►
From YouTube: Weekly Sync: 2019-10-22
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
A
A
B
B
A
I
push
any
more
errors
to
this
or,
like
should
I
remove
the
WIP.
That's
like
you
can
directly.
B
Yeah
I
think
I
think
this
is
good.
Sorry
I
must
I
have
I,
have
get
her
open
on
like
three
different
computers
and
sometimes
I,
don't
I
forget
to
hit
enter
on
my
response.
A
B
A
A
B
And
then
we're
gonna
get
a
chance
to
do
the
model
stuff
by
this
like
within
the
next.
You
know,
oh.
A
A
A
So
yeah
make
changes
okay
soon
and
then
let's
see,
models
urge
a
little,
not
trained
exceptions.
A
B
And
oh
and
then
let's
see
that
that
other
guy
I
can't
remember
his
name
right
now
again,
he
said
he
looks
like
he.
He
did
the
regression.
He
added
a
regression
model
for
tensorflow,
which
is
nice.
B
A
It's
you
you
have
tested
tensorflow
like
is
it?
Is
it
very
heavy
on
the
system
or
like
it's?
Okay,
I
mean.
B
It's
it's
tensorflows
like
this
Beast
of
a
thing
and
it
works
really.
Well,
you
know
for
a
lot
of
things,
but
it
doesn't
it.
It's
definitely
slow
for,
like
things
that
would
be
quick
with
scikit
like
it's,
it's
a
bit
of
a
Overkill
for
what
most
of
what
we're
doing?
It's
great
it's
a
really
great
for
a
lot
of
the
image
processing
stuff
that
many
people
do
and
we'll
have
to
get
into
the
image
processing
eventually.
B
But
for
now
it's
a
bit!
It's
it's
it!
It's
you
you!
Basically,
if
you
import,
tensorflow
like
it
will
take
like
10
seconds
just
to
load
just
to
do
the
import
statement
on
on
even
like
a
fast
computer,
usually,
which
is
just
it's
a
bit
of
a
pain
in
the
ass
but
other
than
that.
I
mean
it's
great.
That's
just
the
you
know.
If
we're
saying
is
it?
Is
it
Overkill?
It's
a
bit
Overkill,
sometimes,
but
it's
nice
to
have
it
in
there
right.
Let's
see,
okay,
what
was
I
closed?
A
I
wonder
where
we
got
they
went
here
right.
You
just
got
the
option
to
squash
and
merge
it
did
you
yeah,
but
it
won't.
Oh
wait!
Now
it's
gone.
Did
you
see
that
yeah
yeah
I
was
I
was
actually
mentioning
this?
Something
like
it
was
squash
and
merged
earlier
and
then
it
became
yeah.
Then
it.
B
A
B
Yeah,
that's
good
all
right,
good,
sweet
done,
good
sweet,
so
yeah,
so
sudarsana
said
that
she
is
gonna,
be
busy
for
the
next
coming
few
weeks.
Here
she
has
some
stuff
going
on
at
work,
so
the
she
was
gonna
get
started
on
the
web
UI.
Maybe,
but
now
it
sounds
like
we're,
not
really
sure
I'm.
Actually
gonna
move
this
stuff
here
that
I
wrote
down
after
the
meeting
last
time
to
down
here,
because
I
didn't
actually
talk
about
this
with
anyone.
B
So
we
can
talk
about
web
UI
stuff.
Now.
A
A
A
Api
talk
to
you
about
the
initialization.
What
initialization.
B
Yeah,
so
we
thought
about
well,
so
we
we
were
thinking
about
it
and
we're
like
okay,
there's
a
couple
ways:
we
could
do
this
right.
You
could
put
it
under
your
org
if
you
wanted
to.
If
you
want
to
keep
it
separate,
if
you
want
to
drive
the
UI
stuff
or
we
could
put
it
under
a
different
branch
under
this,
the
main
repo
or
we
could
put
it
as
a
subdirectory
now.
B
B
I
could
yeah
I
could
I
could
do
that?
It
just
involves
me:
I
have
to
do
a
bunch
of
process
stuff,
if
I
do
that.
B
So,
whereas
I've
already
got
this
one-
and
we
can
put
things
under
this
one
without
without
having
to
go
through
a
bunch
of
process
from
the
Intel
perspective,
so
yeah,
but
then
also
you
know
if
you
want,
if
you
want
to
just
leave
the
web
UI
and
have
it
under
that
and
code
Ed
org,
we
could,
you
know,
start
doing
stuff
over
there
and
you
could
just
you
guys,
could
just
lead
that
and
I'll
just
you
know,
jump
in
whenever
you
need
help
on
things,
but
it
depends
what
you
want
to
do,
whether
you
want
the
work
to
stay
under
the
Intel
Banner
or
if
you
want
to
to
spread
under
that
one,
because
we
can
always
put
it
back
under
here
later.
A
B
Right
yeah,
if
you
don't
have
any
strong
feelings
about
putting
it
anywhere,
I
would
say:
let's
just
put
it
all
in
one
place:
okay!
So
because
you
know
that's,
then
it's
all
in
one
place
and
the
other
thing
is.
A
B
Yeah,
that's
true
yeah.
That
could
be
a
good
one
yeah.
That
could
be
a
really
good
project
idea.
We'll
probably
we'll
probably
do
that
then,
because
I
was
just.
You
know
that
guy
was
asking
me
about
project
ideas
and
I
was
gonna
shoot
out.
He,
it
sounds
like
he's
already,
leaning
on
that
one
thing,
which
is
operations
and
making
operations
into
models
or
models
into
operations
which
is
basically
just
like
using
the
train
or
using
the
predict
function
of
of
a
model.
So
let's
just.
B
A
It's
like
a
very
big
thing
for
some
people
like
when
you
do
it.
You
feel
like
it's
not
that
big
of
deal
but
okay,
some
people
just
do
it.
For
the
sake
of
that.
A
B
Yeah,
that's
yeah!
That's
it's
great
to
have
on
your
on
your
resume:
yeah,
okay,
so
yeah
web
UI
making
products
under
operations.
These
are
a
couple
ideas
here.
The
web
UI
is
definitely
a
very
full-fledged
project,
so
yeah
we
could
leave
that
as
as
gsoc
2020.
I
was
thinking.
So
I've
got
this
presentation
on
Saturday
and
I.
Don't
know
how
far
I'm
gonna
get
on
things,
but
it
depends.
It
depends
how
far
I
get
this
week.
B
I,
don't
I,
don't
know
if
I'm
going
to
get
very
far
or
not,
but
I
might
I
might
actually
just
start
on
the
web
UI,
because
if
I
can
go
to
that
conference
and
then
show
a
web
UI
at
least
part
of
something
that
might
be
helpful,
I
may
I
may
leave
off
part
of
it.
So
I
may
just
do
like
operations
and
then
do
models
and
stuff
as
the
gsoc
project,
but
it
all
depends
honestly
right
now.
I
probably
won't
even
have
time
for
that.
So
it's
it'll,
probably
just
stay
stay.
B
I
could
I
can
if
it's
a
web
websites
websites
I,
can
throw
together
very
quickly
things
that
aren't
websites
take
a
little
more
time.
It's.
A
A
B
A
B
Yeah
so
Hugo
Hugo
I've
used
recently
is
pretty
good.
Have
you
seen
Hugo,
no
okay,
so
Hugo's
a
good
one.
Also
I
think
there
was
one
based
off
Hugo.
That
was
like
really
point
and
shoot.
This
isn't
a
very
good
description
of
what
it
looks
like.
Basically,
it
builds
these
static.
You
build
these
static
websites,
and
so
you
you
can
you
there's
all
these
templates
and
then
you
it
basically
takes
the
markdown
files
and
then
it
puts
them
into
the
templates
and
it
it
looks.
It
looks
good
what.
B
Files
yeah,
so
you
write
yourself
yeah,
you
write
your
blog
post
in
markdown
and
then
it
dumps
them
into
it,
dumps
them
into
HTML
files.
It's
good.
A
B
B
This
could
be
good,
but
oh,
it
might
be
like
a
paid
thing.
Yeah
there's
a
lot
of
stuff
around
like
Hugo
has
been
oh
yeah.
This
could
be
good,
I,
don't
know.
Oh
this
yeah.
This
looks
like
some
kind
of
paid
thing.
Yeah
I
would
recommend
just
messing
around
with
Hugo
at
first.
The
other
thing
that
is
so
this
chuckle
have
you
seen
Jekyll.
B
So
I
mean
this
is
stuff,
that's
gonna!
Let
you
you're
not
gonna
need
to
do
really
web
design
when
you're,
using
these
guys
so
with
with
Jekyll
the
the
bit
so
Hugo
Jekyll
was
sort
of
like
the
big
static
site
generator
and
if
you
make
a
site,
if
you
make
a
Jekyll
site
and
then
use
that
as
like
your
GH
Pages
Branch
or
your
you
know,
coded.github.io
and
just
like
only
put
the
Jekyll
site
in
there.
B
A
B
A
lot
of
people
are
using
it
lately,
because
the
reason
why
people
I
think
a
lot
of
people
switch
to
using
Hugo
is
because
it's
written
in
go
and
so
the
guide
to
gives
static,
binaries
statically
linked
binaries,
which
means
that,
like
there's
no
dependencies,
so
you
can
just
download
this
binary
and
run
it
and
all
of
a
sudden
you
have
a
website,
whereas
with
Jekyll
you
have
to
set
up
your
whole
Ruby,
environment
and
stuff,
and
so
you
know
people
there
were
a
lot
of
people
doing
Ruby
as
web
development
like
still
are,
but
like
Ruby
was
a
big
thing
with
web
development
a
few
years
ago,
especially,
and
so
Jekyll
was,
you
know
very
friendly
to
all
of
them,
because
they
already
have
all
their
Ruby
stuff
set
up.
B
Whereas
now
like
as
when
go,
came
along
in
this
guy,
build
Hugo
and
go
everybody
else
who
wasn't
a
big
Ruby
person
or
you
know,
could
was
you
know.
Ruby
has
some
weird
Oddities
about
setting
up
the
environment.
Just
like
you
get
weird
Oddities
with
python
setting
up
pip,
and
so
people
said.
Oh
okay,
here's
this
Hugo
thing
like
it
does
the
same
thing
as
Jekyll
pretty
much
but
like
I,
don't
have
to
fight
with
it
to
get
it
set
up.
It
just
works,
so
that's
become
big
for
that
reason.
B
B
No
no
problem,
that's
always
always
feel
always
feel
free
to
shoot.
Questions
like
that
and
then,
as
far
as
as
far
as
just
like
JavaScript
is
concerned
like
if
you
wanted
to
write
your
own
website,
I
would
say:
go:
go
mess
with
the
react.
Stuff.
That's
seem.
It
seems
to
be
everybody's
really
into
react
these
days.
Well,
we
talked
about
flutter.
We
talked
about,
and
flutter
is
great
of.
A
B
Yeah,
it
is
aspects
it's
much
better.
What
I've
been
using
for
CSS
stuff
is
just
this
material,
UI
Library,
it's
it
works.
It's
not.
It's
not
always
like
exactly
what
you
want,
but
it
it
does.
It
does
give
you
all
the
CSS
that
you
want
so
like
they'll.
Do,
let's
see.
B
Yeah,
so
this
is
it's:
it's
all
these,
it's
all
the
CSS
and
UI
stuff
and
like
making
it
look
nice
and
then
integrating
it
with
react.
So
that's
this
is.
This
is
good.
If
you
want
to
do
react,
it's
it
works.
I
just
wanted.
A
Okay,
yeah.
B
And
I
wrote
down
I
just
wrote
down
some
notes
for
like
things
that
the
web
UI
should
have
and
I'll
convert.
I
think
I'll
convert
this
into
a
gsoc
project,
so
you
know
we
need
to
be
able
to
create
and
configure
new
models
and
sources.
We
need
to
be
able
to
list
all
the
available
models
and
sources
which
are
available
for
creation.
B
We
need
to
view
all
the
repos
in
a
source
we
need
to
be
able
to
edit
data.
You
know
like
pull
up
a
little
little
form
type
thing
to
edit
the
data
edit.
The
various
feature
data.
If
we
need
to
change
something
and
we
need
to
be
able
to
add,
add
new
repos
and
also
like
probably
a
list
to
see
all
of
them
for
models.
You
know
we
need
an
interface
for
training.
You
got
to
upload
your
data
and
stuff.
We
need
an
interface
for
accurate
assessment.
B
Maybe
you
know
you
upload
your
test
data
set,
or
you
know
that's
that
has
to
do
with
sources
really,
so
you
would
create
new
sources
to
do
that
stuff,
and
then
we
need
an
interface
for
prediction
which
would
basically
just
be
like
hey.
You
know,
here's
this
data
give
me
a
prediction
basic
stuff
like
that
would
be
that
would
get
us
off
the
ground
right,
and
this
is
just
just
I,
don't
know
if
you've
started
to
dabble
with
the
JavaScript
yet,
but
this
is
what
the
API
looks
like
right
now.
B
So
basically,
this
is
just
saying:
take
the
port
of
where
you're
at
and
replace
it
with
eight
zero,
eight
zero.
So
if
you're
on
localhost
5000,
it's
replacing
the
port
with
8080,
which
is
where
we
start
the
HTTP
API.
B
So
we
create
this
API
object,
we
say:
hey
I
want
to
I,
want
to
create
a
new
source
object,
and
then
we
give
it.
We
give
it
the
data
for
the
source.
We
say
upload
that
data
to
my
trainingdataset.csv.
We
configure
it
and
we
say
you
know.
Here's
use
this
use
this.
My
training
dataset.csv
is
the
file
name.
B
This
is
probably
gonna
get,
you
know
turned
into
so
the
all
the
config
objects
that
we
are
using
were
they
they
essentially
their
representation.
Is
this
and
that
allows
it
to
like.
So
you
can
set
like
this,
for
example,
The
Source
ARG
might
be,
you
could
sit,
you
can
set
like
the
what
the
source
should
be
and
then
what
the
config
for
the
source
should
be
only
we
already
know
what
the
source
should
be.
It's
CSV
here,
so
we
we
don't.
B
We
don't
specify
in
there
and
but
but
well,
we'll
make
objects
for
this
that
make
it
make
it
more
clean
looking
than
than
this
sort
of
mess,
and
then
you
create
the
context.
So
basically,
that's
the
same
pattern
that
we're
doing
with
the
async
4
to
create
the
main
object,
that's
the
configure
and
then
we
do
async
4
and
then
we
call
the
main
object
and
then
we
get
the
context.
B
B
And
then
we
do
the
same
thing
for
the
testing
source
and
then
we
create
the
test
Source
context,
and
so
this
is
just
an
example
of
saying:
okay
like
I
want
to
I
want
100
repos
out
of
that
Source,
and
then
you
know
the
repos
are
returned
to
you
in
a
key
value:
mapping
of
the
source
URL
to
the
repo
object.
But
this
is
how
we
would
change
that
into
an
array
and
then
we
create
a
model.
B
We
configure
the
model
scikitlr
and
we
create
the
model
context,
and
so
when
we
create
the
model
context,
this
is
what
I
was
talking
about
with
the
features
and
saying
you
know,
I
wonder
if
we
should
maybe
move
this,
we
should
move
the
features
from
the
creation
of
the
context
like
because
remember
when
we
first
did
the
scikit
model
and
the
scratch
model
I
think
it
was
actually
the
scratch
model.
We
found
out
that
the
features
were
being
passed
to
the
train,
predict
and
accuracy
method,
and
we
said
well
I
mean
why.
B
B
And
now
I'm
realizing,
you
know
as
we
save
and
load
those
models
we
end
up
having
the
model
do
some
saving
and
loading,
and
then
the
model
context
also
doing
some
saving
and
loading
based
off
those
features.
So
maybe
we
should
just
put
that
put
that
config
into
the
model,
rather
than
the
model
context
like
put
the
features
into
the
config
of
the
model
rather
than
the
model
context,
and
then
this
context
call
would
be
the
same
as
the
source.
B
One
where
we're
just
saying
context
like
with
the
name
of
the
context
and
we'd
pass,
the
features
in
with
the
configuring
of
the
model
did
I
can't
I
know
we
talked
about
this
briefly
before.
Do
you
think
that
that's
a
good
way
to
go,
or
is
there.
B
A
A
B
A
A
Let's,
just
let's
just
create
event,
so
so:
okay
model
move
features
from
it
model
context
to
model
so
oops.
B
So
right
now
features
are
passed
to
the
init
method
of
context.
That
says.
A
Okay,
so
we
should
probably
settle
down
soon
like
because
this,
these
kind
of
changes
require
a
lot
of
refactoring
yeah.
B
Yeah
yeah
this
is-
and
this
is
this-
is
where
I
think
this
one
was
an
important
one
to
do,
because
it
this
is
the
this
is
inconsistent
with
the
way,
we're
doing
everything
else.
Everything
else
is
being
done
by
the
config
and
all
of
a
sudden.
We
we
were
doing
this
and
and-
and
we
probably
shouldn't
have
done
that
so,
but
it's
it's
a
learning
process,
so.
A
B
Forgot
about
that
I
need
to
change
the
I
need
to
change
so
yeah.
Basically,
this
I
needed
to
add
spaces
here
and
then
I
need
to
it's
kind
of
things
that,
like
aren't
really
worth
going
back
and
forth,
to
explain
this
needs
four
spaces
indented
here
these
are.
These
are
things
that
would
take
too
much
time
to
go
round
trip
so
I'm
just
going
to
indent
them.
B
But
let's
see
is
it
doing
that
not
a
directory.
Okay
looks
like
he's
raising,
not
a
directory
error.
B
Go
make
that
let's
see
thank
you
for
bringing
that
up.
I
need
to
do
that
then
so
notes
for
myself.
A
I
I
change
it
for
everywhere
you
you
can
just
request
the
change
actually
do
what
do
what
I
actually
changed?
It
everywhere,
I
changed
it
for
tensorflow
2
and
oh
okay,
great.
B
A
A
B
I
saying
his
name
right
again,.
A
B
A
Needs
to
I'm
not
sure
into
model,
not
trained
error.
B
Hey
what
happened
with
the
did
you
get
an
internship
for
this
summer
yet
or
where
I
thought
you
said
you
were
doing
something
doing.
A
So
I
I
am
actually
looking
for
a
summer
internship
still
yeah.
Okay,
if
you
have
any
leads,
do
let
me
know.
Oh
yeah,
you
know
I
will
oh
yeah,
you
know
yeah
John,.
A
I'll
change
her
to
indent,
well,
I,
just
I
wanted
to
ask
you
this
to
like.
Did
you
receive
the
gsoc
hoodie?
You
should
have
oh
I've.
B
I
got
a
t-shirt,
but
it's
probably
because
I'm
in
I'm,
under
the
python
software
Foundation
we're
not
like
yeah
we're
not
like
I,
think
I.
Think
there's
like
people
who
are
directly
under
Google
summer
of
code
and
then
there's
the
software
Foundation
has
like
Carey
one
of
my
co-workers,
she's.
Actually,
the
mentor
for
the
python
software
foundation
and
everybody
else
under
there
is
like
a
sub
org.
So
we
probably
only
get
t-shirts.
B
A
B
I
think
that's
going
on
right
now,
and
that
is
also
because
Harry
is
there,
because
it's
only
the
top
level
Works
but
I'm
gonna
talk
to
her
when
she
gets
back
and
and
see
how
it
went.
I
think
she's,
actually
flying
home
pretty
soon.
Here
it
was
in
Germany.
A
B
Germany
yeah
yeah,
so
I'm
gonna
talk
to
her
as
soon
as
she
gets
back
and
see
see
what
what
the,
what
all
the
information
was
there,
but
it
it
looks
like
from
what
I
saw
on
Facebook.
It
looks
like
it
was
going
well
so.
B
B
B
A
Very
very
python:
python
yeah.
This
looks
cool
yeah,
oh
workshops.
B
Interesting
yeah:
this
is
something
to
look
into:
there's
an
online
Forum
to
submit
grants.
Hey
that
sounds
cool.
B
Oh,
very
oh,
it
looks
like
yeah
kinda
specific
to
to
organizing
python
software
Foundation
stuff
we're
doing
projects
for
for
that
whole
infrastructure
there,
but
worth
taking
a
look
at
for
more
info.
B
All
right,
I,
don't
really
have.
Let's
see
I
think
we
finished
going
through
that
I
can
show
you
I
can
show
you
real,
quick,
the.
How
that
that
that
HTTP
API
looks.
B
Here
I'll
show
you
real
quick
what
it
looks
like
okay,
this
is
this
is
my
work
on
the
on
the
it's
Dash.
A
A
All
right,
let's
try
this
empty
and
and
do
it
for
now-
service,
http,
server,
of
course,
star
and
secure.
Oh.
A
B
All
right
here
we
go
so
it'll,
do
you
know
it'll,
say
created
the
API,
create
a
training,
Source
created
the
training
data
set
uploaded,
it
created
the
model
and
now
see
now
it's
loading
tensorflow
and
it
ran
the
LR
model
and
so
model.
Oh
accuracy,.
A
B
It
looks
like
accuracy,
I,
don't
know
what
happened
with
accuracy
there,
but
okay.
So
here
we
got
our.
We
got
our
our
prediction
back
so
yeah,
so
we
can
use
the
we
can
use
all
the
stuff
that
we
have
within
the
python
stuff.
We
can
use
it
from
JavaScript
now,
which
is
which
is
fun
so,
but
that
is
really
all
I
guess,
that's
all
for
this
meeting.
Unless
you've
got
any
anything
else,
you
want
to
talk
about.
A
A
B
That
would
be
that
would
be
cool
the
more
the
more
the
merrier
right
and
then
the
other
thing
is.
We
want
to
really
think
about.
You
know,
okay,
so
here's
here's!
This
is
sort
of
a
more
major
thing.
Is
images.
B
Oh
yeah
yeah.
That
would
be
that
would
be
interesting
to
see
you
know
what
what
can
we
do
with
that,
because
we've
got
we've
got
all
this
stuff.
I
mean
we're
the
images
you
know.
Python
has
a
bytes
object
right
and
so
we'd
probably
just
store
the
images
as
bytes
object,
and
then
you
know,
there's
not.
A
I
think,
like
the
feature
stuff
was
more
interesting
for
images
like
model
stuff
is
a
later
stage.
After
that
teachers
thing,
we
were
planning
before
G
G-Shock,
which
was
that
the
custom
features
like
DF
FML,
hard
sources
models
and
features
right.
Oh
yes,.
B
Oh
I
forgot,
so
this
I
probably
didn't
make
that
clear.
The
features
the
features
are
now
the
operations,
so
the
feature
stuff
all
still
exists,
but
I
I.
The
way
that
you
generate
the
features
is
via
these
operations,
and
so
I'm
gonna
have
like
that's.
What
I've
been
working
on
like
a
lot
lately
and
I'm
gonna
have
by
the
end
of
this
week
is
okay
here,
I'll
show
you
I'll
show
you
some
of
what
it
oh
dang,
I,
think
I,
just
I,
don't
know
if
I
can
show
you
right
now.
B
Okay,
so
this
is
this:
is
you
write
these
operations,
which
are
just
like
little
python
functions
right
like
this
and
these
ones?
The
example?
Is
that
we're
doing
running
like
sub
process
and
stuff
and
just
parsing
the
Json
output,
and
using
that
as
some
feature
data,
and
so
what
what
we
end
up
with
is
you
know
a
bunch
of
these
little
operations
and
we
chain
them
together
with
their
inputs
and
their
outputs.
B
We
can
generate
all
of
this,
and
so
what
we're
going
to
be
able
to
do
is
have
the
web
UI
be
like
a
drag
and
drop
configuration
of
all
this
stuff
too.
B
So
it's
you're
going
to
be
able
to,
like
you,
know,
really
easily
create
new
data
sets
by
changing
the
Generation
and
and
it
allows
for
obviously
when
we
make
so
when
we
make
models
into
operations.
You
know
these
are
operations
right,
and
so,
if
we
make
a
model's
predict
function
as
an
operation,
then
that
lets
us
really
easily
create
complex
features,
and
so
this
is
going
to
open
up
a
lot
of
really
interesting
things.
B
I
think
things
I
think
it's
going
to
be
really
cool,
and
all
of
this
runs
concurrently
so
like
if
you're
processing
a
you,
know
a
bunch
of,
if
you're
doing
like
a
bunch
of
network
operations
in
all
of
these,
like
it'll
all
run
at
the
same
time,
which
makes
it
really
fast,
but.
A
B
Exactly
and
they're
all
plug-ins
as
well,
so
you
can
we're
we're
going
to
have
a
set
of
them
that
are
within
the
main,
git
repo,
and
then
you
can
publish
your
own
Pi
Pi
package
that
contains
these
operations.
With
that
Dev
create
script.
You
know
how
you
used
it
to
create
a
new
model
package.
B
I
have
I
have
another
one
that
creates
a
new
package
of
operations
and
you
can
publish
that
to
Pi
Pi
and
then
all
you
have
to
do
is
you
know,
tell
people
that
go
install
this
package
and
they'll
have
access
to
all
these
operations.
Now
that
you've
done,
and
so
people
can
just
you
know,
publish
all
these
various
operations
and
use
them
use
each
other's
operations
to
create
new,
interesting
data
sets
and
and
then
obviously,
we've
already
got
the
machine
learning.
B
This
is
sort
of
like
you
know
they
and
and
end
overarching
goal
thing
that
we've
been
working
towards
is
getting
all
this
stuff
working
because
it
you
know
it
provides
a
lot
of
flexibility
with
data
set
generation
and
with
using
the
trained
models
you
can.
You
know
you
don't
have
to
write
you.
You
don't
really
have
to
write
code
to
make
this
work
right,
which
is
a
really
cool
thing,
because
it
lets
people
who
don't
who
don't
know
much
about
programming
like
they
can
they
can
get.
B
They
can
start
doing
interesting
things
right.
They
can
say
hey
when
somebody
emails
me
this
Excel
spreadsheet,
you
know,
use
this
other
Excel
spreadsheet,
so
I
I
have
this
Excel
spreadsheet
of
all
the
things
that
you
know,
I've
manually
classified
or
I've
done
a
prediction.
My
prediction
of
by
Googling
right
well,
so
I
trained
a
model
using
the
web,
UI
I
said,
go
train
me
a
model
and
then
I
say
every
time
you
know
I
could
have
our
operation.
B
B
Yeah,
this
is
going
to
be.
Definitely
this
will
be
gsoc
idea.
Type
stuff
I
still
need
to
flush
out
more.
The
other
thing
is
so
you
can
run
this.
The
the
cool
thing
about
this
is
when
you
define
it
like
this
and
you
just
Define
all
the
operations
and
how
they
connect
I've,
now
got
it.
The
major
pull
request-
that's
going
in
right
now
is
to
let
you
run
the
same
thing
over
the
HTTP
API
as
this
command
line
interface.
B
So
you
can
take
this
this
exports
to
like
a
yaml
or
Json
file,
and
that
shows
all
the
interconnections
and
then
you
can
feed
you
can
give
that
yaml
file
to
a
command
line
interface,
and
it
will
run
that
it
will
run
that
as
the
command
line
program
or
you
can
give
it
to
the
HTTP
API
and
you
can
say:
hey
when
I
hit,
the
slash,
should
I
URL
run
this
data
flow
and
it
it.
B
You
know
it
enables
you
to
to
only
write
the
little
operations
unit
test
those
and
then
the
integration
is
just
taken
care
of
by
this
execution
engine
that
we
have
and
then,
of
course,
it
really
easily
allows
you
to
to
swap
in
and
out
pieces
and
add
machine
learning
places,
because
you
can
just
say
hey.
You
know
after
you've
done
this
safety
check
and
run
Bandit
I
want
you
to
feed
the
output
of
those
two
through.
B
This
predict
function
for
a
model
that
I
chained
trained
on
a
data
set
of
of
assessing
a
bunch
of
these
and
and
saying
whether
and
manually,
classifying
whether
they
are
good
or
bad.
So
that's
going
to
be
an
extended
part
of
this
example
that
we'll
do
eventually
but
yeah.
So
that's
where
that's
where
the
stuff
is
going,
but.
A
B
A
Sorry
you
cut
out
there.
Is
there
a
new
operations
tutorial
update?
No
I.
Did
you
put
up
any
like
new
operations
tutorial
this.
B
Is
so
I've
been
updating
this
I've
been
trying
to
update
this
and
actually
I
need,
there's
an
open
issue.
If
you
had
time
to
go,
go
through
this
I
would
appreciate
that
I
mean
I,
understand
that
you've
already
got
to
do
the
models
to
true
and
stuff,
but
we
need
somebody
to
go.
Make
sure
that
this
thing
makes
makes
sense,
because
I
did
it
that
I,
don't
I,
don't
know
if
it
all
makes
sense.
B
So
if
you
just
read
through
it
and
could
tell
me
if,
if
it,
if
it
sounds
reasonable
or
not,
that
would
be
great.
If
you
have
time,
of
course,
just
comment,
there's
an
issue
open
to
say:
does
this
make
sense?
B
Yeah,
so
that
yeah,
because
I'm
not
sure
you
know
I
wrote
that
stuff
and
it's
tries
to
explain
it
and
that's
the
stuff
that
arvind
was
working
on.
He
was
kind
of
working
on
this
operation,
stuff
and
but
but
I,
don't
he
hasn't
looked
at
the
tutorial
since
I
updated
it?
B
No
one
has
so.
It
would
be
it'd
be
great
to
have
a
second
set
of
eyes,
because
yeah
the
way,
the
way
that
dff
ml
is
structured
right
is
there's
the
sources
where
we're
saving
and
loading
all
of
this
data
there's
the
models
where
we're
doing
machine
learning
and
then
there's
the
operation,
which
is
the
you
know.
B
We
collect
the
data
sets
and
now
we're
going
to
have
the
in
in
the
way
to
collect
we're
going
to
have
this
way
to
collect
data
sets
and
we're
going
to
have
this
way
to
you
know
easily
use
the
trained
models
on
other
data,
so
yeah
yeah,
we'll
I'll,
see
how
this
is.
This
is
received
when
I
go
to
Beachside
this
weekend.
I
want
to
I'm
gonna
present
this
stuff
and
and
talk
about
it
and
we'll
see
what
people
think,
whether
they
think
I'm
nuts
or
whether
it's
going
to
be
useful.
B
B
A
B
Right
well,
if
there's
nothing
else,
then
I'll,
we'll
sign
off
here
and
I'll,
see.