►
From YouTube: Weekly Sync 2020-11-24
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.oex6r7w45b05
C
A
Cool
all
right
cico.
Do
you
wanna
tell
us
what
you've
been
you've
been
up
to
or
anything
you
wanna
talk
about.
B
A
Let's
actually
one
second
here,
so
let's
actually
go
through
what
everybody
wants
to
talk
about.
First
I've
been
deviating.
This
is
so
the
format
that
I'm
trying
to
stick
to
is
that
we
go
through
what
everybody
wants
to
talk
about
first
and
then
we
go
back
and
talk
about
everything.
So,
let's,
let's
is
there
anything
else
that
you
want
to
talk
about.
A
C
Not
much
really,
I
saw
that.
I
saw
that
code
that
you'd
sent
me
for.
I
think
it
was
gradient.
Boosting
and
I've
been
trying
to
train
multiple
features
using
that
and
it's
been
going
well
so
far
and
according
to
a
bit
of
trouble
with
the
accuracy
score,
but
other
than
that.
I
think
it's
different.
A
A
Sudhanshu
in
the
lobby
trying
to
join
this
meeting,
but
I
don't
know
what
happened:
weird:
okay,
so
waiting
or
working
on
so,
and
only
detection
model
with
multiple
features
right
and
yeah.
So,
let's
see
we
may
end
up
just
using
yeah.
We
need
we
need
to.
We
need
to
have
an
example
on
how
to
use
multiple
features
there.
Okay,
so.
C
It's
it's
honestly,
fine
like
it
works
so
before
you
send
me
that
I
tried
drinking
it
using
a
single
feature
and
it
works
for
one
feature.
So
I'm
assuming
generalizing
it
to
multiple
features
shouldn't
be
that
hard.
A
Yeah
I
mean
the
features
is
just
gonna,
give
you
an
array,
basically
so
yeah.
If
it
would
just
be
nice,
you
know
if
we
had
if
we
had
that
as
a
part
of
the
tutorial
I
was
trying
to,
I
haven't
had
a
chance
to
go
through
and
maybe
do
like.
You
know
we
have
the
simple
linear
regression.
I
was
thinking
you
know.
A
I
wonder
if
we
can
do
one
of
those
like
a
multivariate
linear
regression
from
from
without
any
dependencies,
because
that
keeps
it
sort
of
you
know
nice
and
clean
actually,
but
you
know
we
could.
We
could
just
add
some
dependencies
to
it
and
and
do
it
with,
I
don't
know,
numpy
or
psychic
or
something
and
then
show
how
to
do
dependencies.
A
But
I
think
we
have
a
little
section
on
that
too.
So
I
don't
know
it's
something
to
think
about
in
the
future,
all
right,
so
anyways
we're
working
through
anomaly
detection
model
with
multiple
features
and
we'll
put
that
you
saw
x
g
boost
code,
so
things,
okay,
we'll
just
we'll
just
wait.
Okay,
so
all
right
so
and
I'll
cover
my
stuff
last
here,
but
so,
let's
see
so
you
seeker.
A
You
said
you
found
some
stuff
going
through
the
tutorial
and
did
you
want
to
sort
of
show
us
what
you
were
show
us
what's
going
on
here
or
you
want
to
talk
about
some
some.
You
want
to
get
down
some
bullet
points
of
what
you
found
first
or
you
want
to
just
show
us.
B
A
Oh
okay,
so
yeah,
okay,
that
makes
sense
yeah
I
was
looking
at
you
know.
I
was
just
looking
at
tensorflow
because
I
looked
at
your
your
code
and
I
saw
that
tensorflow
has
a
nice
high
level
interface
through
keras
to
the
same
type
of
thing,
and
I
was,
I
was
sort
of
thinking
the
same
same
thing.
So
yeah,
so
name
really
is
not
not
the
best
thing
for
that.
So
let's,
let's
make
a
note
actually
here
I'll
make
my
notes
off
screen.
A
Now
I
close
that
meeting
that
stuck
all
right,
okay,
yeah
here
I'll
make
a
note
of
that.
You
won't.
You
won't
see
me
making
notes,
but
I'm
taking
notes
here
so
got
another
another
another
computer
to
take
notes
with,
let's
see
so
found
some
issues
in
this
tutorial.
Okay,
come
on
all
right,
so
for
model
yaml
files,
we
should
have
the
name
property,
be
something
like
layer
type.
B
About
the
input
channel
of
the
second
layer,
you
have
to
match
the
a
fourth
channel
of
the
first
layer
or
something
nice
third
or
the
second.
A
B
A
So
so
in
layer
or
in
channels
of
next
layer-
or
let's
say
I
just
put
so
out
channel
so
sorry
out,
channels
of
previous
layer
should
be
used
as
in
channels.
B
A
Okay,
all
right,
yeah,
that's,
let's
see
yeah
it's
too
bad.
We
don't
have
sock
shop.
Okay,
let's
see
so
we
need
this
one.
Okay,
let
me-
and
let
me
just
look
at
this:
okay,
so
forward
block
one,
I'm
gonna!
U
pooling
count
three
view
negative
one
and
then
in
features
for
linear,
okay,
yeah,
so,
okay,
interesting
all
right,
so
we
need.
A
A
All
okay,
so
okay
wait
a
minute,
so
it's
basically
saying
so
pooling
all
right
so
yeah.
So
we
have
lru
or
relu.
Then
we
have
pooling.
Then
we
have
linear
so
we're
doing
comp
one.
Our
pooling
conf2
are
pulling
conf
to
our
pooling
com3
our
pulling.
And
then
we
do
view
is
view.
Can
you
scroll
up
to
view?
Is
that
defined
there.
B
A
A
My
guess
is:
I
haven't
looked
at
this
in
a
while,
but
my
guess
is
that
the
negative
one
is
just
whatever
is
coming
in
there
and
then
one
one
or
1296
is
it's
the
out
so
that
it
can
be
the
input
of
linear
but
yeah
I
mean
we
need
that
needs
to
be
explained
so
and
what
is
view,
and
why
is
it
just
all
of
a
sudden
appearing
out
of
nowhere?
Okay,
cool
yeah?
Let's
keep
going
then.
A
Can
also
be
created
by
indenting
the
layers
under
a
key
okay,
sequential
layers
can
be
created
by
indenting
layers
under
a
key.
Oh,
oh
yeah,
yeah.
I
remember
this.
I
think
he
needs
a
separate
tutorial
on
this
because
I
think
that
it
ends
up
being
like.
Where
is
it?
I
think,
there's
an
example
somewhere:
oh
no
go
on.
A
So,
okay,
I'm
going
to
find
an
example
and
I'll
send
it
to
you
in
a
link
in
the
meeting
all
right.
Let's
see,
I
think
it's
under
the
test
cases
so
model
and
then
pytorch
and
then
where's
tess
my
touch
net.
A
Maybe
there
is
not
a
test
case
for
this.
I
swear.
I
saw
that
syntax
somewhere
because
I
think
I
think
this
comes
down
to.
We
need
a
separate
tutorial
showing
that,
because
you
know
it's
basically
he's
basically
saying
that
yeah,
I'm
not
saying
this.
Where
the
hell
did
it
go
yeah,
that's
not
good
yeah.
We
need
that.
A
A
For
example,
if
you
looked
at
like
pooling
or
linear
and
if
you
indented
them
one
and
then
so,
if
you
made,
if
you
made
something
right
so
at
the
same
indentation
level
as
they
are
right
yeah,
so
if
you
indent
them
one
and
then
put
them
under
something
called
pooling,
underscore
linear,
then
you're
creating
a
sequential
layer
with
those
two
inside
it.
I
believe.
A
That's
that's
what
he's
saying,
but
this
needs
a
whole
tutorial
because
that's
sort
of
I
remember
him
telling
me
about
this,
but
I
think
that
maybe
oh
maybe
this
is
the
tutorial
he's
working
on
actually
so
that
might
be
good.
So
this
is
good
feedback
for
suck
shop.
So
all
right
all
right,
so
we
need
a
full
example
around
creating.
A
A
Okay,
cool
cool.
All
right,
that's
great!
Thank
you
for
for,
for
you
know,
making
note
of
this
feedback,
because
this
is
important,
so
you
haven't,
you
haven't
gotten
a
chance
to
go
through
it
or
to
finish
it
yet
right!
Yes,
let's
see!
What's
what's
the
current
error
here?
Oh,
I
think.
Maybe
we
just
missed.
Oh
yeah,
you
can't
put
a
you,
can't
put
the
pound
sign
in
the
middle
of
a
bash
command.
You
can't
comment
like
that.
This
is.
This.
A
Has
gotten
me
many
many
many
times
yeah
you
can
and-
and
I
think,
if
you
just,
I
would
just
delete
that
line.
If
you
don't
have
a
gpu
on
here,
you
can
always
add
it
in
later
or
if
you
want
to
do
a
comment,
you
can
move
it
above
the
command
like
so
yeah.
If
I,
if
I
need
to
comment
in
a
multi-line
command
like
that,
I
usually
just
take
the
line
and
put
it
as
a
comment
above
there.
B
A
A
So
that
looks
correct
to
me:
oh
wait
a
minute,
I'm
oh
yeah!
Maybe
it's
not
quite
correct
and
true
no
century
loss.
Yet
that's.
B
A
Is
probably
gonna
fix
it
for
us,
then?
So
we
need
okay
now.
The
next
thing,
though,
is
that
we
that
was
not
a
helpful
error
message,
so
need
helpful
error
messages
when
loss
function
is
misspelled,
yeah,
okay,
great
all,
right,
yeah,
let's
see
what
happens
here.
A
A
A
A
So
while
this
is
well,
this
is
going
I'll
just
go
over
briefly
some
of
the
stuff
that
I
was
going
to
talk
about
here,
just
because
my
my
sort
of
updates
for
everyone
are
quick.
Basically,
let's.
A
A
So
yeah,
those
are
the
notes
that
I
took
all
right.
So
the
we
have
one
of
one
of
the
models
is
dow
for
pi,
so
model
dell
for
pi.
So
I
talked
so
I
talked
to
the
people
who
do
the
the
release
of
it.
A
A
They
are
hoping
to
have
a
5
5
release
out
by
the
end
of
the
year.
This
should
be
the
last
package
we
last
dependency.
A
A
So
I've
been
trying
to
figure
out
how
to
make
this
happen.
You
should
see
it.
I
was
hoping
to
have
it
right
as
we
right
as
we
started
today,
but
I
think
I
have
a
couple
things
left,
which
is
basically
just
what
was
it
now.
I
forgot
one
of
them,
which
is
not
good
good
log.
A
Okay,
yeah,
I
was
fixing
the
documentation
just
checking
for
the
oh.
I
need
to
make
sure
so
some
so
windows
and
mac
os
are
currently
oh
okay.
So
I
so
here's.
A
Interesting
so
for
for
people
to
know
is
that
so
I've
created
a
new
branch
and
basically
this
is
going
to
be
the
any
if
we
do
like
a
4.1
or
whatever
it'll
go
on
this
branch,
and
one
of
the
things
that
I
did
is
I
set
up
pinning
of
the
dependencies
because
we
found
that
people
would
upgrade
things
because
we
had
the
we
had.
These
version
ranges
on
here
and
it
would
result
in
you
know
as
we
develop
right.
A
We
want
to
make
sure
that
the
latest
version
of
all
these
packages
is
getting
installed
and,
and
that
way
you
know
we're,
always
you
know
our.
We
find
issues
as
they
come
up
if
people
upgrade
you
know
if
people
push
new
versions
of
things
and
our
code
breaks,
because
we
were
doing
certain
things
with
the
old
the
way
the
old
version
worked.
We
want
to
know
about
that
right
away
as
we're
pushing
a
master
branch,
but
then
what
we
found
is
that
you
know
with
the
latest
release
version.
What
hap?
A
What
will
happen
is
oops?
A
Oh,
here's
who
should
honcho
hey
sudanship.
So
what
we
found
is
that,
with
the
latest
release
version,
what
will
happen?
Is
you
know
we
have
these
version
ranges
on
here
and
then
you
know.
Maybe
somebody
one
of
these
packages
updates
and
now
things
are
no
longer
compatible,
and
then
you
know
everything
blows
up.
So
what
I
did
is
I
went
and
I
made
it
so
that
they
are
all
pinned
to
a
specific
version.
A
So
basically
everybody,
if
you
install
the
the
release
version
of
dffml
you're,
guaranteed
that
all
these
packages
have
been
tested
together
at
these
versions.
Now
that
that
may
cause
a
few
issues
with
you
know
when
people
want
to
update
different
packages,
but
they
can
always
use
you
know
a
virtual
environment
or
something,
but
this
should
this
should
result
in
in
in
a
working
environment
with
versions
and
packages
that
all
work
together.
A
So
yeah,
oh
wait.
I
wasn't
sharing
my
screen,
damn
it
an
error
occurred
while
screen
sharing.
Are
you
guys
seeing
my
screen?
A
D
A
To
happen
on
on
this
branch,
which
is
the
we'll
do
branches
for
releases,
okay,
all
right,
so,
let's
go
back.
Does
that
did
that
model
finish
training.
A
Okay,
cool
is
there,
is
it
is
it?
Do
you
want
to
keep
going
through
that,
or
or
does
it
seem,
do
you
think
you'll
you'll
go
through
it
offline.
B
A
Okay,
yeah,
and
so
this
is
part
of
what
happened
here
is
basically,
I
think
I
think
and
yeah,
let's
see,
where
did
that
go.
A
Where
did
that?
Go
where's
the
pin
depths,
okay,
so
yeah.
So
I
think
what
happened
here
is
that
tensorflow
yeah
it
got
pinned
to
2.2.1,
because
I
think
I'm
not
sure
we
need
to
actually
that's
a
good
thing
to
check
on.
We
should
check
on.
I
should
go
check
and
see
if
this
is
I'm
still
an
issue,
because
it
would
be
best
if
we
could
be
doing
a
release
with
two
point.
A
Whatever
the
latest
version
of
tensorflow
is
so
good
good,
good
check,
okay,
so
broken,
and
I
think
these
are
broken
because
of
my
dependency,
pinning,
probably
because
of
dependency
depending
on
mac
os.
A
A
Okay
and
let's
see,
was
there
anything
else
that
we
should
look
at
any
or
always
last
thing
I
had
one
more
thing.
I
was
gonna
note
any
yeah.
Anybody
else
got
anything
that
we
should
look
at
before
this
release.
Candidate
goes
out,
because
I
mean
this
isn't
gonna,
be
the
final
version
of
this
release
and
we'll
do
more
releases
afterwards.
A
But
you
know
what
we'd
like
we'd
like
this
one
to
work:
right
we'd
like
this
to
work
too
so
yeah,
if
you,
if
you
think
of
anything,
just
just
put
it
in
getter
or
if
anybody
has
anything
now.
Let
me
know.
A
All
right
so
yeah,
I
think
these
are
the
main
main
things
and
then
what
was
the
last
thing?
There's
one
other
thing,
but
it's
not
really.
It
doesn't
really
matter
I'll.
Let
you
guys
all
know
when
this
comes
out
and
then
well.
We
can
see
if
it's
working
correctly,
so,
okay,
all
right
cool,
so
we'll
go
we'll
go
through
rest
of
tutorial
offline.
A
A
A
So
yeah
so
ping
team
on
getter
and
we
can
help
and
if
you
need
some
one-on-one,
and
this
goes
for
everybody,
you
know
whenever,
if
you
ever
need
some
one-on-one
time
with
with
me
or
with
anybody
else,
you
can
always
just
reach
out
to
us
and
ask
you
know,
I'm
obviously
like
the
the
primary
maintainer
here
so
I've
like,
I
can
probably
answer
the
most
breadth
of
questions,
but
you
know
for
certain
things
like
if
people
have
written
like,
for
example,
saksham
has
written
this
neural
network
tutorial
and
there's
there's
various.
A
If
you
find
something
that
you
know.
If
you
reach
out
to
me
about
something-
and
I
I
may
refer
you
to
the
person
who
actually
wrote
the
tutorial
or
you
know
who
did
whatever
you're
you're
looking
at
right,
but
you
know
you
can
always
reach
out
to
me
and
you
can
always
reach
out
to
the
gator
channel
and
usually
somebody
will
direct
you
in
the
right,
the
right
direction
of
whoever
knows
most
about
what
you're
asking
about,
and
you
can
always
set
up
a
meeting
with
somebody
right
to
do
some
one-on-one
time.
A
If
you
need
extra
extra
working
through
on
something
all
right,
so
all
right
so
sudhan
I
saw
you
push
some
more
actually
stuff.
How
is
that
going.
D
Yeah,
so
I
needed
help
with
the
the
what
was
it.
I
forgot.
D
Yeah
because
I
I
don't
know
like
what's
what's
wrong
with
that
and
also
down
for
five,
because
in
the
dial
for
pi,
it
is
not
showing
any
error
in
the
logs.
A
A
Is
like
this
is
a
mess
yeah,
okay,
okay,
so
I'll
make
note
of
that.
So
doll
for
pi
alpha
pi
should
be
a
cherry
pick
from
master.
A
Hopefully,
and
then
so
qa
model
john
will
take
a
look
okay,
so
let
me-
and
let
me
just
sort
of
look
at
it
now
since
we're
here
and
we
got
time
all
right.
So
let
me
pull
this
down.
D
And
yes,
some
11
version
is
there
and
because
of
that,
like
some
of
the
tests
in
my
branch
are
actually
failing,
okay,
but
in
the
main
branch,
it's
it's
working,
fine,
okay,.
A
Yeah,
I
think
I
went
and
updated
okay,
so
that's
another
cherry
pick
so
auto
sk
learn
is
failing,
need
to
cherry
pick
from
master
all
right.
A
Okay,
so
cute
take
a
look.
Okay,
I
have
this
open
three
times,
apparently
all
right.
Okay,
what's
going
on.
A
Here
all
right,
so
was
this
the
transformers
one.
D
A
Oh-
and
I
think
we
also
so
yash
and
I
recently
worked
through
something
so
you
know
we
should
probably
rebase
this
whole
situation
pretty
soon
here,
because
there's
been
a
lot
of
work
done
on
master
that
yeah,
we
should
rebase
it
after
we're
done
with
facebook.
A
Yeah
yeah,
I
think-
and
we've
had
a
a
few
things
here-
that
that
have
like
we
fixed
the
caching
on
this
model,
see
it
downloads
multiple
times,
and
we
fixed
that
recently.
So
we'll
need
to
need
to
be
sure
to
rebase
after
phase
five
okay
yeah,
because
we're
cherry-picking
lots
of
things
now.
A
Yeah,
we're
gonna
have
a
lot
of
merge
conflicts.
Yeah
I
mean,
I
hope
we
don't,
because
we
haven't
had
a
ton
of
activity
on
a
lot
of
the
models
other
than
you
know
a
few
fixes
here
and
there.
So
I
hope,
it'll
be
okay.
You
know,
there's
been
no
major
refactors
going
on
to
to
a
lot
of
this
code,
so
you
know.
Hopefully
you
should
have
minimal
conflicts.
D
A
A
A
D
Is
because
the
predict
is
actually
some
different
name.
A
A
A
A
A
A
A
For
yes
reason,
I'm
not
seeing
that
commit
pre-processed
data.
A
A
A
Is
that
are
you?
Are
you
good
on
that
then?
Do
we
need
to
dig
more
into
this,
or
I
mean
I
could
oh,
no.
I
think.
D
I'll,
let
you
know
when
I'll
need
your
help.
Okay,
I
think
you're
good.
A
Okay,
yeah,
I
think
yeah,
probably
just
put
it
in
a
score.
Okay,
sweet
all
right!
Well,
that's
that's
glad.
We
figured
that
out
all
right,
so
it
turned.
It
happened
to
be
that
we
never
removed
the
accuracy
method
from
the
qa
model.
A
I'm
going
to
be
there,
we
never
remove
the
actual
method
from
the
qa
model.
We
just
need
to
make
it
to
a
score
all
right,
cool,
okay,
so
and
then
auto,
esc,
learn
and
dial
for
pi
are
cherry
picks,
let's
see
so
let
me
see
if
we
can
just
fix
those
real
quick.
A
A
Okay,
I
think
it's
this
guy
yeah.
I
basically
did
I
used.
I
figured
out.
We
can
use
that
make
make
numpy
config
on
this,
because
it
was
another
case
where
they
had
changed
the
they
had
changed.
The
the.
A
A
A
A
But
it
then,
I
think
something
else
got
updated.
Somehow,
oh
great,
all
right,
this
is
gonna,
be
this
is
the
thing
all
right,
good
status.
A
I
paid
existing
packages
container
okay,
yeah
at
some
point.
The
dockerfile
got
updated
all
right.
Now
we
gotta
do
yeah.
So
this
is
why
we
got
a
rebase
right.
A
A
A
Okay,
I
guess
we'll
do
that.
First,.
A
A
A
A
Okay,
oh
wait
on
the
master
branch.
A
A
A
A
So,
just
for
anybody,
that's
not
very
familiar
with
what
I'm
doing
right
now.
Basically,
we
can
take
by
by
keeping
commits
nice
and
small.
We
can
we
can.
We
can
take
them
and
we
can
apply
them
on
on
a
different
branch.
So
sudhanshu's
branch
split
off
from
everything
else
quite
a
while
ago,
and
and
so
we've
we
we
haven't
done
because
he's
doing
it
a
lot
of
a
lot
of
work
that
touches
a
lot
of
places.
A
We
have
a
different
branch
that
he's
working
on
and
so
every
once
in
a
while,
we'll
rebase
in
the
changes
from
master,
which
is
basically
it's
kind
of
like
a
merge
if
you're
more
familiar
with
the
merge,
but
so
merge
kind
of
shoves
things
in
in
a
weird
order,
and
then
it
creates
a
merge
commit,
whereas
a
rebase
will
go
and
it'll
replay
your
work
on
top
of
another
branch.
So
basically
it's
like
you
any
commits
that
you
did
get
added.
A
On
top
of
you
know
so
that
you
maintain
a
nice
linear
history,
okay,
so,
okay,
let's
see
models
which
just
don't
support,
windows,
america,
so
dfmo!
A
Okay,
you
know-
I
don't
know
if
we
really
care
about
this
this
one,
because
this
is
the
one
which
is
now
supported
on
3.8
and
we
don't
really
care
about
testing
support
for
3.8
within
this
branch.
So
get
cherry
pick
abort.
A
A
So,
let's
just
check
here
so
so
what
did
we
just?
Do
we
added
these
okay,
so
container
with
all
plugins,
if
not
install,
run
container
run
container?
Okay,.
A
And
more
container
stuff?
Okay,
so
we
applied
this,
make
numpy
config
one
for
auto
sk
learn!
So
here's
your
last
commit
where
we
fixed
you
fixed
the
string
matching
okay
and
then
this
is
the
auto
sk
learn
commit
and
then
this
one
is
the
first
docker
commit
that
we
did
so
then
we
got
that
by
doing
this,
to
look
at
the
changes
to
the
docker
file
here,
so
that
was
there
yeah.
Let's
not
stat,
that's
hard
to
read.
A
So
basically
we
went-
and
we
said,
okay
show
me
the
commits
on
master
branch
that
are
related
to
so
then
this
separator
says:
okay
now
these
are
just
files
right
so
show
me
any
commits
that
changed
the
docker
file,
and
so
we
went-
and
we
said
okay-
this
is
the
last
one
that
we
had
on
this
branch.
So
we
needed
this
one
and
this
one-
and
this
one
at
least
to
keep
going
to
get
to
the
point
where
we
could
start
applying
stuff.
So
we
grabbed
those.
A
So
this
is
that
first
one
8
2
or
oh,
it
creates
a
new
yeah.
The
new
the
sha
will
change
because
we're
applying
it
to
a
different
branch,
and
then
this
is
the
next
one
list
all
models
yeah
and
then
we
should
have
the
update
so
update
to
2020.3.
A
So
this
is
the
one
that
and
then
this
is
the
version
fix
that
actually
fixes
the
problem.
Now.
A
The
problem
is
that
now
that
another
problem
is
that
when
you
go
when
you
upgrade
to
this
version,
there's
another
issue,
which
is
this
issue
that
comes
up-
and
this
is
the
fix
for
the
issue
that
arises
when
you
upgrade
to
this
version,
so
we've
got
a
workaround
to
that
fix,
and
hopefully
this
should
be
good
now
so
I'll
push
this
so
get
log
online
okay,
so
we
added
these
five
commits
here,
all
right
so
and
then,
hopefully
that
fixes
so,
okay
so
needed.
A
A
All
right
the
following
because
of
docker
file
changes
we
needed
some
of
those
commits
too
all
right
and,
of
course,
it's
not
guaranteed
that
you
can
just
pull,
commits
off
things
and
everything
will
work
fine.
So
we
will
we'll
see
what
the
ci
tells
us
after
we
apply
these
commits,
I
I
you
know
usually
when
things
apply
cleanly.
A
That
means
that
you
know
everything
was
added
in
a
reasonable
order,
but
you
know
you
can
just
pull
random
changes
to
you
know
you
give
a
commit
like
change,
for
example,
a
test
file
right
and
you
just
pulled
the
changes
to
the
test
file
and
you
know
there
may
have
been
some
commits
that
changed
the
main
model
that
the
file
the
test
was
testing.
Then
obviously,
you're
now
going
to
end
up
with
a
test
that
doesn't
pass
right.
A
So
you
got
to
be
careful
when
you
cherry
pick
commits
it's
not
a
guarantee
that
everything's
going
to
work
you're
just
pulling
changes
to
certain
files,
certain
changes
to
certain
files.
A
A
A
Then
issue
with
commuter
qa
issue,
with
model
method
still
around.
Okay,
great
all
right,
nice
work,
everybody
we
will
talk
next
week
and
I've
got
a.
We
got
a
holiday
in
the
states
coming
up
thanksgiving
on
thursday,
so
I
may
or
probably
won't
be
around
on
thursday,
but
other
than
that
I
should
be
should
be
around.
So
you
always
ping
me
on
twitter.
If
you
need
anything
all
right
thanks,
everybody
have
a
good
day
or
night,
depending
on
where
you
are
bye.