►
From YouTube: Machine Learning Community Standup - ML.NET in VS Code
Description
On this week's community standup, ML.NET Community member Franz Silva is joining us to discuss his ML.NET Model Builder extension for VS Code and how you can contribute to the tool.
Community Links: https://www.theurlist.com/mlnet-standup-2020-10-21
Featuring: Bri Achtman (@briacht), Luis Quintanilla (@ljquintanilla), Jake Radzikowski, Franz Silva (@FranzSSilva)
#ML.NET #VSCode
A
A
A
C
A
C
B
D
A
A
A
So
this
one
here
is
just
kind
of
like
an
introduction
of
how
you
can
do
email.net
sorry,
how
you
do
option
protection
with
with
model
builder
right.
This
was
something
that
was
fairly
recently
announced
and
and
released
inside
a
model
builder.
So
you
can
go
ahead
and
you
know
basically
use
a
model
builder
to
build
an
object,
detection
model
that
you
can.
Then
you
know
deploy
to
your
net
applications
while
he
was
sort
of
going
through
the
exercise.
A
You
will
take
advantage
of
all
these
scenarios,
all
those
scenarios
in
sort
of
the
the
command
line.
I
think
I'm
seeing
some
comments
in
the
chat
that
there's
a
feedback
loop,
not
sure
what's
going
on
is,
is
everybody
anybody
else
hearing
that
all
right?
Let
us
know
in
the
chat,
if
your
audio,
if
the
audio
sounds
a
bit
messed
up,
okay,
so
the
audio
somebody
says
the
audio
is
good.
Okay,
so
let's
keep
going
maybe
kung
fury
keyboard.
A
Maybe
you
might
want
to
try
logging
off
and
loading
back
into
the
stream
okay.
So
we
have
john
here
who
put
out
an
update
for
his
deep
learning
neural
network
video
right
so
make
sure
to
check
that
out,
I'm
kind
of
going
along
the
line.
So
in
this
case
he
uses
the
the
api
right
to
build
this.
So
you
can
take
a
look
at
that.
We
also
have
a
few
videos
that
visual
ranjan
and
I
apologize.
A
So
if
you
would
like
to
help
shape
the
future,
let
me
make
this
a
little
bit
bigger.
It
might
be
hard
to
see
if
you
would
like
to
help
shape
the
future
of
mlaps
or
machine
learning
operations,
which
is
kind
of
like
devops.
The
closest
thing
you
think
of
it
is
like
devops
for
machine
learning
and
what
that
looks
like
within
the
net
ecosystem.
Please
go
ahead
and
provide
your
feedback
just
take
a
couple
minutes
and
we'd
love
to
hear
from
you.
A
What
are
you
doing,
what
you're
doing
in
terms
of
devops
practices,
even
if
you're
not
doing
ml,
ops
or
using
sort
of
mlaps
in
your
current
workflows?
It's
interesting
to
hear
what
sort
of
devops
tools
you
folks
are
using
that
way.
You
know
we
can
make
that
sort
of
experience
a
little
bit
more
seamless
and
better
so
make
sure
to
let
us
know.
A
We
also
have
this.
This
is
something
that
we
talked
about
a
few
weeks
ago
on
one
of
the
community
standups,
which
is
that
there
was
some
work
happening
to
basically
upgrade
test
floor.net.
It
was
the
previous
version
or
originally
ml.net
was
leveraged
to
tensorflow.net,
which
is
a
project
from
the
cyshar
team
who
we
had
on
a
previous
stream
as
well,
but
it
was
using
version
one
point
x
of
tensorflow.
A
B
Yeah
sure
I
can
talk
about
a
little
bit,
the
the
big
reason
why
we
wanted
to
move
to
tensorflow.
The
latest
version
is
that
there
were
some
issues
with
installing
cuda,
specifically
in
like
the
drivers,
so
there
was
just
it
was
kind
of
difficult
to
actually
get
the
the
gpu
training
up
and
running
without
having
any
sort
of
like
driver
issues
on
your
machines.
B
Unfortunately,
the
the
cuda
version
that
was
for
the
old
tensorflow
was
no
longer
being
supported
by
upgrade
and
we're
excited
that
you
know
this
brings
us
both
to
the
latest
version
of
tensorflow.
So
you
get
all
the
latest
features
from
tensorflow,
but
also
making
it
so
that
it's
just
you
know
easier
to
use
it
from
from
model
builder.
Hopefully
we're
still
working
on.
You
know
integrating
it
into
our
tooling,
but
hopefully,
in
the
next
release,
we'll
be
able
to
give
you
an
easier
set
of
instructions
for
getting
started
with
gpu
training.
B
A
Awesome
yeah.
I
know
that
there's
been
a
lot
of
work
being
done
there
with
direct
l
and
being
able
to
accelerate
right
these.
These
training
jobs-
and
you
know
you-
you
folks,
on
the
tooling
side,
released
the
local
gpu
training
scenario
right.
So
I
can
always
assume
that
that
particular
scenario
as
a
result
of
this
will
will
improve
right
or
we'll
get
better
and
yeah
cool
awesome,
so
still
related
to
open
source.
A
A
C
A
A
So
if
you're
looking
for
speakers,
let's
say
that
you
want
to
zoom
in
here
a
little
bit
make
it
a
bit
easier
to
see,
there's
this
directory,
that
you
can
basically
find
speakers
if
you're
looking
for
particular
topics
right,
you
can
do
a
filter
here
for
specific
for
specific
topics
that
you
would
like
to
to
learn
about,
and
if
you
are
a
speaker,
uh.net
speaker,
right
or
or
a
speak
on
topicsrelated2.net
and
any
of
these
things
here.
You
can
also
add
yourself
to
this
list.
A
So
again,
it's
it's
a
soft
launch
and-
and
the
team
is
kind
of
like
reviewing
these
sort
of
you
know
they're
taking
some
time,
they're
still
trying
to
work
out
the
process,
but
just
it's
nice
that
this
is
there,
and
you
know
for
folks.
You
know,
with
virtual
meetups
and
and
sort
of
mediums
like
this
right,
like
these
streams,
it's
really
nice
that
it's
it
provides
a
solid
or
way
consolidated
way
to
find
folks
who
are
passionate
about
speaking
on
these
topics.
A
Events
there
is.net
conf,
the
agenda
is
out
and
there's
tons
of
really
great
content.
That's
happening
november,
12th
through
10th
through
the
12th
and
there's
tons
of
really
good
stuff.
But
really
we
know
why
we're
here
all
right.
So
there's
two
talks.
I
believe
on
email.net
one
it
brie
is
gonna,
be
delivering
that
talk
and
the
other
one,
I
believe,
is
by
veronica
all
right.
We're
going
to
talk
about
shouldn't
talk
about
ml.net.
C
A
And
xamarin
in
terms
of
for
for
breeze
talk,
I
think
one
of
the
the
things
that
would
be
really
nice
kind
of
like
extending
sort
of
this
format
that
we
have
here,
which
is
really
about
showcasing,
what's
happening
in
machine
learning
space
specifically
within
ml.net.
If
you
are
a
net
developer
or
a
developer,
who
is
using
ml.net
in
production,
we'd
love
to
hear
from
you,
so
you
know
twitter
email.
A
C
Yeah,
no,
I
just
decided
to
do
something
a
little
bit
different
this
time
around.
You
know
I've
done
a
ton
of
intro
workshops
or
intro
talks,
which
are
great,
but
this
time
I
wanted
to
showcase.
You
know
how
people
are
using
ml.net
in
production
or,
if
you're
you
know
not
yet
in
production
and
but
have
a
proof
of
concept
or
about
to
go
in
production
and
want
to
share
that
with
the
world
onto
your
experience.
C
So
just
let
me
know
we
can
get
you
into
my
talk,
I'm
on
day,
one
of
dot
net
conf
I
mean
I
was
also
just
laughing
because
someone
asked
if,
in
the
comments,
do
you
use
a
potato
as
a
mic?
So
I
don't
know
if
that's
directed
at
you
or
me,
I
had
some
audio
problems
so
baby,
but
not
that
I'm
aware
of
it
now
I
kind
of
want
to
make
a
potato
mic
for
the
next
time,
but
yeah.
That's
that's
all
about
the
my
talk
for
nightcomp.
A
The
types
of
things
that
I
might
be
able
to
do
with
uh.net,
and
then
you
can
sort
of
you
know,
run
with
it
and
and
essentially
create
a
project
right
and
compete
in
this.
So
the
13th
it
starts
with
the
workshop
you
it's
not
mandatory
right.
So
if
you
are
experiencing
more.net
you're
more
than
welcome
to
join,
but
you
can
also
just
start
hacking
on
it
right,
it's
gonna
be
the
13th
through
the
18th.
Your
final
submissions
are
due
on
the
18th.
A
C
A
First,
50
people,
the
shirt
right
so
get
a
shirt
sign
up,
and
on
top
of
that,
once
you
once
you've
gone
ahead
and
signed
up,
you
can
go
ahead
and
create
a
project
right
and
that's
it's
really
pretty
easy.
You
essentially
create
a
project
by
opening
an
issue
here
in
the
repository
which
I
can't
log
into,
because
I
am
not
loved
until
github,
but
essentially
you'll
be
able
to
open
up
a
or
create
a
project
and
basically
have
a
detailed
description
of
it.
C
A
Provide
some
information
about
it
here
we
have,
we
have
brett's
here,
let's
save
it,
for
example
right
and
here's
just
a
in
terms
of
name.
Perhaps
you
might
include
you,
you
can
work
on
this
in
both
as
an
individual
right.
So
if
you
want
to
work
on
this
by
yourself,
you're
certainly
more
than
welcome
to,
but
we
highly
encourage
you
to
form
teams
around
to
try
and
sort
of
continue
building
that
community
get
to
work
with
new
folks
and
and
yeah.
So
you
provide
a
brief
description.
A
You
also
provide
information
about
whether
you
are
looking
for
teammates
and
also
if
you
would
like
to
mentor
for
your
team
and
essentially
what
that
mean.
What
that
means
is
if,
if
you
want
a
little
bit
of
guidance
in
terms
of
like
hey,
is
this
something
that's
feasible?
What
do
you
think
like
that?
Obviously
we
wouldn't
you
know
the
mentors,
wouldn't
necessarily
go
ahead
and
directly
be
involved
in
the
project,
but
at
least
they
can
provide,
and
those
folks
who
have
experience
with
level.net.
A
A
And
then,
last
but
not
least,
this
is
sometime
in
the
future,
but
this
is
the
applied
sharp
challenge,
which
I
believe
the
first
one
was
last
year
and
essentially
what
this
is.
It's
almost
like
a
hackathon,
it's
a
somewhat
longer
event
that
takes
place
and
you
get
to
basically
submit
projects
or
create
projects
around
different
areas,
one
of
which
includes
f,
sharp
sorry,
which
one
of
which
includes
machine
learning
and
data
science.
Of
course,
there's
these
other
topics,
but
if
you're
interested
in
pursuing
participating.
A
I
believe
that
there's
going
to
be
more
information
on
this
sometime
in
the
future
at
the
moment,
they're
looking
for
judges
right
in
each
of
these
categories.
So
if
you
would
like
to
be
a
judge
in
one
of
these
categories,
make
sure
to
reach
out,
I
think
the
information
is
on
here
on
how
you
can
get
involved
as
a
judge
and
again
in
terms
of
being
a
participant.
A
I
believe
that
there's
gonna
be
more
information,
as
you
know,
as
time
goes
on,
and
the
judges
are
all
sort
of
selected
for
this.
Let's
see
here
and
then.
Finally,
if
you
are
not
already
on
discord-
and
you
would
like
to
continue
the
conversation-
perhaps
you
your
questions-
weren't
answered
on
this
on
the
stream
make
sure
to
join
the
net
evolution.
There
is
a
machine
learning
channel
where
you
can
discuss
all
things:
machine
learning.
A
A
D
D
Okay.
Well,
I
think
introductions
with
vaude
dunham.
So
once
again,
I'm
I'm
a
full-time
cto
at
teravision.
I
work
with
closed
and
open
source
technologies
and
I'm
a
full
stack
developer.
Also,
I
used
to
be
a
dot
net
developer
years
ago,
but
lately
slowly,
I've
been
transitioning
to
javascript
based
projects
and
node.js,
but
I've
never
lost
my
love,
for
you
know
microsoft,
technologies
and
now
that
it's
all
open
source,
even
even
more
so
to
explain
a
bit
what
the
extension
actually
does
in
visual
studio.
D
D
The
idea
is
to
offer
a
faster
iteration
to
to
actually
using
the
cli
and
well
one
of
the
reasons-
and
I
already
mentioned
one-
was
to
increase
developer
productivity
and
also
to
have
like
an
alternative
to
the
visual
builder,
the
model
builder
in
in
visual
studio
2019,
which
is
awesome
that
extension,
but
in
linux
or
mac,
or
even
people
that
just
used
lightweights
lightweight
code
editors.
It's.
D
They
have
a
an
alternative
to
be
able
to
do
it
themselves
and,
of
course,
I've
always
wanted
to
contribute
something
to
open
source,
because
I
just
love
how
open
source
works
in
general
and
contributions
so
the
whole
the
idea
behind
this
was
well.
Let
me
let
me
contribute
something,
and
I
can
actually
you
know,
do
and
help
with.
E
A
D
Okay,
so
so
mainly
the
only
things
you
need
to
be
able
to
use
the
actual
extension
are
to
have
the
sdk
installed.
It
can
be
net
core
31
and
up
and
dot
net
five.
It's
working
on.net
five.
I
tried
that
this
morning
on
the
rc2
version-
and
you
know
also
need
the
ml
mlnet
cli,
which
you
can
install
using
the
global
tool
for
net.
Just
installing
with
this
command,
you
can
get
the
mlnet
cli
installed.
D
It
so
what
we're
going
to
do
here
is
a
quick
test
so
running
from
the
command
palette
for
the
ones
that
use
visual
studio
code.
You
can
press
ctrl
shift
p
to
open
the
command
palette,
and
then
you
you
can
find
this
ml.net
model
builder
in
in
the
commands,
and
the
first
thing
it's
going
to
ask
you,
for
is
what
type
of
what
type
of
order
train
you
want
to
use.
That's
from
the
cli.
D
D
It
will
ask
you
if
you
want
to
ignore
any
columns
like
if
you
don't
want
to
use
every
single
column
in
your
data
set,
you
could
just
use
the
ones
you
could
just
click
out
the
ones
you
don't
need
and
after
that
it
asks
for
the
training
time.
In
this
case,
we're
just
going
to
put
10
for
demo
purposes
and
the
last
set.
D
The
last
step
is
just
going
to
say:
where
do
you
want
to
output
the
actual
code,
the
the
train
model,
in
this
case
I'm
going
to
put
in
this
folder
output,
and
we
could
just
let
it
run
you
just
minimize
this
here.
I
think
there
we
go
so
just
like
in
the
cli
or
what
it
will
do.
It
will
run
the
cli
on
the
on
the
on
a
child
process
and
then
send
back
whatever
information
the
cli
is
already
using.
D
D
D
Okay,
so
what
other
things
so
right
now
you
can
do
binary
classification
and
regression
using
the
actual
extension
in
visual
studio
code,
the
future
things
that
I'm
planning
on
adding
and
is
to
actually
have
a
bit
more
feature:
parity
with
the
actual
cli.
So,
like
any
of
the
commands
you
can
do
on
the
cli,
you
can
do
them
from
the
actual
extension.
D
So
that's
adding
image
classification
and
recommendation
to
the
to
the
to
the
extension,
a
more
intuitive
ui.
This
is
something
I
was
planning
I
was
discussing
with
jake,
also
like
implementing
a
step
wizard
where
you
can
go
back
and
forward
on
the
steps,
because,
right
now,
it's
just
like
one
flow
and-
and
you
really
don't
know
how
many
steps
you've
got
left
to
be
able
to
to
to
finish
the
command
and
the
other
one
is
probably
implementing
web
views
where
we
could
actually
create
a
nice
ui
similar
to
the
one.
D
D
The
other
thing
I'm
working
on
is
a
config
file
so
that
all
those
steps
that
you're
running
right
now,
you
could
just
have
them
already
like
pre-defined
in
a
config
file
and
also
you
could
just
actually
say
model
builder
run,
and
then
it
will
take
whatever
files
you
have
in
your
configuration
file
and
run
it
for
you
and
export
it
and
do
everything
for
you.
So
it's
not
you
don't
have
to
go
through
all
these
steps.
Every
time
you
need
to
generate
a
a
train
model.
B
D
D
You
know
all
the
methods
to
consumer
and
everything,
something
that
can
be
just
nice,
quick
and
easy
to
play
around
with
the
with
the
actual
train
model,
which
would
this
is
gonna,
be
very
interesting.
We
I
still
have
to
investigate
how
to
do
it,
but
it's
it's
coming.
It's
coming
for
the
future,
so
ways
you
can
contribute
to
the
project.
D
I
still
don't
have
like
guidelines
and
things
like
that
in
the
repo
I'm
planning
on
you
know
creating
guidelines
and
things
like
that
in
the
actual
github
repository,
but
mainly
that
the
ways
that
you
can
actually
contribute
is
test
it
I
mean,
please
use
it
in
your
daily
projects
report
any
issues
you
have,
I
mean
right
now.
It
reads
from
a
csv
file,
a
tab,
separate
tab,
separate
separated
file,
but
I
don't
know.
D
Maybe
I
do
have
a
mistake
somewhere
and
it's
you
know
some
large
data
set
is
not
it
won't
read
it
correctly.
So
please
please,
try
it
out
and
of
course
you
can
in
the
same
github
repo.
You
can
request
any
new
features.
You'd
like
to
see
like
you
can
say:
oh
you
can
do
this
better
this
way,
or
maybe
I'd
like
to
see
some
specific
feature
added.
D
Well,
we
can
we
can
in
general,
we
can
say
if
the
cli
does
it,
then
probably
we
can
add
it
to
the
the
extension
and,
if
not,
then
we
can
even
escalate
it
even
more
and
probably
open
up
a
issue
on
the
machine
learning
repo
and
see
if
we
can
have
it
added
to
the
actual
cli.
Also
and
pull
requests
are
welcome.
So
if
you
see
any
thing
that
you
can
help
on
the
actual
source
code,
you
can
you're
more
than
welcome
to
create
a
pull
request
right
now,
I'll
see.
D
B
B
We
had
a
question
in
the
chat,
so
I
just
want
to
talk
about
it
a
little
bit
as
far
as
so
yeah,
whether
or
not
you
have
to
implement
the
algorithms
yourself,
and
the
answer
is
no,
so
this
is
running
using
the
the
cli
and
the
cli
is
based
on
our
our
auto
ml
solution
that
we
also
use
in
model
builder,
and
what
we
do
is
we
there's
there's
several
algorithms
within
the
within
classification
and
we
we
iterate
over
them.
So
using
our
automl.
B
B
B
So
if
you,
if
you
are
knowledgeable
enough
to,
you,
know,
go
and
you
know
fine-tune
the
algorithm
yourself
or
you
want
to
choose
a
specific
algorithm,
because
you
know
it'll
be
a
good
fit
for
for
your
data.
You
can
you
can,
you
know,
start
with
start
with
this
extension
in
nvs
code
or
start
with
the
extension
and
model
builder
to
kind
of
get
a
working
pipeline
and
then
from
there
sort
of
fine
tune.
D
D
B
Cool
so
yeah,
you
can
see
here
it's
already
gone
through
three
different
algorithms,
and
if
you
give
it
more
time,
you
would
see
that
maybe
these
three
algorithms
pop
up
more
with
different
sort
of
parameters
and
fine-tuning
things
or
or
potentially
more
algorithms,
I'm
trying
to
remember
how
many
there
are
for
classification
and
not
oml,
but
oh
and
one
one
thing
to
note
too.
I
think
you,
you
said
in
your
presentation
that
it
was
binary
classification,
but
I
think
this
if,
if
it's
using
the
cli,
it's
actually
multi-class
classification,
so
it'll
it'll
train.
D
D
Is
that
with
me?
Yes
yeah,
I
probably
will
yes
nice,
so
yeah,
here's,
the
actual
github
repository
you
can.
You
can
clone
it
from
here.
If
you
want
and
then
submit
a
pull
request,
whatever,
with
whatever
changes
you
need,
if
it's
to
actually
install
the
the
extension
it's
freely,
you
know
you
could
just
search
for
the
extension
on
the
on
the
extension
store
on
visual
studio
code.
A
A
D
D
Yeah
I've
built
extensions
but
they're,
not
public,
they're,
more
personal
and
for
the
actual
inside
of
the
company
they're
more
like
auto
autocomplete
for
projects,
for
example,
we
have
internal
projects
where
we
have
a
lot
of
code.
We
need
to
rewrite
every
single
time,
so
we
have
like
code
snippets
and
things
like
that
for
specific
projects,
so
we
actually
create
our
own
extensions,
sometimes
to
to
improve
developer
productivity.
A
A
D
I
was
a
donna
developer.
I
do
not
actually
do
net
development
nowadays,
mainly
because
most
of
the
projects
that
we
work
on
are
you
know:
node.js
python,
golang
and
any
other
languages,
but
it's
more
like
we're.
Looking
for
the
opportunity
to
work
with
also
uh.net
open
source-
and
I
think
it's
something
like
we're
just
looking
like
for
the
right
moment
to
implement
something
using.net,
at
least
in
in
the
company
by
myself-
it's
I've
always
I
have
never
stopped
learning.net.
D
I've
always
kept
up
to
date
with
whatever
new
technologies
that
they're
launching
you
know
it's
like
every
single
dot
net
conf,
I'm
always
like
number
one
guy
to
listen
to
scott
hanselman.
You
know
show
whatever
demo
he
has
for
net
and
things
like
that.
So
yeah
I
mean
that's
kind
of
me
in
in
that
sense,
and
of
course
I
I
I
did,
I
do
sometimes
workshops.
I
did
a
workshop
recently,
I'm
I'm
actually
even
here
in
ecuador.
D
Well,
it
was
mostly
because
I'm
always
looking
like
for
ways
to
solve
different
problems.
It
can
be
for
client
projects,
so
it
could
be
personal
projects.
One
thing
was,
like
you
know,
doing
image.
Classification
was
one
was
kind
of
the
things
I
was
looking
into.
In
that
moment
I
was
actually
using
tensorflow
and
python,
but
I
noticed
afterwards
that
you
know
there
was
a
version
for
dotnet.
It
was
a
lot
easier
to
use.
D
And
that's
kind
of
it's
more
testing
than
you
know
like
actual
production
use,
but
it's
it's
got
so
much
potential.
You
know
just
to
be
able
to
use
net
for
everything
I
mean
it's.
It
just
makes
a
lot
easier
and
that's
kind
of
how
I
found
ml.net.
I
started
with
you,
know,
learning
with
tensorflow,
and
then
I
noticed
that
microsoft
was
investing
a
lot
in
you
know,
machine
learning
as
well
and
that
ml.net
existed.
So
that's
when
I
started
to
test
it.
A
Cool
nice,
so
so
taking
a
little
a
bit
of
a
step
back
there.
What
was
your
experience,
because
you
mentioned
that
you
started
with
tensorflow
and
we're
using
python?
So
you
know
what
was
your
experience
with
machine
learning
prior
to?
I
guess
you
know
starting
to
use
these
tools
and
what
sort
of
projects
were
you
were
you
building?
You
mentioned
animation
classification,
but
I
mean
if
it's
okay,
to
speak
more
about
it
like
what
sort
of
problems
were
you
trying
to
solve.
D
Yeah,
it
was
mostly,
it
was
a
the
problem
that
we
were
looking
to
solve
was
to
find,
if
someone
not
to
give
up
too
much
about
the
project,
but
it
was
to
give.
It
was
mainly
because
we
needed
to
know
if
people
were
using
the
correct
equipment
on
a
let's
say,
a
construction
site.
So
we
were
looking
for
ways
to
say.
Okay,
so
is
this?
Is
this
labor
using
a
hat?
Is
he
using
his
safety
jacket?
D
Is
he
using
these
kind
of
things
and
that's
when
we
had
started
to
like
investigate?
How
was
this
done
and
that's
kind
of
the?
The
reason
why
I
went
into
image
classification
is
to
you
know
like
to
we
started
with
a
whole
bunch
of
pictures
like
saying
this
is
with
a
hat
on
half
hat
and
then
started
to
train
all
these
models
to
be
able
to
identify
those
those
things.
So
that's
that's
the
reason
why
I
got
into
the
image
classification
and
at
the
same
time
you
know,
there's
you
know.
D
There's
always
one
thing
I
say
like:
there's
always
a
way
that
in
dot
net
you
can
do
it
better,
and
it's
not
just
like
saying
I
love.net
and
everything.net
it's
just
it's
just
true.
You
know
in
the
sense
that
it's
you
know
you
want
to
do
image
classification
using
tensorflow
or
whatever
other
technologies.
D
Okay,
you
can
do
it.
You
can
build
your
entire
like
python
file
and
you
can,
you
know,
find
the
the
correct
commands
to
install
and
then,
if
you're,
using
a
gpu,
you
can
go
through
all
the
installing
the
specific
drivers
and
the
cuda
drivers
and
everything
you
need
to
be
able
to
run
them
on
your
own
machine.
If
you
have
a
gpu
for
it,
but
then
afterwards,
you
know
you
find
solutions
like
ml.net,
which
is
oh
just.
You've
got
the
cli
with
three
lines
of
code
or
just
one
line
of
code.
D
C
So
we
don't
currently
have
anything
like
that
in
our
tooling,
but
that's
actually
something
that
we're
working
on
and
something
we
want
to
add,
especially
with
integrating
with
the
notebooks
experience,
not
necessarily
for
live
training,
but
just
more
model
explainability
in
general
and
and
more
visualizations.
So
you
can,
you
know,
track
your
model
or
see
you
know
accuracy
and
all
the
different
metrics
that
you
have
and
be
able
to
see.
Explainability.
C
We
currently
have
two
methods
of
explainability,
pfi
or
permutation,
feature
importance
and
feature
contribution,
calculations,
and
so
those
are
part
of
the
api
right
now
we're
still
working
on
making
those
a
bit
more
user-friendly
and
then
eventually
integrating
them
into
our
tooling
as
well,
but
yeah,
anything
to
add
there
from
jake
or
luke.
Louise.
B
C
We
have
some
other
questions
in
the
chat.
Let's
see
someone
asked
what
happened
to
microsoft,
virtual
academy,
it
used
to
be
a
good
video
collection
and
resource
for
learning.
I
don't
know
the
answer
to
what
happened
to
it.
I
do
have
some
alternatives.
Louise,
maybe
you
know,
because
you
work
on
documentation.
A
C
John
wood,
well,
I
was
recognizing.
He
has
great
videos
I'll
post
it
in
the
chat,
the
the
link
he
was
on.
I
think
a
lot
two
weeks
ago.
He
has
a
ton.
A
whole
playlist
for
email.net
has
almost
60
videos
from
beginner.
You
know
getting
started
to
to
more
advanced
topics,
and
I
think
I
don't
know
if
you
mentioned
this
earlier-
louise,
but
as
part
of
the
hackathon
the
first
day
we're
actually
going
to
have
our
intro
workshop.
B
A
Oh
no,
I
think
I
froze
all
right
because
you
guys
can
still
see
see
me
screen
right,
yeah
yeah.
So
that's
a
really
good
question
and
going
back
before
I
get
into
this
to
a
point,
jdr
and
txru
microsoft,
virtual
academy
is
equal
to
emerson,
learn
now
yeah
I
mean
you
can
find
a
lot
of
learning.
Content
in
microsoft
learn
all
right.
So
that's
a
really
great
platform.
You
can
also
get
certification.
A
A
So
yeah,
there's
tons
of
stuff
that
you
can
check
out
for
that,
but
yeah,
okay,
so
themes
getting
into
the
hackathon,
and
so
there's
a
few
resources
here
that
can
might
maybe
help
you
get
started
right.
So
one
of
the
things
that
I
mentioned,
I
I
talked
about
creating
a
new
project
right,
but
what
happens
if
you
see,
for
example,
you
look
through
these
projects
and
the
issues
right
and
you're.
Like
you
know,
I
don't
really
either.
A
A
I
believe
it's
only
him
and
he's
still
not
sure
whether
he's
looking
for
teammates,
but
if
it's
something
that
you'd
like
to
contribute
to
or
you'd
like
to
work
on,
you
can
just
add
a
message
here
right
in
in
the
issue,
and
you
can
just
say
you
know
I'm
interested
in
contributing
and
and
then
you
can.
You
know
just
sort
of
update
this
issue
to
reflect
who
who
your
team
is
and
who
forms
part
of
that
team
right.
A
A
B
A
There
is-
and
there
isn't
so-
this
is
another
resource,
that's
also
in
the
repository.
You
can
look
at
it
project,
ideas
and
things,
so
we
don't
have
categories
like,
for
example,
like
hey,
you
know
you,
here's
a
price
for
tooling
here's
a
price
for
documentation,
and
we
don't
have
it
broken
out
like
that
right.
C
A
For
example,
we
talked
about
that
pfi
or
or
permutation
feature
importance
and
feature
contribution,
calculator
right.
Those
are
two
interpretability
techniques
that
are
currently
available
inside
of
email.net,
as
we
mentioned,
there's
still
some
work
being
done
to
make
those
a
little
bit
more
user-friendly
right
and
there's
other
interpretability
techniques
beyond
those
right.
So
perhaps
you
may
be
interested
in
like
in
in
implementing
a
new
interpreter,
real
technique
or
even
contributing
to
the
existing
ones
right.
A
Something
else
you
might
want
to
add
basically
contribute
extension
to
that
interactive
if
you're
not
familiar
with
net
interactive.
It
is
this
one
of
the
things
that
it
provides
you
with
it
allows
you
to
work
with
jupiter,
notebooks
and
and
use
dotnet
code
inside
of
jupyter
notebooks
right.
So
perhaps
you
might
add
an
extension
such
as
I
know,
one
of
you
folks
on
the
chat
was
asking
hey:
are
there
visualizations
right
for
these
training
jobs?
So
that
may
be
something
that
you
might
be
interested
in,
adding
as
an
extension
right?
A
Something
else
is
contributing
to
the
tensorflow.net
right.
So
that's
that's
not
necessarily
that's
something
that
again
it's
used
as
part
of
ml.net
right.
We
just
talked
about
it,
how
it
was
upgraded,
so
0.20
to
version
0.20
and
and
perhaps
something
that
you
might
be
interested
in
doing
there
is
tensorflow.
Has
this
object
detection
api
right?
A
So
currently,
if
you
want
to
do
object,
detection
number
one
for
training,
you
have
to
do
it
into
tooling
right
and
number
two
currently,
because
it
is
a
resource,
intensive
sort
of
workload
it
it
leverages
azure
right,
but
perhaps
there
may
be
a
way
that
you
might
be
able
to
implement
the
tensorflow
object,
detection,
api,
internship,
flow.net
and
then
that
could
eventually
get
consumed
by
ml.net
right.
So
those
are
just
some
ideas
of
things
that
you
might
want
to
consider.
A
A
Here
the
only
requirement
like
strict
requirement,
there's
actually
a
few,
but
one
is
that
it
has
to
be
a
new
project
right.
So
as
awesome
as
you
know,
the
model
builder
extension
previous
code
is
unfortunately
you
know
it's
not
a
new
project,
it's
a
16
project,
but
you
can
work
on
essentially,
like
let's
say
I
don't
know,
adding
visualizations
to
to
the
you
know,
model
builder
extension
or
ml.net
extension
and
vs
code
right.
So
that
would
count
right.
A
So,
although,
although
maybe
not
you
know
doing
something
like
that,
but
but
adding
on
to
this
existing
project,
that's
out
there.
That
would
certainly
count
as
a
new
project
right
and
then
the
other
thing
is,
you
must
use
sample.net
in
some
way,
shape
or
form
right.
So
so,
as
long
as
those
two
things
are
sort
of
satisfied
and,
of
course,
that
you
agree
to
to
to
some
of
these
things
here
that
that
should
be
enough
right
to
get
you
started
so
hopefully
that
kind
of
gives
yeah.
A
B
I
can
talk
on
that
a
little
bit,
so
I
don't
actually
know
what
the
roc
m
is,
but
I
believe
we're
gonna
kind
of
increase
our
compatibility
with
gpus
using
direct
direct
ml.
I'm
not
sure
if
it'll
have
support
for
that,
I
can
kind
of
figure
out
offline,
but
that's
supposed
to
anything
that,
with
the
way
that
the
direct
ml
is
sort
of
architected,
anything
that
directx
gets
sort
of
like
gpu
optimization,
for
I
believe
that
the
direct
ml
should
be
getting
some
sort
of
performance
boost
from
it.
B
C
C
A
Yeah,
we
actually
have
another
question
here
from
s
ready,
it's
kind
of
unrelated
talking
about
authentication
and
project
templates
with
entity
framework
so
on
this
particular
stand-up
we
cover
you
know.
We
cover
machine
learning,
topics
right,
so
anything
there's
some
somewhat
of
a
bit
of
a
focus
on
ml.net.
A
But
again
we
thought
we
bring
on
community
members
to
to
come
talk
and
show
off
their
awesome
projects
like
like
france
right,
but
if
you're
interested
in
cc
framework,
there
is
a
stand
up
for
entity
framework,
I
believe
which
is
gonna.
Take
it
takes
place
every
other
wednesday
whenever
we're
not
on
right.
B
And
then
one
thing
to
to
add:
if,
when
you're
working
on
hackathon
projects
or
just
contributing
in
general,
we
kind
of
mentioned
the
discord
before
but
feel
free
to
join
there.
If
you
have
questions
and
you
need
some,
you
want
me
to
try
to
help
connect
you
with
experts.
Inside
of
microsoft.
I
can
try
to
you,
know,
get
some
knowledge
from
inside
of
microsoft.
B
If
you
need
it,
one
project
that
I'm
kind
of
excited
about
I'm
going
to
actually
maybe
add
a
share
to
your
luis
for
the
the
hackathon
is
there's
there's
things
that
are
sort
of
ml.net
adjacent,
one
of
them
being
like
data
frame.
So
the
data
frame
is
useful
in
again,
I'm
not
actually
an
expert
on
dataframe,
but
I'll.
Just
talk
about
a
tiny.
C
B
C
This
great
question:
is
it
still
worth
trying
the
hackathon
even
with
very
little
experience?
Yes,
definitely
like
we
mentioned
earlier,
there's
a
ton
of
resources
to
get
started
and
start
learning.
We're
always
here
for
support.
If
you
have
questions
and
then
actually
to
kick
off
the
hackathon
we're
going
to
be
doing
our
intro
to
ml
workshop,
I
think
it's
like
a.
I
don't
know
if
we're
gonna
do
a
three
hour,
six
hour
version,
probably
somewhere
in
between.
C
And
then,
let's
see
someone
else,
if
donut
has
any
role
in
machine
learning?
Yes,
it
does.
If
this
is
your
first
time
tuning
in,
we,
you
know,
are
the
machine
learning
with
net
stand-ups.
So
ml.net
is
the
the
main
product
we
work
on,
which
is
a
machine
learning
framework
for
net
developers.
I'll
put
up
the
url
list
too,
that
will
have
a
lot
of
the
resources,
and
so
you
can
get
started
there.
C
C
D
C
A
Yeah
absolutely,
but
I
do
see
that
we're
getting
to
the
top
of
the
hour
so
yeah
we
see
a
vote
there
in
the
chat
for
anomaly
since
it's
anomaly:
detection.
Okay,
so
that's
that's
really
great
to
see
but
yeah
we're
getting
to
the
top
of
the
hour.
We
just
want
to
thank
franz
franz.
We
believe
we
shared
the
twitter
other
than
that.
Is
there
something
else
or
a
better
way
to
get
in
contact
with
you
or
is
twitter
twitter?
Just
fine?
It's
just
fine!
A
All
right
awesome,
all
right,
so
you
know
sign
up
for
the
hackathon.
You
know
like
subscribe
all
the
things
first,
50
people
to
sign
up
for
the
hackathon
get
get
a
free
t-shirt
and
next
time,
next
time
it's
gonna
be
really
exciting.
We
have
don
time
joining
us
and
he's
gonna
be
talking
about
automatic
differentiation,
so
make
sure
to
tune
in
for
that
on
november,
4th.