►
From YouTube: SimPEG meeting Sept 16, 2020
Description
SimPEG meeting and intro to iooxa
A
Drops
in
cool
well
thanks
for
joining
ron,
I'm
excited
to
to
see
some
of
what
you're
up
to
show
it
off
with
everyone.
Yeah.
So
rowan's
came
to
give
us
a
demo
of
some
of
the
work
that
he's
been
doing
with
steve
and
others
at
ioxa,
but
also
maybe
if
you
have
some
notes
or
things
you
want
to
be
sure
we
get
to.
A
Please
just
drop
them
in
the
in
the
notes
here
and
we'll
make
sure
that
we've
got
enough
time
at
the
end,
to
yeah
cover
any
sort
of
important
items
or
burning
comments,
and
things
like
that:
yeah
cool
rowan.
Do
you
wanna?
Take
it
away?
Do
you
know
everyone
here
by
the
way,
I
think
you
know
most
people?
I
don't
know
if
you
know
john.
B
C
B
If
I
like,
look
out
to
the
future
of
how
I
would
like
academia
and
education
to
work,
these
are
the
sort
of
resources
that
I'd
like
to
see
a
lot
more
of
in
the
world
where
you
have
like
embedded
code
that,
like
you
know
that
it
runs,
you
can
jump
into
that.
You
can
see
the
source
code,
you
can
dive
deeper
as
well
as
you
can
contribute
back.
So
right
now.
B
B
B
B
Does
that
sort
of
make
sense,
and
so
inside
of
here
I
can
either
work
at
the
notebook
level
or
actually
the
cell
level
and
so
similar
to
over
in
the
ioxa
land.
I
can
actually
save
this
cell
to
that
project,
so
I
can
save
that.
What
happens
when
I
do
that
is
it's
now.
Actually,
tracking
the
versions
of
that
and
it's
over
in
ioxa
land.
B
B
B
C
B
Awesome
so
this
this
is
now
like
a
linked
cell,
that's
sort
of
managed
by
iosa.
That
also
allows
you
to
jump
to
and
see
the
changes
so
that,
if
you're
working
with
someone
who's,
not
in
like
is
maybe
like,
wants
to
contribute
to
the
writing
side
of
a
jupiter
notebook,
but
not
necessarily
jump
in
and
change
the
code
or
if
you're
like
versioning
code
cells
between
notebooks,
that's
sort
of
allows
you
to
pull
those
changes
in
and
now
that
math
and
sort
of
all
of
the
text
is
updated
right
away.
B
E
E
B
So
I
this
isn't,
this
isn't
hooked
up
to
the
gsi
at
all.
Right
now,
so
I
was
just
sort
of
using
some
of
these
example
notebooks
to
to
demonstrate
that
sort
of
that
sort
of
work,
and
so
this
this
is
sort
of
like
facilitating
that
out-of-band
collaboration
loop,
so
say
lindsay
like
sent
you
a
notebook.
This
is
a
different
way.
B
B
A
C
C
C
B
B
B
G
Yeah,
so
I
think
this
kind
of
opens
up
the
door
of
like
editing
some
of
the
contexts
in
our
like
notebooks
or
apps,
because
we
have
some
introduction
of
the
apps,
but
I
mean,
I
think,
that's
a
rough
sometimes
like
it's
not
perfect.
Writing
and
people
may
want
to
change,
but
we
don't
have
a
pretty
good
mechanism
to
change
that,
but
this
would
be
a
kind
of
nice
tool
to
do
that,
but
rowan
I
I
just
had
a
chat
with
rosemary.
G
It's
not
really
directly
related
to
this,
but
one
thing
that
she
didn't
like
not
about
the
ioxa,
but
the
idea
of
white
paper
so
sometimes
like
they
just
produce
some
scientific
result
without
the
peer
review,
and
then
they
figure.
Oh,
that
was
wrong
and
she
kind
of
hated
that
so
I
I
thought
like
there's
some
somewhere
in
between
blog
and
the
peer-reviewed
article
where
like,
if
you
have
some
functions,
that
you
can
revisit
and
the
peer
reviewed
by
other
people
and
have
some
functionality
to
edit
and
at
least
make
some
sort
of
suggestion.
G
I
think
there
there.
I
think
there
are
places
this
can
play
a
huge
role.
I
I
just
thought
when
I
was
just
listening
to
her.
This
could
be
an
interesting
tool
to
be
used
because
I'm
just
thinking
about
oh,
what
is
the
real
difference
of
medium
blog,
but
that
media
blog
is
sort
of
on
the
fly.
So
people
write
things
and
very
fast.
It's
actually
nice
to
have
that
fast
mechanism,
but
you're
not
going
to
go
backwards
in
that
medium
block
and
that's
probably
not
designed
to
be.
G
F
G
Comments,
but
what's
actually
the
downside
of
that
white
paper
idea,
I
I
think
I
kind
of
let's
say:
stanford
university,
like
big
name
university
right,
something
about
that
and
then
some
people
figure.
Oh,
that's
not
right,
and
that's
that's.
That
makes
a
huge
problem
so
like
that's,
where
kind
of
why
she
didn't
like
it
that,
because
there's
no
going
backwards
and
wait
to
fix
it,
but
that
working
in
the
peer-reviewed
journal
it's
just
too
slow,
so
you
need
somewhere
in
between.
F
F
F
B
A
F
A
C
B
If
I,
if
I
just
copy
this
and
open
up
a
tree
view-
and
I
say
a
new
python-
3
notebook,
so
this
the
extension
also
works
over
here.
So
I
can
go
back
to
here
copy
this
link
over
here
and
import
a
cell
paste
it
in
for
f6.
The
paste
doesn't
work
and
I
haven't
fixed
that
one,
but
that
that
would
then
show
up
right
there.
B
F
F
B
C
B
Yeah,
so
this
this
is
like
a
notebook
that
I've
just
saved
to
ioxa.com.
So
I
can,
if
I
wanted
to
share
this
with
someone
like
they
can
go
in
copy.
This
link
go
to
their
jupiter
wherever
it
is,
and
then
import
it
from
here.
So
this
is
sort
of
what
I
was
showing
they're
trying
to
show
before
and
then,
when
you
hit
the
import
button
that
like
just
brings
in
the
notebook.
B
So
this
is
like
a
a
pretty
easy
way
to
share
notebooks
as
well
between
two
different
people,
and
so
this
output
actually
is,
is
saved
on
ios.
So
we're
not
able
to
show
that
right
now,
hopefully,
by
the
end
of
this
week
we
will
be.
But
then
this
output
specifically
can
be
added
into
different
different
articles
as
well.
So
like
pulling
this
image
into
a
report
or
a
paper
or
something
like
that,.
F
C
B
That
would
be
the
goal
is
trying,
because
I,
I
think
that
I
don't
know
publishing
everything
to
pdfs.
You
lose
a
lot
in
that
sort
of
conversion,
and
if
you
keep
some
of
those
links
alive
over
to
notebooks
and
make
it
more
reproducible,
you
can
also
have
interactivity
potentially
in
there
and
all
sorts
of
other
stuff,
and
it
gets
a
little
bit
more
exciting.
F
B
G
G
G
D
Over
from
technical
report
to
techno
technical
report
to
technical
report,
however,
figures
they
seem
to
change,
so
I
don't
know
how
that
would
maybe
benefit,
but
certainly
like
yeah,
just
like
the
basic
like
the
equations
for
yeah.
What
is
a
pseudosection?
How
is
it
plotted
and
all
that
kind
of
stuff
yeah?
I
I
like
the
idea,
I
just
I
don't
know
I
was
even
trying
to
get
people
onto
latex,
because
then
you
can
at
least
store
some
variables
and
stuff,
but
it's
hard
to
get
people
like
away
from
this
industry.
D
D
B
I
think
in
in
like
talking
to
some
other
folks
hearing
like
their
technical
reports
that
are
basically
like
copy
pasted
for
different
clients,
or
something
and
there's
like
paragraphs
in
there
that
just
like
carry
on
typos
for
years
and
like
because
nobody's
there's
no
easy
way
to
actually
fix
it,
because
you're
always
working
from
that
root,
template
and
it
just
like
propagates
forever.
And
it's
like
not
worth
fixing.
D
A
I
A
I
C
I
I
It
doesn't
really
it's
not
really
a
thing.
It
doesn't
represent
the
shape
of
anything
that
lives
on
the
mesh
at
all.
It's
just
a
tuple
of
the
number
of
s
number
of
like
for
edges,
it'd,
be
the
number
of
x,
edges
number
of
y
edges,
number
of
z
edges.
Those
are
just
I
mean
those
are
three
different
things
that
each
live
on
the
mesh
right.
B
I
Right
and
that's
why,
like
I
suggested,
naming
those
to
be
like
shape
cells
and
shape
faces?
Why?
Because
it's
the
shape
of
like
it's
kind
of
like
a
numpy
shape
right
like
it's
the
shape
of
the
array,
it's
the
shape
of
things
that
live
on
the
mesh
and
that's
when
I
that
kind
of
made
me
thinking
like
okay.
Well,
what
is
the
point
of
this
via
e,
like
you
could
have
a
shape
face
as
y,
because
that's
something
that
lives
on
the
mesh?
It's
the
shape
of
the
y
face
array.
I
H
I
do
think
that
yeah,
probably
if
you're
in
doubt
rather
drop
one
or
two
or
three
things,
because
there
are
so
many
things
and
that's
also
what
I
thought,
maybe
about
the
n
c
x
and
c
y
and
z
sets
because
in
the
end
there
are
n
c,
0
and
c
1
and
c
2
yeah
or
just
keep
the
short
names,
but
they
point
to
shape
cells,
zero
shape
cells,
one
I
don't
know
yeah,
but
there
are
so
many
methods.
So
because
it
happened
to
me.
I
think
I
asked
rowan
more
than
once.
C
I
I
mean
I
kind
of
like
just
relying
on
them
like
as
far
as
like
shape
cells,
because
then
it's
not
really
because,
like
there's
a
little
bit
of
discussion
on
the
cylinder
mesh,
the
cylindrical
measure,
at
least
I
brought
up
like
okay.
Does
it
really
make
sense
to
call
this
call
these
the
x-faces
on
a
cylinder
mesh
like
or
it's
like?
I
I
don't
know
you
know
what
I
mean
like
it's,
okay.
What
is
you
kind
of
have
to
know
that
x
means
radial
on
a
cylinder
mesh?
You
have
to
know
that
these
things
mean
that
yeah
but
kind
of
going
along
that
way.
So
that's
kind
of
one
of
the
thoughts
and
discussion
we
were
talking
all
right
brought
up
there.
Another
thing
was,
and
so
we
have
an
ncx
ncy
ncz
property
right,
which
we
understand
that
those
are
the
number
of
cells
in
the
x
direction,
but
then
we
also
have
an
like
nfx.
I
What
number
of
faces
number
of
x
faces
right
that
those
are
two
different
like
types
of
things,
but
looking
at
how
they're
named
I,
it
would
lead
me
to
believe
coming
in
from
the
outside
that
they
would
mean
the
similar
things,
but
they're,
not
very
like
one
quantity
represents
the
number
of
cells
in
the
x
direction.
The
other
quantity
represents
the
number
of
x.
I
A
I
A
I
My
thought
was
okay,
the
only
point,
the
only
time
when
they're
grids
is
when
they're
on,
like
the
tensor
mesh
or
tensor
type
meshes
or
right,
they're,
not
grids
on
the
tree,
mesh
they're,
not
grids,
and
if
we
had
like
a
triangular
mesh,
so
it's
like
okay!
Well,
should
they
just
I
mean
they're,
not
even
grids
really
they're,
just
locations
they're
the
xyz
values
of
all
of
these
things.
I
The
locations
of
all
these
things
that
so
it's
like
okay
does
grid,
really
make
sense
as
a
name
and
they're
not
grids
like
and
they're,
not
necessarily
grids.
So
that's
why
I
was
like
okay
well
either.
That's
what
led
me
to
calling
them
locations
or
lindsay
had
brought
up
just
dropping
grid
or
location,
just
calling
them
cell
centers
nodes,
edges
faces.
Why,
and
I
kind.
J
You
know
yeah
this
came
up
with.
I
think
I
designed
something
that
would
return
the
the
cell
width
for
the
tree
meshes
and
then
we
went
through
a
whole
host
of
names
and
we
ended
up
calling
it
like
gridded,
h
or
something.
I
think
it's
like
h,
gridded
right
now,
yeah,
which
I
mean
just
come
to
consensus
on
something
and
then
stick
with
it,
but
yeah
like
cell
thicknesses
or
whatever.
J
I
G
E
G
Unstructured
grid
have
could
have
like
random
points,
so
I
don't
know
I
I
I
probably
just
used
to
it
and
I
reluctant
to
change
usually,
unless,
like
you
have
very
clear
reason
why
we
need
to
change
yeah
yeah
and
this
yeah.
At
this
point,
I
I
don't
know,
I
think
I
I'm
kind
of
fine
with
great
cc,
with
fx
and
sims
making
sense
to
me.
A
So
do
you
have
sort
of
the
timeline
in
mind
like
do
we
want
to
maybe
leave
this
document
open
until
the
end
of
the
week
and
let
folks
kind
of
jump
in
or
do
you
want
to
sort
of?
I
mean
I
think,
there's
some
things
that
it
seems
like
we
have
pretty
solid
consensus
on
yeah.
How
would
you
like
to
proceed
joe.
I
I
A
H
H
H
J
Say
any
so
many
of
the
examples
tutorials
all
of
these
things,
I'm
guessing,
maybe
a
large
portion
of
the
tests,
or
at
least
a
certain
portion
of
the
tests
that
we
have
are
all
going
to
use
the
the
sort
of
original
way
of
calling
this
stuff.
So
I
mean
once
we've
come
to
consensus.
That
seems
like
a
fairly
easy,
albeit
boring
task
that
maybe
I
could
throw
some
powers
at.
I
J
H
I
A
Yeah
and
then
I
guess
the
other
thought-
and
this
is
a
bit
more
general,
but
both
with
senpai
again
with
discretize
is
it
would
be
great,
and
I
don't
know
how
to
do
this,
but
I
know
it
can
be
done
on
our
docs
right
now.
We
only
ever
have
the
latest
version,
and
so
it
would
actually
be
nice
if
we
could
have
be
able
to
go
back
to
previous
versions
and
right
now,
they're
deployed
discretizes
on
github
pages.
A
I
I
I
H
A
H
J
Yeah
so
started
to
work
through
the
rest
of
the
tests
for
the
simpeg
em1d,
so
I've
got
kind
of
everything
there,
except
for
the
stuff
involving
the
dual
moment
and
to
make
sure
I
understand,
is
this
sort
of
the
idea
that
you're
using
different
sources
for
different
ranges
of
time
channels,
like
you'll,
have
one
source
that
is
going
to
be
corresponding
to
early
time
channels
and
then
one
source?
Maybe
it's
a
different
waveform
or
it's
like
a
different
dipole
moment
for
the
late
time?
G
J
G
J
J
J
But
at
this
point
it's
it's
really
coming
together
and
I'd
say
95
of
the
tests
work
and
we
have
a
whole
suite
of
examples
that
will
let
you
do
forward
modeling
and
inversion
for
a
single
sounding
and
the
the
stitched
1d
forward,
modeling
and
inversion,
both
frequency
and
time
domain.
So
the
core
set
of
of
tutorial
examples,
and
basically
all
the
tests
are
working
now.
J
A
J
Yeah
yeah
there
there
is
a
mapping
for
h.
That
kind
of
framework
already
exists.
I
just
don't
want
a
situation
where,
for
some
applications
the
zed
position
is
the
actual
elevation
and
then
for
this
particular
situation,
it's
defined
as
being
an
offset.
I
just
need
to
think
about
it.
I
do
I
do
know
where
it's
happening,
but
yeah.
I
I
want
to
make
it
not.
I
don't
want
to
make
it
confusing
for
the
end
user,
that's
that's
kind
of
who
I'm
thinking
about
when
I'm
making
these
decisions.
G
J
Is
we
could?
This
is
actually
what
we
do
when
you
define
the
receiver,
we
can
put
in
a
flag
where
you
would
say
like
absolute
or
offset,
and
you
could
actually
just
choose
whether
or
not
you
want
to
define
your
receivers
relative
to
the
source
or
if
you
want
to
define
your
receivers
as
the
absolute
position.
G
D
Yeah,
just
a
quick
one.
There
yeah
dom's
got
me
my
head
a
little
wrapped
around
the
tiled
simulation
problem
there.
So
I've
started
that
up
yeah.
I
push
it
to
my
empty
branch,
but
I'll
make
another
separate
one
so
that
it's
because
I'm
a
little
bit
behind
and
it'll
just
be
easier
to
pull
that
in
so
I've
just
put
it
all
onto
the
dash
side.
So
I
set
it
up
like
how
you
did
for
a
simulation
there,
joe.
So
it.
A
D
D
F
G
D
Then
a
note
that
I
actually
something
that
interesting
I
found
when
I
was
just
working
with
the
paralyzations
of
just
with
multi-thread
or
multi-processing
and
desk
yeah.
So
I
found
I
was
just
running
an
example
and
I
found
that
my
cores
with
utilization
was
showing
like
100
right
off
the
bat
using
just
like
a
simple
multi-processing
problem
and
then
all
of
a
sudden
it
just
kind
of
went
down
to
like
one
percent
and
it
just
continued.
My
code
was
running.
D
Everything
was
showing
that
it
was
running,
but
it
started
off
for
full
power
and
then
it
just
died
out.
So
then
I
looked
into
it
a
little
bit
more
and
apparently
the
problem
is:
is
the
blast
library,
that's
numpy,
scipy
and
pretty
much
everything
is
built
with
a
blast
or
open
blast.
I
think
it
was
both
actually
that
had
the
effect
they
reorder
your
your
resources
as
soon
as
on
import.
D
So
if
there
I
found
a
little
I'll
have
to
post
the
link,
but
there's
a
command
line
that
you
can
use
to
the
os
system,
and
I
think
it
only
works
on
maybe
linux
and
mac,
and
you
can
tell
it
to
go
back
to
a
single
threaded
and
not
use
multi-threaded
blasts
and
then
all
of
a
sudden
you're
you
stay
at
100
core
usage.
All
the
time.
D
Okay,
well,
it's
not
necessarily
this
well,
I
guess
yeah,
it's
the
speed.
It's
just
like
all
of
a
sudden.
It
is
using
everything,
but
then
it
just
dies.
Oh
it
was
well.
I
was
super
curious
on
that.
It's
like.
Why
does
that
happen?
So
I
don't
know
if
people
see
that
yeah
blast
is,
I
guess
it's
multi-threaded
and
does
say
that,
but
it's
yeah
that
must
be
the
root
of
all
these
problems.
I'd
say
with
pardeso
and
without
that
being
thread
safe
yeah.
D
A
Interesting
cool
thanks
john,
so
I
have
a
quick
update.
Is
that
next
week
I'm
giving
a
talk
julio
reached
out
from
the
swung
group.
So
this
is
targeted
at
an
audience
in
mexico
and
he
asked
if
I
would
talk
a
bit
about
sort
of
open
source
software
development.
It's
mostly
it's
a
geoscience
and
geophysics
audience.
So
talk
a
bit
about
simpeg
and
some
of
the
work
we're
doing
and
also
kind
of
open
source
practices
and
open
source
communities.
A
E
E
Anyway,
we
haven't
figured
out
how
to
celebrate
yet,
but
that's
on
tebow's
hands,
so
hopefully
there'll
be
something
tomorrow
or
short
shortly
after
that,
and
we
haven't
figured
out
whether
it
could
be
virtual
with
you
know,
people
in
calgary
and
california
or
where
they're
actually
going
to
try
to
have
something
in
the
park.
But
diva
is
going
to
do
well
and
yeah
people
want
to
give
them
a
shout
beforehand.
That'd
be
great,
sounds
good.