference
the
source
label,
and
this
is
all
what
needs
to
change,
because
obviously
you
should
just
be
able
to
sort
of
post
the
whole
config
all
in
one
and
do
it
there.
A
But
you
know
it
wasn't,
wasn't
what
we
did
when
we
did
it
so
because
we're
trying
to
we
also
want
to
provide
direct
access.
You
know
the
idea
behind
all
of
this
with
the
python
library
and
the
command
line,
library
and
the
http
services
to
provide
the
exact
same
interfaces,
so
you
could
do
the
same
thing
across
them
now
that
doesn't
mean
that
we
don't
need
to
have
more
helper
methods
right,
so
we
don't
have
to
do
so
many
requests,
and
so
that's
where
the
refactoring
will
come
into
play
all
right.
A
So
I'm
gonna
post
this
code
here
and
I
will
post
you
know
what
I
use
to
run
it,
and
this
is
sort
of
you
know
something
that
that
we'd.
This
is
yeah
a
little
little
console
test.
Example
here.
A
A
But
but
we'll
we'll
play
by
ear-
and
let
me
know
so,
let
me
know
if
thursday
is
a
good
a
good
day
for
you
all.
If
not,
then
we
can
do
friday,
but
you
know
I
hesitate
to
to
schedule
friday
and
friday
meetings,
because
that
would
be
friday
evening.
Your
guys
time,
I
believe
I
know
that's
what
we
had
last
year,
but
also
you
know
nobody's
really.
No
none
of
us
are
doing
anything
right
now.
A
I
would
assume
there's
you
know
still
coronavirus
going
on
where
everyone
else
is.
So
you
know
I
I
understand.
If
no
one
cares
about
friday
meeting,
I
think
that's
actually
what
happened
last
year
too.
That's
how
we
ended
up
with
a
friday
meeting,
but
you
know,
let
me
know
whether
you
want
thursday
or
friday
or
another
day.
You
know
I
want
to
avoid
doing
it
on
the
weekends
but
yeah
cool,
so
I'm
going
to
post
this
and
then
I
we
will
call
it
a
day
thanks.
C
Just
contact
donkey.