►
From YouTube: SimPEG meeting Oct 30
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).
A
Well,
oh
well,
thanks
everybody
for
making
the
time
for
this
there's
a
few
so
like!
Well,
maybe
just
go
over
like
what
we
want
to
chat
about
today.
I
was
hoping
to
make
some
more
progress
on
parallel,
but
didn't
get
a
chance
to,
but
I
think
John
did
some
playing
around
there.
So
it'd
be
great
to
hear
how
that's
gone
and
just
like
what
your
take
is
on
moving
forward
with
paralyzing
stuff
in
Tsim
bei
yeah.
B
A
A
And
it's
er
what
our
game
plan
is
for
that
and
on
a
related
note,
there
is
a
graduate
student
at
Colorado
School
of
Mines
who's
been
working
on
visualization
stuff,
and
so
it's
I
just
would
like
to
like
bring
that
to
people's
attention.
Sio2
a
quick
little
demo
in
one
of
the
floor
requests
so
I'll
pull
that
up
and
that's
also
tied
to
the
Python
t7
conversation,
because
right
now,
vtk
only
sports,
Python
2.7.
They
are
moving
and
they're
supposed
to
be
on
3.6
by
the
summer,
but
they're
not
there
yet
so
yeah
thats
related.
A
A
C
C
So
just
to
create
the
tiles
currently
before
I
was
using
multi-processing,
but
then
Lindsay
showed
me
how
to
use
the
ask
and
stuff,
and
that
seems
to
work
well,
so
the
next
step
will
be
to
do
a
second
level,
so
here
I'm,
creating
problems
in
a
loop
right
and
then
each
in
each
problems.
I
will
be
also
computing
sensitivities
in
the
problem,
and
so
the
next
one
will
beat
you
to
add
the
advaithic
here.
Instead
of
multi-processing,
you
just
loop
with
desk
right
in
here.
A
C
So
all
you
need
to
do
that.
Xnz
is
a
little
fun
function
that
you
want
to
loop
over
multiple
times.
So
you
could
predict
her
right
right
around
that,
like
this,
it's
a
bit
like
the
pool
air
socket,
so
you
just
just
creating
a
creating,
basically
less
or
that's
a
dictionary
of
this
function
that
has
the
decorator.
So
this
is
the
same
one
here
all
right,
I'm,
just
calling
this
guy
multiple
times,
creating
a
dictionary,
oh
they'll
and
then.
A
The
video
when
quick,
bring
to
people's
attention
here-
and
this
is
something
we'll
have
to
just
explore
a
bit
so
last
week's
doggy
presented
tasks
and
showed
the
futures
approach,
and
so
the
futures
approach
actually
competing,
as
things
are
asked
for,
whereas
the
gasps
dot
delayed.
So
if
you
see
the
way
the
dogs
done,
this
line
73
through
76,
nothing
actually
happens
in
those
lines
that
nothing's
computed.
It
is
just
like
setting
up,
basically
all
of
the
infrastructure
to
perform
the
computation.
A
The
doused
builds
the
graph,
they
call
it,
and
this
is
something
that
we'll
have
to
look
into
a
bit
more,
but
basically
fiddles
how
it's
going
to
perform
the
computation
and
then
in
line
78
when
gassed-up
is
called
that's
when
everything
is
actually
run,
and
so
I
think
that
this
is
going
to
end
up
being
important
when
we
do
multiple
levels
of
parallelization,
because
then
it
will
go
through
and
actually
create
the
graph
first
and
then
execute
it
rather
than
executing
it
as
it
goes.
Yeah.
D
C
My
hope
is
that
so
you
see
right
now,
like
a
in
the
in
the
problem
and
I'm
assigning
different
number
of
workers
right,
and
my
hope
is
that,
if
we
use
desk
all
the
way
through,
is
that
we'll
just
be
able
to
allocate
the
maximum
number
of
workers
on
at
all
levels
and
they
will
figure
out
how
to
optimize
it
right.
So
we
don't
need
to
worry
about
over
over
allocating
processors.
That
makes
sense.
C
D
A
A
C
C
I,
don't
have
64
threads
right,
so
it's
kind
of
it's
slowing
and
slowing
it
down
because
I'm
it
starts
to
conflict.
So
in
order
to
be
able
to
max
it
doodle
max
my
resources
properly,
and
we
need
to
do
like
a
four
here.
Let's
say
on
a
like
a
multiple
of
eight,
so
I
have
to
manually
manually
change
that
right.
C
E
C
A
I
know
where
he
is
I
mean
this
is
a
good
place
to
play
and
first
just
build
an
understanding
of
what's
going
on
before
we
start
implementing
it
everywhere.
It's
nice
to,
hopefully
at
least
get
figured
out
that
two
levels
of
parallelization
and
once
we
have
that
I
think
we'll
be
in
much
better
shape.
C
D
Think,
like
sounds
I
should
really
go.
Look.
How
DeVito
does
so
I
think
that
it's,
you
think,
mask
it's
nice
and
very
abstract
level,
but
in
very
low-level
paralyzation
like
for
you,
the
sensitivity
computation,
for
instance,
it
seems
like
people
are
using
like
kind
of
low-level
like
paralyzation
like
MP,
odd
and
then
use
the
desk
at
very,
like
a
top-end
to
run
the
whole
bunch
of
them.
So
well,
I'm,
not
not
sure
we're
not
there.
Yet
probably
way
to
go.
Is.
C
F
D
Mean
like
it
depends
because,
like
if
your
run,
let's
say
it
takes
about
10
seconds
or
20
seconds
in
it.
Okay,
that's
where
like
task
is
really
powerful
like
it's
easy
to
paralyze
and
it
use
multiple
force,
but
I
think
that
there's
other
level
like
if
you
really
is
the
code
like
each
brother,
cakes,
sometimes
like
no
and
no
second
and
overhead,
don't
ask
I
think
it's
about
that.
I
guess
I'm,
not
exactly
sure
what
was
the
number.
D
A
I
mean
I
think
for
the
first
class
of
problems
were
looking
at
like
what
Dom
showed
with
computing
sensitivity
like
the
overhead
of
computing.
The
sensitivity
I
think
is
gonna
still
be
high
enough
that
you
know
the
overhead
of
gas
doesn't
really
make
a
huge
difference,
but
there
absolutely
is
a
point
where
that
could
be
the
case.
But
if
we
at
least
first
I
think
like
gask
is
a
nice
place
to
just
get
a
handle
on
how
it
works
and
then.
F
A
A
All
right
so
then,
what
I
would
like
to
show
is
a
few
conversations
with
respect
to
Python,
g7
and
some
stuff.
So
yeah
I
see
that
screen.
A
Perfect,
okay,
so
looking
at
a
lot
of
it
started
with
teeters
pull
request
actually,
and
so
there
were
just
some
upstream
dependency
issues
with
testing
on
Python
to
seven
and
sort
of
looking
at
it.
At
that
point
in
time,
I
was
going
back
and
forth
with
dieter
and
he
was
finally
just
like
you
know:
it's
not
worth
spending
time
trying
to
figure
this
out.
Nobody
supporting
Python
do
so
anymore,
drop
it
and
we'll
move
on
and
I
was
like
that's
a
very
fair
point.
A
So
it
started
the
pull
request
to
drop
support
for
Python
to
seven
and
so
right
now
it's
failing
because
I
merged
some
stuff
that
conflicts,
but
that's
okay!
Well,
we'll
get
there,
and
so
some
of
the
reasons
to
think
about
its
dropping
to
seven,
both
numpy
and
sci-fi
are
dropping
support,
I
think
at
the
end
of
the
year.
Oh
they'll
do
so
no.
A
And
so
yeah
so
I
think
like
this
is
a
good
point
to
start
having
that
conversation,
and
we
can
do
something
similar.
So
we
can
push
bug
fixes
down
to
previous
versions
if
they
come
up,
but
I
think
that
yeah,
we
should
basically
start
there's
a
lot
of
good
stuff.
That's
been
done
in
Python
36
that
if
you
want
backwards
compatibility
you
just
can't
take
advantage
of,
and
so
I
think
we
should
just
pull
the
plug
soon.
A
So
the
only
reason
to
not
do
that
exceptionally
rapidly
is
because
so
vain
Sullivan
here
he
is
that
Clara
school
Minds
dimension
and
he's
been
working
on
visualization
and
so
he's
done.
A
lot
of
nice
work
with
pear
review
and
right
now.
I'm
here
view
of
course
requires
BTK,
which
requires
python
to
seven
and
so
I
should
show
you.
A
He
started
at
the
request
with
creating
vtk
objects
from
discretize
meshes,
and
so
one
of
the
that
he
sent
along
just
this
morning
that
I
was
quite
excited
about
it,
so
that
he's
created
a
curvy
mesh
still
just
a
very
basic.
He
grabs
the
example
grid
through
some
data
on
their
wrote
it
to
a
BTK
file,
and
then
this
is
what
it
looks
like.
A
A
So
he's
done,
he
does
what's
it
called
geo,
biz,
yeah
and
so.
B
A
Will
so
I
think
he's
done
a
lot?
I
need
to
take
a
better
look
and
I
would
encourage
everybody
else
to
as
well
at
what
is
in
the
geophys.
We
should
get
the
mean
token
geofence,
I,
think
and
so
PVG.
Oh
sorry,
so
I
believe
that
he's
actually
got
some
really
interesting
stuff
hooked
up
to
the
like
to
Python
code
into
the
notebook.
So
you
could
potentially
drive
what
you're
viewing
in
paraview
from
the
notebook
and
so
to
have
that
communication
layer.
A
A
A
book
on
that
kid
VG
is
bottleneck,
yeah
yeah,
but
they
are
working
on
it
and
so
yeah.
So
we'll
get
Bane's
pull
request
in
before
we
do
the
Python
3,
merge
I,
don't
think,
there's
like
a
huge
rush
on
getting
the
Python
3
going,
but
it
needs
to
be
done
soon,
so
yeah
just
putting
that
on
people's
radars
and
making
sure
that
it's
not
going
to
cause
any
major
headaches.
D
E
A
A
C
A
A
D
D
D
A
B
D
A
A
A
There's
a
couple
things
I
would
like
to
chat
about
and
get
people's
opinions
on.
So
Deeter
added
just
some
comments
to
the
big
picture
and
like
in
our
documentation
and
so
I
would
appreciate
if
a
few
other
people
just
took
a
look
and
Dinn
I.
Think
it's
a
nice
addition.
But
of
course
this
is
sort
of
changing
some
of
the
scope
and
wording,
and
all
of
that
so
take
a
look,
give
a
thumbs
up
if
you're
happy
with
it.
E
A
E
Contact
we're
not
able
to
call
something
and
I
just
added
the
option
to
the
club
Mississippi
of
the
save
directive.
So
your
motion,
it
was
before,
where
I
think
you
were
floating
5
d,
5m
and
now
you
can,
you
can
just
say:
okay
block
5
d,
5
5
s
and
also
like
something
will
not
be
returning,
which
was
with
the.
E
E
A
Okay,
actually,
and
do
you
have
an
example
that
makes
use
of
that
or
would
you
be
able
to
add
a
second.
A
The
other
one
that
I
would
appreciate
some
input
on
here:
I've
added
some
derivatives
for
DC,
so
I
added
electric
fields
and
currents.
I
also
had
a
charge
density,
200
the
fields
of
oh
yeah
and
then
for
the
directives.
I'll
show
you
this
because
this
is
I,
think
something
nice
and
something
to
think
about
when
you're
writing
out
or
you're
creating
a
directive.
A
That's
gonna,
write
stuff
out,
try
and
do
something
like
this,
so
that
you
can
set
the
directory,
because
it's
really
annoying
just
like
dumping
everything
that
you're
saving
into
your
current
working
directory.
So
then,
here
you
can
save
it
to
somewhere
else
and
then
similarly,
like
setting
the
name,
it's
just
helpful
to
be
able
to
do
so
that
lets
you
now
save
the
inversion
model
and
outputs
in
different
places.
If
you
want
so,
if
you
already
run
one
run,
all
of
that
is
now
possible.
A
One
more
quick
thing:
this
sort
of
concerns
I,
think
cute
knee
probably
the
most,
but
I
would
like
to
just
bring
it
up
to
everybody.
So
the
simulation
stuff
I
started
playing
around
with
some
ideas
a
while
ago,
but
the
code
base
is
like
moved
beyond
where
that
floor
request
is
so.
I
would
actually
like
to
like
abandon
this
pull
request
and
start
fresh
I.
A
A
A
A
A
D
F
Yeah
so
by
the
by
didn't,
really
want
to
interrupt,
but
I
was
just
wondering
if
there
were,
you
know
other
communities
and
surely
they've
grappled
with
you
know:
multi
the
multi-level
paralyzation
just
find
out
what
they
did
or
maybe
have
a
template
or
just
help
us
from
rediscovering
some
wheels.
Yeah.
A
That's
a
good
call,
I
think
the
pan
geo
group
has
thought
a
lot
about
paralyzation.
It's
more.
F
A
F
Don't
know
any
that
computational
people
in
mystics
or
in
astronomy
and
people
who
might
actually
be
working
on
these
higher-level
problems.
I
know,
Doug
Scott
was
one
person
who
might
be,
but
I
think
the
physicists
are
using
Python
a
lot
and
yeah
it
just
might
be
worthwhile
trying
to
connect
yeah
I
I
can
take
a
look
and
see
go
through
the
physics,
roster
who's
actually
working
on
a
large-scale
numerical
problems
or
computation.
A
F
D
A
B
G
G
G
Look
at
the
data
file.
We
have
like
a
column
for
each
one
like
a
reject
column,
it'll
say
either
accept
or
reject.
So
the
code
just
goes
through
our
database
and
yeah
outputs,
the
simpe
survey
class
as
just
the
accepted
ones,
but.
D
D
That
yeah
so
I
mean
some
functionality
like
at
least
like
indices,
that
I
can
pull
out
or
what,
if
I'm,
not
hundred
percent
sure
so,
but
if
that's
actually
different,
then
there's
some
I'm.
Sorry
because,
like
the
main
thing
I
need
to
convert,
for
instance,
your
apparent
charge
ability
to
a
voltage,
for
instance,
okay,
so
I
mean,
if
I
don't
have
that
information
about
like
geometric
factors
or
apparent
resistivity.
Although
it
is
that's
null,
though,
I
cannot
convert
that
into
a
format
that
likes
impact
was
invert,
so
yeah
I
think
we
still
need
that.
G
G
D
G
Pushed
anything
to
simpang
processing
yet
just
because
I've
been
playing
around
or
the
last
little
bit
here,
but
I
finally
have
deconstructed
it
into
similar.
How
we
do
with
ye
are
with
the
DC
stuff.
I
just
have
a
nice
kernel
that
you'll
just
make
an
MT
when
doing
kernel,
and
that
take
your
time
series
and
just
do
a
simple
syntax
of
a
multiplication.
Yeah
I
just
wrap
that
up
this
weekend
and
I
was
gonna
push
that
here
later
click
hey.
G
D
G
A
D
A
D
Yeah
next
week,
I'm
going
to
get
together
with
the
guy
named
matheus
bunker,
so
this
guy
is
actually
doing
an
IP
modeling,
whether
it's
not
like
our
sort
of
like
IP
inversion
and
simulation.
It's
very
like
a
microscopic
scale.
Let's
say:
you've
got
a
single,
green
and
then
electrolyte,
okay,
what's
pottery
actually
simulate,
so
you
need
to
input
as
a
concentration
of
ions
and
need
to
set
up
a
boundary
condition.
D
So
he
actually
did
that
did
a
lot
of
work
on
that,
but
he
used
to
come
Saul,
which
is
the
commercial
software,
so
it's
very
hard
to
open
to
like
the
public
or
their
community.
So
yeah
I
had
a
conversation
who
I
can
rewrite
something
very
simple
within
sim
pack,
so
he
seems
pretty
interested,
so
anybody
interested
like
I
can
I
can
let
you
know
when,
like
we're
having
a
conference,
so
yeah
I'll,
I'll,
post
it
but
feel
free
to
come.
If
you
want
it'll
be
interesting,
yeah.