►
From YouTube: 1. Welcome
Description
From the NERSC NVIDIA RAPIDS Workshop on April 14, 2020. Please see https://www.nersc.gov/users/training/events/rapids-hackathon/ for all course materials.
A
So,
of
course,
this
is
a
training
event
for
learning
to
use
the
Nvidia
rapids
ecosystem
on
on
GPUs,
and
the
idea
is
that
we
want
to
help
you
get
ready
for
being
able
to
do
data
analytics
on
perlmutter
when
it
arrives
later
this
year,
all
right
so
moving
forward
all
right.
I'll
start
again,
thanks
for
zooming
in,
we
are
originally
hoping
to
have
this
as
an
on-site
training
with
remote
attendants
possible,
but
the
remote
attendants
of
course
became
mandatory.
We
want
to
thank
everybody
for
continuing
to
be
interested
in
in
the
training
to
join
us.
A
This
way
it
did.
We
did
do
a
lot
of
logistical
preparation
beforehand,
but
a
bunch
of
that
just
had
to
be
cancelled.
This
also
had
implications
for
how
we're
gonna
do
the
training
I'll
get
into
that
a
little
bit
later,
but
again,
I.
Just
wanted
to
thank
everybody
for
putting
up
with
this
again.
I
just
want
to
remind
everybody
to
stay
muted
if
possible.
That
way,
we
avoid
echoes
and
and
other
kind
of
things,
I.
Think
probably
by
this
point,
people
are
probably
pretty
good
at
that
Zoo
America.
A
This
event,
of
course,
is
being
recorded.
It
should
be
showing
up
at
some
point
on
the
nurse
web
site
soon
on
the
web
site
for
the
for
the
training
event.
I'm,
not
I'm,
not
gonna,
say
it's
gonna
happen
this
week,
maybe,
but
it
might,
it
should
probably
be
happening
at
some
point.
I
mean
next
week.
I
would
say
the
slides
that
we're
gonna
be
presenting
today
are
on
the
nurse
website
on
the
training
event
page.
So
you
can
look
at
the
download
them
review
them
later
on.
A
A
So
if
you
had
problems
using
Jupiter
notebooks
on
the
shared
GPU
node
instance
just
immediately
prior
or
this
morning
to
this
event
to
calling
in
it's
probably
because
we
were
taking
most
of
the
GPUs
and
you're
waiting
outside
the
reservation
and
the
scheduler
was
kind
of
just
saying
sorry,
you
can't
run
a
job.
Yet
if
it's,
if
it
winds
up
having
sustained
problems,
you
know
I'll
be
debugging
knows
but
I,
don't
think
that'll
be
too
difficult
for
me
to
resolve,
while
other
people
are
doing
presentations.
A
So
definitely,
if
you
can't
run
the
notebooks
right
away,
don't
panic
in
the
afternoon
I
think
everything
will
probably
be
working.
Just
fine
and
you'll
be
able
to
do
it
then,
and
afterwards,
for
questions
and
comments.
I
think
the
best
thing
for
us
to
do
right
here
is
to
use
the
chat
window
instead
of
like
all
a
meeting
unless
I
would
say
unless
it
was
the
request
of
the
speaker,
but
let's
listen.
A
If
you
default
to
using
chat
window
for
passing
questions
and
Laurie
or
I
will
be
keeping
an
eye
on
that
I
think
that's
just
about
it,
everything
usually
for
an
in-person
event.
We
have
something
like
you
know,
earthquakes
and
how
to
recycle
and
where
the
bathrooms
are
and
how
to
escape
the
building.
You
don't
have
to
do
any
of
that.
Presumably
you
know
how
all
of
that
works
that
your
your
place,
but
I
can
say
you
know
just
stay
relaxed
in
your
bathroom
or
whatever
with
coffee.
A
A
So
I
just
wanted
to
remind
everybody,
of
course,
everybody's
here,
because
they
know
this
that
promarker,
the
next
production
system
that
we're
gonna
be
running
at
nurse
is
gonna,
be
arriving
later
started
wrapping
later
this
year
and
it's
gonna
be
completely
delivered.
Sometime
during
next
year,
it's
going
to
have
at
least
three
times
the
computational
power
of
Cori.
It's
going
to
be
a
heterogeneous
CPU,
node,
GPU,
node
architecture.
It's
going
to
bring
together,
AMD
CPU
base
nodes
and
NVIDIA
based
GPU
nodes,
and
we're
very
excited
about.
A
This
look
has
a
new
internet
from
Cray
which
is
Ethernet
compatible
and
is
supposed
to
be
very
good
at
supporting
data.
Centric
workflows
and
we're
gonna
have
an
all
flash
platform,
integrated
storage.
So
there's
a
lot
of
really
cool
things
that
are
coming
in
Perlmutter
and,
of
course,
we're
very
excited
about
all
of
those
things.
A
A
So
to
that
end,
we've
been
hosting
a
number
of
training
events.
Workshops
hackathons,
we're
gonna
continue
doing
this
throughout
this
year
next
year.
Other
laboratories
have
been
posting
these
as
well.
A
lot
of
these
are
around
applications,
not
really
workflows,
but
applications
that
are
written
and
compiled
languages
like
C,
C++
or
Fortran,
but
we
didn't
want
Python
users
to
be
left
out
and
you
wanted
them
to
have
some
of
the
fun
too
and
I
actually
think
they're.
A
Probably
they
have
kind
of
maybe
more
fun,
because
Rapids
is
a
really
powerful
library
upon
which
you
can
build
a
lot
of
stuff.
So
it's
hopefully
it's
going
to
make.
You
feel
productive
and
powerful.
At
the
same
time,
when
you
do
something
they
arrive,
why
the
focus
on
Python
I,
don't
know
that
I
really
need
to
say
this
to
you
because
I
mean
it's.
It's
obviously
part
of
why
you're
here,
but
I
mean
it's
an
integral
programming
language.
That's
something
that
we.
A
We
kind
of
perceive
that
it's
necessary
to
close
the
gap
between
the
software
needs
of
view
of
nurse
users
and
current
state-of-the-art
GPO,
analytics
libraries,
and
so
that's
why
we
have
a
partnership
with
Nvidia,
but
here
I've
got.
You
know,
other
other
statistics
to
just
mention,
which
is
that
I
mean
basically
every
month,
there's
over
a
thousand
unique
non
staff,
Python
users
and
the
risk,
and
that's
a
fairly
sizable
fraction
of
everybody
that
logs
in
are
they
all
using
high-performance
libraries
for
everything
that
they're
doing
in
Python?
A
Probably
not,
but
a
good
fraction
of
those
are
a
quarter
of
all
projects
that
run
jobs
and
nurse.
So
this
is
on
the
compute
nodes
are
using
Python
quarter
of
jobs,
they've
been
preparing
Python
in
some
form
or
fashion,
but
so
it's
obviously
an
important
part
of
our
workload
and
that's
why
why
we're
having
this
training
training
event?
Why
we
have
this
engagement
is
with
nvidia.
I.
A
Think
I
missed
the
slide
yeah,
so
I
just
wanted
to
say
a
word
about
our
engagement
with
nvidia
this
as
part
of
the
non-recurring
engineering
contract.
That
nurse
has
was
nvidia
to
help
to
help
users
of
python
get
ready
for
problem
letter,
as
I
said,
we're
trying
to
close
the
gap
between
what
our
users
need
and
what
is
available
in
their
data
and
like
ecosystem.
A
So
part
of
this
has
involved
finding
workflows
that
run
on
the
CPU
architectures
at
nursed
from
real
nurse
users,
some
from
inside
the
nurse
centers
of
some
staff
workflows,
but
also
crucially,
nurse
user,
workflows,
sharing
those
lifting
media
engineers
in
order
to
help
inform
product
development
and
get
feedback
from
the
experts
on
how
users
can
optimize
their
GPU
utilization.
So
those
have
been
very
productive
for
us,
and
this
hackathon
is
part
of
that
effort,
and
we
hope
that
you
find
it
a
useful
experience.
A
Okay,
so
here's
today's
agenda.
It
is
a
number
of
talks,
but
there
are
a
number
of
exercises
as
well.
However,
due
to
the
remote
nature
of
this
they've
been
turned
into
flipped
classroom
exercises.
So
if
you
had
an
opportunity
to
try
the
notebooks
ahead
of
time,
that's
great
if
you
haven't
been
able
to
do
it,
maybe
you
have
a
chance
to
do
today
when
the
reservation
is
working
or
after
we're
it's
up
to
April
20th
and
then,
of
course,
when
promoter
arrives,
you
can
of
course
try
those
notebooks
later
try
your
own
workflows.
A
Obviously,
what
we
hope
is
that
you
can
look
at
these
notebooks
and
learn
and
kind
of
finally
get
a
sense
of
like
okay,
I
kind
of
see
a
way
that
I
should
be
thinking
about
using
the
GPUs
and
what
the
tools
are
in
Python.
That
can
help
me
use
the
GPUs
and
how
to
talk
to
them
and
how
it's
different
from
or
similar
to
the
kind
of
workflows
that
you're
used
to
on
CPUs
they're
going
to
be
three
breaks
today.
So
there's
a
couple
of
15-minute
breaks
and
then
there'll
be
a
break
for
lunch.
A
You
know
during
that
time.
Let's,
let's
all
you
know,
stretch
and
do
things
we
need
to
do
during
that
time.
I
would
also
ask
that
people
be
respectful
of
our
presenters,
because
some
of
them
are
in
many
different
time
zones
and,
like
some
of
them
are
a
few
hours
ahead
of
us.
So
it's
gonna
be
seven
o'clock
by
the
time
we
get
done
here.
So
let's
try
to
keep
on
schedule.
Of
course,
the
the
part
of
that
is
going
to
be
on
the
presenters
but
I.
A
You
know
as
an
audience
if,
if
it
makes
sense
to
question
and
turn
it
into
an
email
question
or
something
like
that,
then
we
can
help
to
route
those
later
if
it
turns
into
a
longer
discussion,
also
as
I
mentioned
before,
there's
homework
that
was
shared
with
you.
I
know
that
a
number
of
users
were
able
to
get
on
with
the
GPUs
in
the
past
several
days
and
try
out
the
notebooks.
A
Today's
presenters
are
coming
to
us
from
Nvidia
they're
from
the
rapids
team
and
also
a
Solutions
Architect,
so
the
presenters
are
IU
SH
and
V
boo
and
Nick
from
the
rapids
team
and
Zahra
Rowe
Nagi
who's,
a
Solutions
Architect
as
our
as
a
former
nurse
postdoc.
Actually
so
it's
nice
to
have
her
back
and
presenting
as
well
I
wanted
to
take
a
minute
to
especially
thank
our
presenters
from
Nvidia,
so
much
for
being
able
to
deal
with
difficulties
around
a
remote,
remote
training
event.
A
We
felt
that
it
was
important
to
try
to
hold
this
at
the
at
the
time
that
we
originally
scheduled
it
rather
than
pushing
it
off
to
later
in
the
year,
because
we
wanted
to
give
you
the
users
time
to
become
familiar
with
the
tools
well
in
advance
of
the
arrival
of
the
hardware.
But
again
you
know,
I
know
that
people
are
in
different
time
zones
and
I.
You
know
I
hope
we're
not
messing
up
people's
schedules
too
bad
for
this,
but
I
did
want
to
again.
A
Thank
you
so
much
for
for
a
green
participate
and
make
your
presentations,
but
also
for
the
hard
work
that
you've
done
with
the
workflows
that
we've
shared
with
you
a
few
further
Thanks
I
want
to
thank
Laurie,
who
has
done
a
lot
of
the
setup
and
testing
of
notebooks
and
making
sure
that
things
are
happening
at
the
right
time
and
watching
the
waiting
room
for
me
right
now,
and
all
of
that.
So
thanks
a
lot
Laurie
and
also
Celeste
for
administrative
support
and,
of
course
you
who
are
remote
and
socially
distanced.
A
This
last
slide
here
contains
a
number
is
which
will
be
directly
useful
during
the
training
event
today.
But
we
include
them
here
kind
of
mostly
for
reference
for
later
on
when
you're
looking
at
the
signs.
So
with
that
I
think
that
stop
sharing
my
screen
and
then
what
we'll
do
is
we'll
move
to
the
first
presentation.