►
Description
On this week's community standup, Microsoft MVP Jon Wood is joining us to discuss various ways of contributing to ML .NET
Community Links: https://www.theurlist.com/mlnet-standup-2020-10-07
Featuring: Jon Wood (@JWood)
#MLNET #OpenSource
A
A
Just
some
youtube
videos
yeah
we'll
go
over
those
later
john,
has
been
super
great
and
and
important
in
building
the
ml.net
community
and
helping
people
get
started
and
we'll
show
off
his
work
in
a
bit
but
yeah.
First,
let's
go
through
some
community
links.
Here
is
the
url
list
for
that.
The
link
for
that
and
we'll
put.
A
Comments
of
the
recorded
youtube
video,
so
I'll
share
my
screen
here.
All
right,
so
we've
got
a
huge
list
today,
I'll
go
ahead
and
get
started.
The
first
one
is
just
kind
of
an
announcement.
A
few
weeks
ago
we
hit,
I
guess
in
september
10th
we
hit
100k
installs
of
model
builder,
which
is
awesome,
so
that
was
a
really
really
big
time.
A
A
Yeah,
this
is
exactly
why
so
this
is,
if
you're
not
familiar.
This
is
reddit.
This
subreddit
is
our
slash
programmer
humor
and
someone
actually
posted
a
screenshot
of
adding
machine
learning,
which
is,
of
course
for
model
builder,
and
they
said,
oh,
it's
as
easy.
As
that,
no
wonder
it's
popular.
I.
I
think
this
is
a
very
big
success,
making
it
to
programmer
humor,
you
know,
mainstream
tech
is
great
and
it
was
actually
great
to
see
someone
leaving
really
positive
feedback
about
how
it
is
actually
really
simple
to
use.
A
So
that
was
really
awesome
to
see.
Then
here
we've
got
christopher.
So
after
our
last
community
stand
up,
we
had
bart
on
talking
about
dot
net
5
blazerwebassemblyml.net.
He
actually
said
that
he
got
inspired
by
the
talk
and
I
think,
before
the
stand-up
was
even
over.
He
followed
on
the
tutorials
and
put
and
made
it
in
blazer.
So
he
had
good
or
bad
text
or
that
sentiment
analysis
and
was
able
to
implement
that
into
a
blazer
web
assembly.
Application.
B
A
A
I
know
exactly
what
that
is:
it's
the
data
set
that
was
used,
which
just
proves
the
importance
of
your
data
set
that
you
use.
This
is
what
the
wikipedia
toxic
versus
non-toxic,
I
think,
there's
only
like
250
examples
in
it.
So
christopher
I'll
challenge
you
to
actually
train
with
the
yelp
label
data
set,
which
will
actually,
I
think,
give
you
a
bit
better
results
there.
B
Yeah
yeah,
so
we
noticed
all
the
cool
cats
were
getting
on
discord,
so
we
now
have
a
channel
on
discord
as
well
that
you
can
join
and
come
chat.
We
know
that
there's
a
few
channels
for
you
to
you
know
get
into
contact
with
the
team.
You
can,
of
course
you
know,
open
up
issues
and
start
conversations.
There
there's
also
the
getter
channel,
which
is
again
tied
to
the
to
the
core
repo
right
to
the
machine.net
machine,
learning
repo,
but
discord
or
the
channel.
That's
in
discord.
B
A
A
B
B
A
B
D
C
A
B
A
B
A
B
It's
next
tuesday,
I
believe
yeah.
So
if
you're
in
ndc,
if
you're
attending
ndc
sydney,
I
believe
it's
virtual
now
so,
if
you're
attending
it-
and
you
want
to
learn
a
little
bit
more
about
net,
you
know
jks
he's
a
very
great
presenter
and
he's
this.
You
know,
presentation
of
his
size
is
really
great,
so
make
sure
to
check
it
out.
A
And
then
so,
I'm
for
most
of
you
know
alexandra
by
now
he
actually
was
on
the
six
figure
developer,
talking
about
ml.net
and
ml
ops,
which
is
his
open
source
project
and
so
to
talk
a
little
bit
more
about
ml
or
sorry
mlx.net.
I
guess
I
should
say
emma
lops,
if
you
don't
know,
is
managing
the
model
lifecycle,
so
it's
kind
of
like
devops,
for
machine
learning
or
for
your
machine
learning
models.
A
It
can
be
things
like
you
know,
versioning
your
models,
deploying
your
models,
keeping
track
of
all
your
data
and
different
versions
of
the
data.
There's
a
lot
of
different
applications
here,
and
this
is
actually
an
open
source
project
that
alexander's
been
working
on.
I
believe
he
was
on
our
first
community
stand
up
for
machine
learning.
He
talked
a
bit
about
it,
so
this
is
his
blog
post
about
it
and
he
actually
had
a
new
release.
B
Yeah
yeah,
so
1.0
was
actually
not
too
long
ago.
It
feels
like
it
wasn't
that
long
ago,
so
getting
to
1.2.
It
was
really
great
and
there's
a
lot
of
really
there's
a
lot
of
pieces
coming
into
play
here,
right
and
actually
alexander's
in
the
chat.
So
if
you
know
you
have
additional
questions,
you
can
ask
him,
but
so
one
of
the
pieces
and
then
some
of
the
work
that
he
started
doing
was.
He
started
developing.net
templates
right
so
so
template.
B
So
you
can
say
dot
net
new
and
then
you
can.
You
know
basically
auto
generate
a
startup
project
right
for
remote.net
and
then
he
took
it
a
step
further
where
he
essentially
integrated
this
and
made
it
so
that
you're
able
to
create
or
basically
scaffold
a
sort
of
asp.net
core
web
application
right
and
then,
if
that
wasn't
enough
right,
he
went
ahead
and
made
it
super
easy
to
basically
containerize
and
then
deploy
to
kubernetes.
B
The
application
in,
I
would
say,
a
few
lines
of
code,
maybe
three
three
to
five
lines
of
code,
maybe
less
and
yeah.
So
it's
it's
super
simple.
If
you're
looking
for
that
sort
of
highly
scalable
and
distributed
sort
of
deployment
via
containers
and
an
asp.net
core
applications,
this
this
release
is
going
to
enable
you
to
do
that,
so
definitely
check
it
out.
Congratulations,
alexander
and,
and
really
the
community
that
he's
built.
I
know
that
there's
a
lot
of
passionate
community
members
that
are
contributing
to
it.
B
A
B
A
B
Yeah
so
michael,
I
hope
that
I'm
pronouncing
that
one
right
he
put
together
an
article
before
about
how
you,
how
do
you
get
values
out
of
prediction
values
out
of
basically
just
general
models
in
nimble.net?
In
this
case,
though,
he's
doing
it
specifically
with
honest
models
so
for
folks
who
may
not
be
aware
in
ml.net
you're
able
to
score
or
make
predictions
with
onyx
models
right.
B
A
B
So,
for
example,
if
you
have
data
health
data
on
on
a
patient
right,
you're
able
to
sort
of
give
priorities,
certain
patients
who
may
be
more
of
at
risk
of
certain
heart
conditions
right,
so
they
leverage
dental.net
to
build
that
model
and
I
believe
it
might
be
in
production.
So
it's
really
great
to
see
that
companies
are
again.
Companies
are
really
embracing,
ml.net
and-
and
it's
actually
out
there
in
the
wild
and
then
in
production
right
so
yeah.
This
is
a
really
cool
use.
Center
use
case.
A
Not
only
in
production
but
there's
just
so
many
different
use
cases
for
it
and
so
many
different
kind
of
areas
like
healthcare
law.
You
know
my
favorite
hazelnut
story
like
it's
pretty
crazy
to
see
what
people
are
coming
up
with
yeah
great,
so
this
one
was
actually
in
a
visual
studio
magazine
written
by
david.
A
And
this
is
pretty
similar,
so
this
is
just
a
blog
post
about
these
some
september
updates
for
ml.net,
one
of
them
being
object,
detection
and
model
builder
and
this
tool
called
vot.
So
currently,
in
model
builder,
we
don't
offer
a
way
for
you
to
tag
your
photos.
We
or
your
images.
We
expect
you
to
have
that
data
already
labeled
and
we
suggest
a
tool
called
bot,
which
is
an
open
source,
microsoft
tool
that
you
can
use.
A
A
B
A
Yeah
definitely-
and
then
I
think
we
talked
about
this
last
time,
but
we
released
one
ml.net
152..
We
there
was
a
regression
in
151,
so
we
recommend
that
you
skip
that
one
and
go
straight
to
1.5.2,
and
you
can
see
the
release
notes
here.
I
think
we
had
our.
I
had
a
previous
blog
post
that
talked
about
all
the
new.
A
B
Yeah
yeah
yeah
totally,
so
this
one
is
it's
actually
something
that's
been
out
for
quite
some
time
and
the
team.
The
net
interactive
team
has
been
making
a
lot
of
improvements
to
this
feature,
which
is
without
an
interactive
folks.
Who
may
not
be
aware
that
interactive
is
this
global
tool,
global.net
tool
that
you're
able
to
one
of
the
things
that
it
provides?
You
is
with
the
kernel
for
jupyter
notebooks
you're,
able
to
write
c,
sharp,
f-sharp
and
powershell
code
inside
of
jupiter
notebooks.
B
Now
something
else
that
people
may
now
the
web
interface
jupiter
notebook
super
lab.
It
has
certain
shortcomings
and
that's
one
of
the
nicest
about
jupiter
that
it's
open
source
is
accessible,
so
you're
able
to
use
many
different
sort
of
front
ends
right,
so
you
can
use
interact,
and
in
this
case
you
can
one
of
the
things
that
you
can
do
is
you
can
use
visual
studio
code
right
now.
B
In
this
case,
you're
gonna
need
the
insiders
version,
because
this
is
still
a
very
very
it's
it's
a
preview
sort
of
feature
and
it's
it's
it's
under
active
development
and
taking
feedback,
but
one
of
the
really
cool
things
is
as
opposed
to
jupiter
notebooks,
at
least
from
what
I
tried.
You
don't
need
anaconda.
B
A
B
So,
with
this,
what
you
can
do
is
you
can
use
variable
sharing
to
basically
take
the
values.
For
example,
let's
say
you
have
a
list
of
data
points
right
and
you
can
then
pass
those
in
to
javascript
your
perhaps
a
d3.js
or
whatever
it
may
be,
and
and-
and
you
can
get
some
very
rich
resolution-
visualizations
that
maybe
perhaps
some
of
the.net
libraries
may
not
give
you
so
so
it's
it's.
It's
a.
B
Yeah
and
we
actually
have
a
question
from
theorem
geek:
is
there
a
standard
approach
to
the
file
type
now?
Is
it
diverging
towards
ipi
and
b
from
what
I
tried
you
can
use
ip
file.
So
it's
not
that
div
format
anymore,
john,
I'm
not
sure
if
you've
seen
differently,
but
I
think
that
that
ipi
and
v
extension
files
sort
of
work
out
of
the.
A
C
A
So
if
you
you
might
have
joined
us
earlier
in
the
year
for
when
was
that
march,
or
I
can't
remember
what
exactly
it
was,
but
we
had
our
first
virtual
ml.net
community
conference,
it
was
super
successful.
We
had
a
ton
of
people
a
lot
of
engagement.
We
started
off
with
the
workshop
that
louise
and
I
led
and
then
from
there
we
went
into
a
day
full
of
community
sessions
and
yeah.
So
this
is
kind
of
the
the
second
one
virtual
ml.net
conference,
but
hackathon
version.
A
So
we'll
start
off
with
that
workshop
again
and
then
then
we'll
go
into
a
hackathon.
So
this
is
the
repo
for
this
is
where
you'll
actually
submit
your
project
ideas.
So
you
can
create
a
project
by
creating
an
issue
in
the
repo
put
in
the
details
of
that
project
and
then
or
if
you
you
know,
want
to
join
an
existing
project,
whether
you
don't
want
to
come
up
with
your
own
idea,
or
you
see
something
that
looks
really
interesting.
You
can
comment
on
an
existing
issue
and
yeah.
A
We
have
you
know
the
rules
here.
Some
example
data
sets
some
project
ideas,
some
faqs,
the
code
of
conduct.
We
have
a
slack
channel
for
it
too.
So
if
you
haven't
joined
the
virtual
ml.net
slack
community,
yet
you
can
do
that
here
and
then
here's
where
you
would
so
there's
actually
a
signup
form.
So
if
you
click
on
this
there'll
be
a
form
here
with
some
important
dates,
and
this
is
how
you'd
sign
up
and
the
first
50
people
to
sign
up
will
get
an
exclusive
ml.net
t-shirt.
A
So
make
sure
you
get
your
sign
up
soon
and
that
will
be
on
november
13th
or
20th
so
it'll
be.
You
have,
I
think,
a
little
less
than
a
week
or
maybe
about
a
week
yeah.
I
think
a
little
less
than
a
week
to
get
your
submissions
in
and
then
on.
The
20th
is
when
we
will
actually
announce
the
winners
and,
of
course
this
is
going
to
be
all
virtual.
B
Yeah,
so
in
terms
of
ideas
and
types
of
projects,
we
really
aren't.
We
look
forward
to
seeing
what
you
come
up
with
right
and
there's
no
sort
of
you
know
I
guess
standard
in
terms
of
what
we're
expecting.
What
we
do,
though,
expect
is
that
you
use
ml.net
in
some
way,
shape
or
form
right
and
and
the
different
ways
that
you
might
use.
It
is,
for
example,
one
thing
that
you
might
want
to
do
is
contribute
to
the
actual
to
the
core
repo
right.
B
So
you
might
say:
hey,
there's
this
algorithm
or
this
thing
that
I
see
it's
not
there
and
I
think
it's
something
that
will
be
valuable.
Let's
go
ahead
and
do
that
something
does
that
you
might
consider.
Speaking
of
you
know,
brie
mentioned
that
about
sort
of
local
training
scenarios,
perhaps
maybe
adding
like
a
high
level
object.
Detection
api
that
leverages
tensorflow.net's
object,
detection
api
to
to
basically
enable
that
sort
of
local
object,
detection,
training
scenario
that
would
be
really
cool.
B
So
so,
at
the
end
of
the
day,
it
doesn't
matter
what
you
do
as
long
as
it's.
It
includes
demo.net
in
some
way,
shape
or
form,
and
just
as
a
reminder,
it's
got
to
be
a
new
project
right,
so
existing
projects
won't
necessarily
be
considered
for
submission.
You
can
most
certainly
find
people
and
work
with
them
on
existing
projects,
but
they
won't
necessarily
be
considered
for
submission.
Only
new
projects
will
be
considered.
A
Yeah,
actually
so
we're
still
figuring
out,
you
know
final
prizes,
but
we
do
know
that
one
of
them
will
be.
The
winners
will
get
to
come
on
to
the
community,
stand
up
with
us
and
and
talk
about
their
submissions,
so
that
that
will
be
one
of
them.
We'll
have
a
judging
criteria
on
here.
I
don't
know
if
it's
on
here
already
with
a
few
different
categories,
so
there
will
be
multiple
winners
but
yeah
the
winners
will
get
to
come
on
with
us
and,
I
hope
that's
a
good
prize.
A
D
A
A
A
So
if
we
go
here,
you
can
see
that
he
has
a
ton
of
different
videos
like
he
spent
so
much
time
on
this
and
he's
always,
you
know
contributing
he
actually
just
created
a
he
just
submitted
a
sample
or
yeah
a
sample
for
automl
ranking
scenario,
which
is
awesome
so
definitely
take
a
look
at
his
videos.
He
has,
you
know
from
getting
started
to
kind
of
digging
deeper
into
it
like
creating
a
deep
neural
network
model.
A
So
yeah
well,
of
course,
we'll
have
the
link
in
the
url
list,
and
such
someone
asked,
who
is
john
john
is
john.
Is
that
he's
our
guest
today
yeah?
So
john?
Is
there
any
well?
I
guess
we'll
have
you
on
to
talk
about
all
these
later,
so
I
have
one
more
thing
I
want
to
show,
but
first
let
me
go
through
and
make
sure
we
don't.
I
know
that
we
might
have
had
some
questions
that
we
missed
yeah.
B
I
do
so
there's
this
one
from
gregor.
So
is
it
a
problem
if
you
can't
be
there
for
the
first
two
days,
totally
blocked
every
friday
on
saturday,
not
a
problem
at
all,
so
the
first
day
is
going
to
be
a
workshop
and
that
if
your
experience
with
email.net,
it's
not
required
that
you
attend,
but
you
know
it
might
be
more
of
a
refresher
for
new
users.
B
We
encourage
it,
though,
that
way
you
get
a
little
bit
acquainted
with
with
ml.net
and
all
it
has
to
offer
and
yeah
there's
not
a
requirement
that
you
have
to
be
there
sitting.
You
know,
sort
of
so
you
know
attached
to
your
seat
for
all
days
of
the
hackathon.
If
that's
what
you
choose
to
do
for
whatever
number
of
hours
or
days.
That's
that's
your
prop.
You
know
prerogative
but
yeah.
That's
totally!
Okay,
we're
totally
flexible.
As
long
as
you
know,
you
submit
by
the
deadline.
A
Yeah,
so
I
see
another
one
here,
still
developing
a
chat
bot.
I
assume
that's
ml.net.
What
are
the
options
and
not
english
languages
or
translating
the
data
set
is
the
way
yeah.
So
it's
all
going
to
depend
on
your
data
set
that
you
have
you'd
have
to
translate
that
data
set
or
have
a
data
set.
That's
already
in
a
different
language,
louise
or
john
anything
to
add
there.
B
Yeah,
it's
constantly
a
challenge
right
for
for
not
getting
data
sets
in
in
non-english
form.
I
mean,
I
think
you
even
see
it
with,
for
example,
with
alexa
right
and
then
these
sort
of
smart
home
devices
that
are
voice
activated
a
lot
of
times.
They
have
trouble
understanding,
for
example,
things
like
what's
it
called
like
accents
and
things
like
that,
and
a
lot
of
that
comes
from
that.
B
A
So
this
question:
can
I
participate
with
my
friend
or
do
I
have
to
be
solo?
You
can
totally
sign
up
with
a
friend
if
you
want
there's
options.
So
when
you
submit
a
new
project,
it'll
be
a
new
issue
in
the
github
repo
and
you
can
tag
you
know
whether
you're
looking
for
teammates
or
whether
you
don't
want
team.
You
know
outside
teammates,
and
you
can
tag
that
so
you
can
totally
sign
up
with
a
friend
or
friends.
B
A
A
B
No,
I
don't.
I
don't
see
anything
I
mean,
I
guess,
there's
a
question
here
and
just
to
for
folks
who
may
be
new
to
the
you
know
to
the
stream
or
just
channel
that
in
general,
ml.net
is
light
tensorflow.
It
is
not
light
tensorflow,
but
you
can
leverage
tensorflow
via
some
third-party,
not
third-party
but
sort
of
external
libraries
like
tensorflow.net
to
help
build
uh.net
models.
A
And
we
had
someone
comment
that
there's
sunsharp
check
that
out.
It
has
tensorflow
in
it,
so
we
actually
had
high
ping
on
last
week
who
works
on
the
slide,
sharp
team
or,
and
he
it's
an
open
source
framework
for
a
bunch
of
data
science
and
machine
learning,
tools,
4.net
or
specifically,
c
sharp.
So
tensorflow.net
is
one
of
their
projects
that
has
c-sharp
bindings
for
tensorflow,
and
so
tensorflow
is
a
bit
lower
level.
A
I
would
say-
and
we
actually
have-
I
think
we
just
said
this:
we
have
our
image
classification
api,
which
is
actually
on
top
of
tensorflow.net,
for
training,
custom
image,
classification
models
with
a
method
called
transfer
learning.
So,
okay,
so
let
me
switch
over
to
visual
studio
here.
I
just
want
to
show
off
one
thing
and
then
we'll
hand
it
off
to
john.
A
I
right
click
on
my
project
and
add
machine
learning,
as
you
saw
in
our
slash
programmer
humor,
and
so
then
you
see
all
these
different
scenarios
and
I
don't
know
if
we
explained
this
before
and
our
previous
and
maybe
the
first
one
first
stand
up.
We
did
but
these
right
here
these
scenarios
are
all
ones
that
are
supported
by
automl
or
automated
machine
learning.
So
this
means
that
you
choose
these
scenarios,
you
input
your
data
set
and
then
automl
will
iterate
through
different
models
and
choose
the
best
model
for
you.
A
So
you
don't
have
to
deal
with
choosing
the
data
transformations
or
the
algorithms
that
you
want
to
add
to
the
pipeline,
so
it
trains
it
for
you.
So
the
limited
scenarios.
These
are
scenarios
that
are
still
supported
by
ml.net.
You
know
you
can
build
the
pipeline,
you
can
choose
your
algorithms
and
all
done
it.
Has
algorithms
applied
for
these
scenarios,
but
it's
not
yet
supported
by
automated
machine
learning.
So,
for
instance,
if
I
click
on
anomaly
detection,
you
can
see
it
skips
all
the
usual
steps,
and
it
goes
straight
to
this.
A
You
know
consume
the
model
and
what
we
do
is
link
to
our
sample
code
in
in
our
sample
repo
on
github
and
open
on
my
other
screen
here.
So
it
actually
opens
up
to
this
tutorial
where
you
can
download
that
repo
and
get
started
and
then
start
plugging
in
your
own
data,
but
so
just
because
they're
not
supported
by
automl
now
doesn't
mean
that
they
won't
be
in
the
future.
A
A
B
Yeah
now
with
ml.net
ml.net,
actually,
if
you're
training,
image
classification,
custom,
image
classification
models
with
the
image
classification
api,
it
actually
uses
transfer
learning,
so
only
the
last
layer
is
retrained
and
you
don't
actually
build
the
layers
yourself.
If
you
want
to
do
that,
you
can
certainly
check
out
tensorflow.net
and
there
you
have
a
little
bit
more
lower
level
control
over
your
layers
and
your
network.
A
A
C
A
B
Time
series
data-
and
I
think-
and
I
may
be
wrong
in
this-
but
I
think
at
the
moment
it's
only
univariate
right
so
that
what
that
means
is
it
only
allows
one
input
or
one
feature
to
to
train
your
forecasting
model.
I
do
I
believe
that
at
some
point
there
may
have
been
work
on
multivariate,
but
I'm
not
sure
if,
if
that
actually
is,
is
actually
part
of
the
penalty
yet,
but
yeah.
A
B
A
I
will
stop
sharing
my
screen
and
let's
turn
it
over
to
john.
So
john,
do
you
want
to
talk
a
little
bit
about
you
know
all
the
ml.net
stuff
you've
worked
on
and
maybe
why
you
love
ml.net
so
much?
How
did
you
learn
about
it?
That
was
a
lot
of
questions
and
ones,
but
you
know
I
feel
pretty
piglet
to
everyone.
So
I'm
almost
interesting.
C
When
I
was
first
introduced,
I
think
I
was
watching
it
live
as
well
and
me,
and
a
colleague
of
mine
at
the
time
were
pretty
interested
in
the
data
science
and
machine
learning
space
and
we've
been
learning
some
python
stuff
and
actually
how
I
got
into
that
was
f,
sharp,
which
louise
you
probably
would
enjoy
that
I
remember
a
code
camp
I
went
to.
It
was
a
guy,
he
did
on
f
sharp
and
I
always
remember
what
he
said.
C
He
said
you
can
always
go
into
the
the
web
space
or
the
data
space
and
he
taught
how
you
can
do
the
data
space
with
f,
sharp
and
so,
and
I
got
so
intrigued
from
it.
I
started
playing
around
with
it
and
I
saw
this
is
really
fun
actually
enjoyed.
Programming
like
I
did
when
I
first
started,
and
I
messed
with
f
sharp
and
data
so
and
then
at
some
point
I
thought
man
I
need
to.
If
I
want
to
get
serious
with
it.
C
A
A
C
That
announcement,
so
I
kind
of
dug
into
it,
played
around
with
it,
and
I
was
like
you
know:
if
you
really
want
to
get
into
it.
I
want
to
just
help
contribute
kind
of
do
some.
Some
pull
requests
in
there.
So
that's
pretty
much
how
I
got
started
and
for
the
videos
it
was
kind
of
a
thing
I
was
doing
for
work
at
the
time,
just
creating
a
bunch
of
presentations
and
I
was
creating
videos
for
those,
and
you
know
I
was
still
into
the
emma.net
and
I
was
like
man.
C
C
Api
yeah
and,
interestingly,
that
video
is
still
one
of
my
most
popular
videos
but
yep.
I
enjoyed
that
and
enjoyed
the
feedback
and
how
mainly
how
much
it
was
helping
people
from
the
comments
I
was
getting
and
I
was
saying
well
why
don't
I
just
keep
making
videos
with
mmo.net
and
occasionally
do
some
calling
the
services
and
azure
email
in
there
as
well.
So
that's
how
that
got
started
and
in
the
meantime
I
continue
to
look
for
issues
on
the
repo
to
kind
of
help
with
mostly
documentation.
B
Yeah-
and
I
think
that
that's
really,
you
know
interesting
that
you
and
koi
tree,
I
am
not
listening
to
metal,
but
what
I
was
saying
was
that
it's
really
interesting
that
you
mentioned
that
right,
like
that.
You
saw
this
thing
that
you
liked
you
were
passionate
about
and
you
didn't
see
anything
there
for
it,
and
you
were
just
like
well
yeah.
Let
me
just
go
ahead
and
create
something
for
it,
and-
and
I
think
that
that's
something
that
we
encourage.
We
definitely
said
when
we
highlight
these
community
links
right.
B
We
want
folks
in
the
community
to
to
if
they
don't,
if
they
see
that,
there's
a
gap,
if
there's
things
that
are
missing
that
perhaps
are
being
covered
and
that
they
would
like
to
see
as
part
of
their
learning
process
or
whatever
it
may
be.
We
we
definitely
encourage
you
and
we
we
look
forward
to
to
those
types
of
contributions
right.
C
C
C
B
And
for
those
for
those
issues
that
you
mentioned,
that
you
know
you
feel
like
they
may
be
way
over
your
head
like
how?
How
has
your
experience
been
in
terms
of
like
you
know?
Maybe
there
you
need
guidance
and
you
know,
or
you
need
some
input.
It
sounds
like
okay
like
this
is
something
I
think
you
work
on,
but
you
know
I
just
need
a
little
help.
How's
that
experience
been.
C
It's
been
great,
a
few
people
in
particular
eric,
has
been
really
awesome
in
helping
me
on
a
few
issues.
Some
of
them
also
was
like.
I
think
I
know
what
to
do
so
I'll,
get
a
pull
request
and
then
I'll
tag
them
say.
I
think
this
is
right.
Just
give.
Let
me
know
what
I
need
to
change,
or
anything
everybody's
been
like
super
nice
on
the
comments
and
giving
further
guidance
on.
C
B
And
there's
also
I'll
add
one
more
just
to
throw
it
in
there
shameless
plug.
We
also
accept
contributions
on
docs,
so
the
docs
are
open
source
as
well
at
it's
a
dotnet
stocks
repo,
and
we
also
encourage,
if
you
see,
even
if
it's
just
a
typo,
please
just
you're
only
helping
the
next
person
by
making
the
documentation
better.
So
yeah
we'd
certainly
accept
that
and
if
there's
any
questions,
I'd
be
more
than
happy
to
sort
of
you
know,
guide
you
and
making
contributions
so.
B
B
A
C
C
C
If
that's
possible
again,
it's
it's
one
of
those
things
that
it
is
evolving
and
there's
just
nothing
really
good
out
there.
Yet
it's
it's
probably
not
the
time
for
that
in
in
inman.net,
because
I'm
I'm
sure
people
will
want
some
other
features
like
like
the
neural
networks.
The
thing
that
we
just
got
asked
earlier.
B
I
believe
it's
the
first
hundred
rows
or
in
a
data
data
view
before
after
or
whatever
transformation
should
perform
to
it,
that's
kind
of
a
step
in
that
direction,
right
towards
some
sort
of
developability
of
transformations
and
what
exactly
is
going
on
in
my
model
and
what
these
transformations
are
doing
but
yeah.
I
agree
that
it's
still
a
long
way
there
and
even
the
preview
that
one's
not
really
recommended
that
you
use
it
in
production
right,
it's
it's
more
of
like
hey!
It's
there.
A
D
C
A
B
B
A
We
got
a
question
about
what's
the
proper
hashtag
to
use
for
on
twitter
for
mlmet,
I
went
back
and
forth
about
a
year
ago
about
what
it
should
be.
I
believe
mlnet
is
the
one
that
we've
kind
of
decided
on
mlnet
and
that's
where
I
see
the
most
the
most
stuff.
So
that's
a
great
question,
though,
because
I
also
question
that
sometimes
but
hashtag
about
that
is
the
one
that
I
usually
use,
I
think
also
uses
for
sure
yeah
but
yeah.
That's
a
great
question.
A
A
Yeah,
that's
the
other
one
I've
seen
but
yeah
john's
also
on
twitter,
so
that's
his
twitter
handle
there
he's
always.
I
always
see
him
answering
questions
people
in
the
community
and
you
know
he
tweeted
out
about
the.
I
love
him
on
that
shirt
a
while
back,
and
so,
if
you
want
to
follow
him,
that's
his
twitter
handle
like
swipe
up
to
subscribe.
All
the
all
that
good.
A
B
B
I
guess
you
tend
to
have
more
ammo
dynamic
videos,
but
you
also
cover
other
technologies
right.
I
think
you've
done
azure
machine
learning.
You
also
done
cognitive
services.
In
your
experience
and
again
coming
from
the
standpoint
where
it's
like
hey,
you
know,
I
didn't
really
have.
I
guess
a
machine
learning
background.
You
know,
I
guess,
for
somebody,
that's
getting
started
right,
what
sort
of
approach
or
or
how
do
you
see?
Those?
Maybe
you
know,
be
different,
but
at
the
same
time
sort
of
you
know
working
together
nicely.
B
No,
no
sorry,
I
meant
in
terms
of
solutions
right
so
like
if
I'm
looking
at
hey,
I
want
to
do
ml.net.
Sorry,
I
don't.
I
want
to
do
machine
learning.
I
don't
have
a
lot
of
background.
Should
I
use
content
services?
Should
I
use
azure
ml?
Should
I
use
xml.net
like
and
and
if
I
go
down
one
path
right?
Does
that
mean
that
I
then
can't
sort
of
you
know
use
the
other
solutions.
C
Well,
you
can
certainly
go
toward
different
solutions.
I
think
each
of
these
are
pretty.
They
have
a
lot
of
documentation
on
all
of
them
where
you
can
get
started
with
them
pretty
easily,
but
in
terms
of
which
one
to
kind
of
start
using
for
one
thing,
I
would
say,
depends
on
what
your
end
goal
is.
Is
your
end
goal
to
create
an
application
and
get
something
published
or
deployed,
or
is
your
end
goal
to
kind
of
tink
around
with
something
learn
more
about
machine
learning
and
some
of
the
steps?
C
B
Yeah,
no,
I
I
think
that
certainly
makes
sense
and
yeah
depending
on
your
needs
right.
You
can.
You
can
certainly
mix,
mix
and
match,
like
in
the
example
of
like
in
the
example
of
azure
machine
learning
right
so,
for
example,
the
azure
training
enable
training
scenarios
through
the
tooling
they
leverage,
azure,
machine
learning,
right
and
and
with
interoperability
tools
like
onyx
you're,
able
to
basically
sort
of
cross
that
boundary
a
bit
and
then
be
able
to.
B
So
it's
really
nice
that
it's
not
you're,
not
just
coming
at
it
from
you
know,
from
sort
of
a
it's
like,
hey,
you
know
done
it
is
all
I've
tried
and
that's
sort
of
you
know.
You
know,
definitely
try
it
out
with
that
net,
but
certainly
that
you've
sort
of
tried
the
spectrum
right
and
yeah,
and
there.
C
D
A
C
C
A
A
You
know
and
usually
they'll
use,
automl
or
tooling,
and
then
there's
kind
of
another
group
of
people
who
they
want
to
learn
more
about
the
process
and
have
a
bit
more
control
and
actually
learn
a
little
bit
more
about
the
mechanics
of
machine
learning
as
they're
they're
building
the
model.
I
would
love
to
know
you
know
from
you
all
that
are
watching
feel
free
to
comment
like
what
group
you
kind
of
fall
into.
B
You
are
more
than
welcome
to
build
tooling
right,
so
we
have
the-
I
guess,
official
right,
first
party
tooling,
if
you
will
in
model
builder
and
cli
right,
but
as
john
mentioned
right,
there's
tooling
that
perhaps
there's
other
tooling
that
you
may
find
is
missing
enamel.net
and
we
certainly
welcome
those
types
of
contributions.
You
know
a
week,
maybe
I
guess
a
bit
of
a
short
time,
but
at
least
like
a
prototype
for
poc.
That,
certainly
you
know,
would
be
welcome.
B
C
C
B
Yeah,
but
it's
really
cool
and
and
that
you
know
it
literally
started
from
a
tweet,
and
I
think
that
that's
essentially
what
we
would
like
to
see
at
this
hackathon
right.
It's
just
a
bunch
of
people
like
hey
here's,
this
crazy
idea.
Let's
try
and
I'm
sure
that
there's
plenty
of
people
who
are
more
than
willing
to
pitch
in
and
don't
really
think
it's
a
crazy
idea,
and
I
think
that
that
it's
really
valuable
and
would
be
more
than
happy
to
collaborate
and
contribute.
B
Yeah,
the
first
conference
we
had
about,
I
think
at
the
highest
point
in
viewership.
It
was,
I
think,
about
500,
so
it
was
a
pretty
good
turnout.
Considering
that
I
think
we
had
roughly
700
signups
so
yeah
I
mean
for
it
being
the
first
one
and
again
you
know
the
magnetic
community
is
certainly
growing,
but
you
know
rel,
relatively
speaking,
right.
It's
still
fairly
fairly
small
and
that's
okay
right.
You
know
mediums
like
this.
B
Also,
making
the
product
and
and
everything
better
right,
so
yeah.
We
certainly
hope
that
you
folks
are
are,
you
know,
engaged
and
excited
for
this
hackathon,
and
we
really
we
are
just.
We
have
no
idea
what
to
expect,
and
I
don't
say
that
we
don't
have
expectations.
But
what
I
mean
is
by
that.
It's
not
that
we
think
that
the
submissions
are
not
there's
no
submission,
but
that
you
know
we
have
no
expectations
in
terms
of.
A
A
Exactly
then,
of
course,
we'll
continue
to
promote
it,
and
you
know
you'll
be
sick
of
hearing
about
it
after
a
while,
but
that's
okay,
let
me
see
if
there's
any
last
questions
gregor
said,
I
have
an
idea,
but
I'm
not
100
sure
that
it's
technically
doable
good
enough
for
submission
yeah
totally,
so
we
had
actually
talked
about.
You
know
having
a
a
prize
for
what
was
it
like
most
ambitious
or
you
know
something
similar.
A
A
B
A
Yeah,
some
friends
here
he
actually
reached
out
a
while
ago.
I
think
last
year
about
he
created
a
ml.model
builder
extension
for
vs
code,
which
is
awesome.
So
it's
built
on
the
cli
yeah
and
it
looks
like
there's
a
new
version
out
or
he's
announcing.
Maybe
he
didn't
announce
before
I
can't
remember,
and
yeah
has
added
compatibility
for
codespaces
in
the
new
cli
friends.
I'm
sure
we'll
have
you
on
at
some
point
to
talk
about
it
because
you,
you
know
you
can
tell
us
a
lot
more
about
it.
A
B
That's
awesome
yeah,
so
I
did
notice
that
time
flew
and
we're
actually
hit
the
top
of
the
hour
so
yeah.
So
I
mean
we
definitely
want
to
thank
john,
not
only
for
coming
on
today,
but
for
all
the
contributions,
whether
that's
you
know,
organizing
community
events,
the
videos,
of
course,
actually
making
contributions
to
the
to
the
core
repo
right.
The
list
goes
on,
so
we're
very
grateful.