►
From YouTube: Weekly Sync 2020-12-08
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.wtsh98t1iotr
A
All
right,
so
we're
talking
about
doing
sweep
is,
as
a
visualization
plug-in
so
you're
going
to
probably
want
to
look
in
the
the
command
line
stuff
for
listing
records.
Python
dffml,
dffml
cli
list.
A
So
here
is
the
list
records
command,
yeah,
okay,
so
it
goes
through
and
it
does.
It
yields
everything
unless
it
yields
all
the
records.
Unless
you,
let's
see
resources
and
sources,
yeah
yields
all
the
records,
unless
you
say
pretty
print
them.
So
you
may
want
to
wrap
this.
Yes
yeah.
I
think
you
can.
You
may
even
be
able
to
like
subclass
from
this,
and
that
way
you
would
just
make
the
the
run
loop
of
this
guy
be
something
like
you
know.
A
Async
four
run
in
super
dar
run
and
you'll
need
to
make
sure
that
you
set
pretty
to
false
because
you're
not
going
to
have
pretty
and
then
you
now
you
have
or
let's
see,
xsync
4
record.
A
Now
you
have
all
the
records
right
here.
Just
like
the
user
would
be
doing
the
list
command.
So
you
probably
want
to
build
an
array
or
a
panda's
data
frame
with
those
and
then
and
then
pass
it
to
sweep.
Is
all
right
so.
A
A
C
Yeah,
that's
that's
cool,
I
think,
as
you
have
given
me
a
right
path
that
how
can
I
start
so?
I
think
let
me
start
first
of
all
and
then,
if,
if
I
got
any
kind
of
issue
then
I
will
definitely
yeah.
C
A
Sweet
sounds
good
cool,
all
right,
sudanshi
how's
it
going.
A
I
saw
you
added
a
bunch
more
actually
scores.
B
A
Okay,
oh
one,
second,
actually
real,
quick.
So
for
the
the
fact
that
we
have
regression
of
exergy
boost,
but
no
classifier
can
we
add
an
issue.
A
So
can
you
add
an
issue
to
track
this
and
and
how
I
just
want
to
make
sure
that
I'm
pronouncing
your
name
right?
How?
How
did
you
pronounce
your
name.
A
Okay,
yeah,
could
you
add
an
issue
here
so
that
we
can
track
that?
That
would
be
great.
A
So
all
right,
yeah,
I
haven't-
had
to
take
the
chance.
A
A
B
Yeah
so
yeah,
so
because
of
these
two
imports,
I
was
actually
getting
that
error.
So
now.
Actually
the
problem
is
like.
I
have
added
the
scorer
for
ner
qa
and
classification.
D
B
The
transformers-
but
there
is
also
like
some
issue,
like
the
module,
is
not
getting
called
like
I'm
getting
some
error
with
that.
A
A
Okay,
great
yeah,
nice,
very
nice,
all
right,
you
want
to
present
and
then
oh,
I
didn't
remove
imports
causing
circuit.
I
was
like,
where
did
that?
Circular
dependency
happen:
okay,
accuracy,
accuracy;
okay,
so
these
were
causing
a
circular
dependency.
So.
A
A
So
that
when
we
do
go
to
implement
them
all
they
still
work.
So
what
did
you?
What?
What
tests
were
you
running
to
find
that
out
like
what
what
tests
blew
up
before
you
did
this,
so
we
can
run
them
again
and
make
sure
that
that
fixed
it.
B
A
Okay,
all
right!
Well,
let's
all
right
to
model
transformers.
A
B
Yeah,
so
we
might
not
run
it
because,
like
there
are
also
some
issues
with
that.
Also.
A
Okay,
all
right
all
right,
so
I'm
just
gonna
put
this.
I'm
just
gonna
make
a
note
of
the
fact
that,
because
I
don't
want
to
take
up
too
much
of
our
time
right
now
but
yeah,
I
think
that
we
can
replace
this
commit.
A
We
can
remove
this
commit
and
do
you
know
how
to
do
that?
It
is
just
in
case
you
haven't
seen
this
before,
but
if
you
do
get
rebase
dash,
I
and
then
you
can.
You
do
head
minus
30.
D
Okay
should
be
able.
A
To
replace
it
with
removal
of
mine
from
diff
mouse
model
slash
okay,
so
we
should
be
able
to
replace
it
with
the
removal
of
that
line.
A
Hopefully
I
mean,
I
believe,
that
that
looked
like
the
circular
dependency
to
me,
so
hopefully
that
gets
it
which
makes
sense
because
we
used
to
need
that
all
right.
Okay,
so
sorry,
what
was
the
next
thing.
B
So
the
next
thing
was
actually
in
the
transformers
model.
D
A
B
I'm
not
running
that
right
now,
okay,
but
so
actually
we
can
look
into
into
the
logs
actually.
A
B
So
I
was
like
my
first
guess:
was
that
maybe
like
we
are
taking
all
that
and
converting
it
into
an
accuracy
model,
that's
why
it
is
not
working
and
in
the
setup.pi
I
have
just
done
accuracy,
colon
and
then
I've
added
the
class
in
the
transformer
setup.I.
Maybe
I
think
that
was.
A
This
all
looks
correct.
Let's
see,
how
do
we
test
this.
A
A
Accuracy
something
about
loading,
this
file
is
making
it
mad.
Now
from
dffml.util
import
entry
point
so
yeah,
it's
dffml.util.entry
point
import
entry
point
I
think,
if
you
actually
just
do
this
is
one
of
those
where
you
might
might
benefit
from
this
type
of
thing
and
where
we
can
just
import
them.
A
A
A
A
A
A
B
A
A
Thank
you
for
doing
the
commits
like
this.
This
is
making
it
so
much
easier
to
review
and.
A
A
D
A
Good
sweet,
hey
nice
progress
on
this,
we
were
just
actually
talking
about.
We
were
just
actually
talking
about
how
xg
boost
could
benefit
from
different
kinds
of
accuracy,
and
I
was
saying
well
you
know:
sutanji
is
going
to
he's
going
to
expose
that
whole
framework
for
doing
that.
So
not
only
will
new
accuracy
scorers
be
able
to
apply
to
xg
boost,
but
they'd
be
able
to
apply
to
everything
if
we
wrote
them
so
very
exciting.
B
E
B
B
Like
need
to
do
that
in
this
case
or.
B
About
spacey,
like
I
was
thinking,
maybe
I
can
first
like
push
that
thing
into
master,
but.
A
So,
let's
see
so
I
mean
let's
just
double
check
on:
let's
see,
okay,
so
let
me
just
capture
this.
A
B
A
B
In
the
spacey
model
it
actually
does
pc
model
actually
has
its
own
spacey
dot.
Scorer
thing:
oh.
A
B
Yes,
so
we
actually
need
to
wrap
the
whole
spacey
dot,
scorer
module
and
then
like.
We
can
use
that
particular
score.
Oh.
A
B
A
I
see
I
see
what
you're
saying
yeah
so,
but
it
was
using
one
specifically
before
right.
A
Yeah
yeah,
let's
see
and
we've
we've
taken
it
out
since.
A
A
Score
score:
okay,
yeah!
Let's
score
what
was
score
I
must
have.
We
must
still
be
in
there,
so
we're
going
to
hurt
something:
that's
scored,
no
just
import
score.
A
A
Let's
see
score
scored
at
score:
oh
okay,
gold
bars.
A
Yeah
this
is
a
bit
interesting
because
it's
like
well,
I'm
not
saying
I'm
not
familiar
with
this,
but
you
know
what
what
was
he
doing
with
gold
parts
and
stuff
and
is
that
gonna
need
to
change?
I
mean
who
knows
that
might
be
like
an
argument
to
the
accuracy
score
yeah.
Unfortunately,
I
haven't
messed
with
this,
so
I
don't
know
yeah
I
mean
I
think
at
this
point.
If
it's
just,
I
would
just
replicate
this,
because
this,
I
don't
think
is
I
mean
this
is
a
pretty
the
overhead
on
on
you.
B
A
B
A
Because
you
know,
when
you
go
to
do
that,
rebase,
just
it's
quite
quite
often
that
that
that
you
yeah
you
want.
F
A
Everything
going
in
as
clean
as
possible
because
that
can
get
really
messy,
really
fast
and
and
really
confusing
about.
What's
breaking
so
I
think
it's
probably
worth
it
just
to
to
put
in
the
same
body
here,
and
we
can
put
a
to
do
on
that
to
look
at
it
further.
B
So
I
will
have
to
just
create
these
two
scores:
yeah.
A
I
mean
I
would
just
create,
I
would
just
say
you
know,
spacey
score
right
and
then
do
this,
and
then
we
can
find
you
know
the
more
specific
or
more
generic
role.
We
can
just
find
the
more
specific
name
for
it
later
or
tweak
it
later
right,
but
just
make
one
called
spacey
score
or
something
and
then
you
know
do
do.
A
This
was
right.
We
don't
need
to
work,
worry
about
too
much.
Yes
right.
So
just
as
long
as
we're
replicating
what
we
had
right,
then
we
could
change
more
later.
Okay,
so
let's
just
replicate
the
function.
A
We
had
yeah,
and
then
I
mean
that'll
that
yeah
the
the
the
cleaner
testing
environment.
We
can
have
going
into
that
rebase
that
the
happier
we're
going
to
be
yeah,
I'm
wondering
what
the
hell's
going
on
here.
This
is
this
other
stupid
unit,
testing.
A
A
Yeah
yeah
we're,
I
think
what
we
also
want
to
do
is
we
may
want
to
just
switch
the
because
you
know
what
I
I
just
saw
is
that
where
was
this
when
I
ran
the
tests,
yeah
like
this
with
unit
test
b,
it
didn't
it
didn't
seem
to
blow
up.
In
that
way,
I
mean
who
knows?
Maybe
it
did.
There
were
a
bunch
of
things
that
were
yeah.
A
Okay,
none
of
those
errors
look
like
related
to
that
which
we're
gonna
have
to
change
it
to
say,
run
it
that
way
in
the
python
three
to
support
python
3.9
anyways,
because
they
apparently
are
removing
the
setup.py
test
way
of
doing
it.
So
if
this
ends
up
still
being
an
issue
at
the
end
of
the
rebase,
we'll
just
we'll
just
switch
the
docs
to
say
that,
since
we
have
to
do
that
at
some
point,
anyways,
I
think
I
already
just
removed
the
test
requires
stuff
and
everything,
and
so
we
should
be.
A
We
should
be
ready
to
do
that.
Okay,
anything
else
on
that.
A
Good,
I
saw
you
so
where
did
that
go?
Why
did
I
keep
scrolling
all
the
way
up?
Okay,
I
just
saw
your
pull
request.
D
A
Right,
sweet,
very
exciting,
okay,
all
right!
So
this
stuff
here
the
context
yes.
G
So
I
I'm
not
able
to
build
the
docs,
and
I
later
saw
that
the
code
block
thing
was
showing
I'm
not
able
to
see
how
it
it
looks
in
the
documentation.
So.
G
A
Yeah,
that's
weird:
let's
see.
A
A
A
A
No
okay,
so
yeah,
it's
got
to
be
a
local
issue.
Let's
check
it
out,
though,.
F
F
Yeah
who's,
this
guy
right,
it's
in
tutorials,
oh
the
neural
network,
tutorial
yeah,.
F
A
A
A
But
otherwise
this
is
looking
great.
Okay,
that's
weird,
though
okay,
so
I
think
I
mean
I
think
you've
definitely
got
an
install
issue.
A
A
Cool
great
yeah,
so
let's
just
make
sure
that
we
get
that
negative
one
in
there
and
then
I
think
all
right
and
thanks
for
commenting
more
okay.
So
let's
see.
D
G
A
Yep
there
you
go,
let's
see
what
are
you
doing?
Let's
see
so,
okay,
great
okay,
you
got
that
you
got
that.
Let's
just
check
these
okay,
so
we'll
just
I'll.
Just
just
anything
else.
You
want
to
talk
about
sexual.
A
Yeah,
I
would
try
to
uninstall
every
every
dfml
package
and
I
think
that
I
think
that
they
have
no
version
yeah.
So
there's
nothing
that
just
gives
you
a
nice
sort
of
list,
except
for
this
is
the
closest
thing
I
don't
know
if
this
does.
Should
I
too,
but
I
would
just
do
you
know,
let's
see
yeah
and
then.
A
Python
uninstall
dash
y
yeah.
This
basically
would
be
the
best
way
to
uninstall
all
of
them,
so
I'll
just
paste
that
in
the
chat.
A
Okay,
all
right,
so
let
me
just
make
note
of
this.
So
pull
requests
for.
D
A
Okay
and
then
with
the
oh
issue,
building
docks,
so
this
is
probably
an
install
issue.
Let's
try.
A
A
Okay,
okay,
great,
so
you
change
this.
Okay,
let's
see.
A
All
right:
okay,
perfect,
yeah,
okay
and
yeah,
and
then
I
was
waiting
for
saksham
to
give
it
a
look,
because
saksham
had
written
this.
Initially
all
right,
great,
we'll
merge
this.
Let's
see,
I
think
these
were
all
conda
download
failures.
C
A
Sorry,
so
I
think
the
thing
is
well,
it's
not
in
the
ci
right.
So
the
thing
is
that
the
auto
formatter
that
we
use
black
it
needs
to
be
installed
to
a
specific
version,
and
that
version
is
document
it's
in
this
setup
file
and
I
think
it's
documented
in
the
contributing
docs
as.
A
Oh,
you
just
need
to
format
it:
okay,
yeah,
okay,
yeah!
You
probably
just
need
to
run
the
auto
formatter,
then
yeah,
and
it's
probably
this
line.
I
would
guess
that.
A
Yeah,
so
let's
see
yes,
this
looks
good.
I
think
it's
just
because
you
need
another
line
here.
You
need
another
blank
line
and
then
you
need
to
to
add
back
this
line.
But
if
you
run
that
black
auto
formatter,
it
will
do
it
for
you
now,
let's
see
style
so
yeah
here
it
is.
This
is
what
you're
looking
for,
because
they,
basically
they
don't
guarantee
that
the
formatter
is
going
to
produce
the
same
results
between
versions,
and
so
you
know
sort
of
sort
of
the
point.
A
But
at
the
time
we
formatted
it.
It
was
on
that
version
and
if
we
move
to
the
latest
version,
then
we
have
to
reformat
all
the
code
which
creates
a
giant
reformat
commit.
So
I'm
not
sure
we're
gonna
do
that
anytime
soon,
at
least
maybe
before
the
beta,
all
right,
okay,
so
style
issue.
A
Okay,
all
right,
okay,
so
and
I'm
going
to
merge
all
these
pieces.
So
I
think
this
one
you
just
need
to
run.
A
Looks
good
and
oh,
I
don't
think
you
need
this
pass
line
here.
Let's
see
cool
very
nice
all
right!
Well,
I
will
talk
to
you
guys
all
next
week
or
I'll
talk
to
you
on
gitter
and
I
hope
you
guys
have
a
great
week
and
good
luck
with
any
exams.
If
anybody
has
exams.