►
From YouTube: SimPEG meeting May 13
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
B
Yes,
I
can
I
have
minded
these
two
questions
of
each
one.
You
know
one
is
that
the
disk
retires
and
apparently
go
never
deployed
on
Conda.
Is
there
a
particular
Oh?
Does
anyone
know
what
they
show
that
all
the
actual
pipelines
fail,
because
it
appears
because
it
since
January
that
they
have
been
run,
the
locks
are
not
retained?
So
I
can't
really
look
at
the
locks
why
they
failed.
D
B
D
D
B
A
That's
where
so,
maybe
just
because
they
do
with
this
small
in
the
3d,
it.
A
We,
like
I,
mean
we've
the
testing.
That's
in
simple
right
now
is
a
bit
loose
on
that,
like
we,
we
haven't
actually
done
super
in-depth,
really
trying
to
match
analytics
part
like
I
mean
it's,
so
we
like
with
the
example
that
Peters
referring
to
yen
posted
a
time
domain
example
where
he's
trying
to
just
compare
with
a
half
space
solution
for
Morden
Helmand,
and
we
can't
get
the
error
down
really
below
6%
and
it
looks
like
there's
a
slight
shift
in
like
it
looks
like
we're
slightly
above
the
solution.
I
believe
time.
A
Soggies
noticed
this
in
the
like.
In
an
airborne
example,
we
really
tried
to
match
hard
a
solution.
It's
not
significant
I
mean
not
hugely
significant,
but
six
percent
is
still
non-trivial
and
we
should
be
able
to
fix
that.
So
it's
something
where
I
think
we
basically
just
need
to
go
through
the
derivation
again
and
just
check
I
mean.
So
all
the
sensitivities
are
correct
and
everything,
but
it
is
I,
think
I
think
we're
probably
just
have
like
you
know,
a
DT
or
an
index,
that's
one
off
somewhere
that
is
causing
this
issue.
B
B
For
that
simple
I'd
particular
if
you
have
stores
and
receivers
exactly
on
the
faces
and
things
like
to
see,
yeah,
okay,
yeah,
another
wise
how
ping
so
garyun.
If
I
don't
hear
on
slack
back
yeah
I,
don't
know
that
no
I
was
looking
a
lot
at
the
need:
emulation
branch
from
synthetic,
how
you
handle
all
the
surveys
and
sources
and
receivers
and
bait
done
so
I
tried
to
learn
and
get
a
bit
out
of
all
the
things
you
have
discussed
already
and
thought
about
her.
B
So
I
don't
have
to
do
everything
again,
it's
really
great
and
yeah.
That's
it
I've,
also
working
on
young
saving
and
loading
and
backends
a
while
ago,
with
the
hdf5
Dom
helped
me
a
bit
and
now
also
for
simpler
cases,
the
Chasen,
which
is
nice
just
to
look
at
it.
If,
if
it's
a
small
problem
that
was
quite
sukham,
Isaac.
A
B
The
interesting
thing
is:
if
you
have
really
small
problems,
then
then,
actually
chasing
is
smaller
than
numpy,
compressed
or
hdf5
compressed,
because
they
already
had
some.
They
have
some
overheads
to
accommodate
all
the
stuff
but
yeah.
But
then
we've
talked
about
kilobytes
and
otherwise
gigabytes.
A
B
B
F
So
I
guess
before
the
weekend
put
some
work
into
doing
the
em1
D
stuff,
so
I
mean
there's
still
lots
of
work
that
needs
to
be
done
on
it,
but
right
now,
I
was
able
to
update
it
so
that
it
would
run
on
the
latest
simulation
class
and
I
was
able
to
do
a
local
Wendy
inversion
and
actually
do
the
global
1d
inversion.
So
that's
kind
of
a
first
step
and
I
was
just
sort
of
going
through
a
lot
of
the
I
guess
the
notebooks
that
were
there
and
trying
to
get
those
notebooks
to
work.
F
F
Maybe
it's
gonna
go
into
a
utilities
file,
that's
at
a
higher
level,
so
I
got
to
figure
out
where
I
want
to
put
some
stuff
and
then
also
make
her
a
proper
survey,
source
and
receiver
class
right
now,
everything's
dumped
into
the
survey
class
and
I
guess:
yeah,
the
properties
are
sort
of
being
assigned
to
stuff.
That's
not
standard
standard
sim
peg.
So
it's
really
about
what
properties
are
assigned
to
which
classes
and
parsing
it
out.
But
I've
got
the
first
step
and
yeah
I
just
wanted
to
figure
out
Dom.
F
E
Here
you
only
driver
if
you
have
it
and
I
that's
the
one
I
gave
to
you
right.
That's
the
it's
the
same,
one
that
we
had
use
for
jiff
tools,
bag
down
and
I
repurposed
it
for
what
I'm
doing
right
now
with
with,
but
it's
the
same
stroke
but
I,
don't
think
st.
Peggy
MIT
itself
has
has
any
driver
in
it.
There's
gonna
be
a
lot
of,
as
you
said,
like
utils
and
yeah.
E
D
F
A
Thanks
Devon
one
thing
with
the
utilities
it
might
be
worth
just
like
sketching
sketching
that
out
just
in
an
issue
before
sort
of
moving
stuff
around
to
just
sort
of
sort
that
we
can
all
kind
of
think
through
that,
because
I
think
there's
pros
and
cons
to
both
Divac,
creating
like
a
utilities
folder
or
to
trying
to
keep
them
them
stoked.
So
yeah.
If
there's
something
that
we
can
all
have
just
a
bit
of
a
discussion
before
you
make
a
decision.
That
would
be
great.
D
There
is
already
like
a
utilities:
folder
the
natural
source,
stuff,
I'm,
simple
sory,.
F
E
F
I
mean
there's:
this
is
a
actually
a
fair
amount
of
decisions
that
need
to
be
made,
which
was
whether
or
not
we
want
to
make
a
simulation
1d
class
in
frequency
domain
and
a
simulation
1d
class
in
time
domain
or
whether
or
not
we
want
to
make
just
a
like.
The
e/m
1d
problem,
a
step
up
and
then
put
the
the
frequency
domain
and
the
time
domain
in
there
because
they
use
so
much
of
the
same
stuff
so
that
that's
kind
of
a.
E
F
E
H
Yeah
yeah
I
gotta
got
sidetracked.
Last
week
we
got
a
data
set
processed
for
our
tipper
data
and
I've
been
trying
to
invert
earth.
Not
trying
I
was
actually
successful
in
inverting
it
yeah
it's
over
the
vert
of
red
of
fort
Dome
in
South
Africa
there
yeah
it's
it's
coming
out,
pretty
good.
It's
a
it's!
A
large
survey,
so
I
had
to
kind
of
dumb
it
down
a
bit
with
the
number
of
cells,
but
yeah
the
inversion
seems
to
be
working.
H
The
new
speed
up
of
the
MT
code
for
JT
back
and
Javik
seem
to
be
working
and
producing
proper
stuff,
yeah
I'm
still
just
working
on
that
we're
gonna,
try
and
invert
some
impedance
data
over
that
next,
but
that
again
has
to
I
need
to
use
the
sensitivity
matrix
so
hope
I'll
be
getting
back
to
that.
I
really
want
to
finish
that
off,
but
yeah.
H
H
Hopefully
it
gets
back
to
me
and
then
yeah
I'm,
just
still
fighting
with
interfacing,
with
the
simple
inversions
on
our
cluster
I,
everything
was
working
and
then
I
went
to
compile
it
for
Windows
and
the
SSL
libraries
are
completely
different
and
I've
got
to
change
a
whole
bunch
of
stuff
for
security,
wise,
my
forte,
but
I
guess
I'm
learning.
If.
H
D
H
G
H
G
H
So
we
I'm
just
using
four
frequencies
right
now
and
range
from
50
to
700
Hertz,
but
we
did
pull
out
like
46
frequencies,
so
from
50
Hertz,
55
Hertz
to
about
10,000
Hertz
we
pulled
out.
But
again
not
all
the
data
is
is
worth
30
with
so
I've
just
started
with
four
just
because
of
my
computer
can
only
handle
so
much
but
yeah.
H
The
only
problem
I
really
had
was
getting
the
the
background
set
as
soon
as
I
had
a
nice
background
things
just
kind
of
took
off
so
so
far,
that's
really
been
the
only
hurdle,
it's
still
a
little
slow
but
yeah.
It's
a
bit.
It's
faster
now
and
took
me
about
eight
hours
to
get
eight
iterations
so
about
an
hour,
an
iteration
at
four
frequencies.
Okay,.
H
So
I
made
the
mesh
pretty
tiny.
It's
only
about
two
hundred
and
some
thousand
cells
and
about
well
I
also
threw
out
a
bunch
of
stations
that
weren't
that
great
so
only
about
1900
stations
in
total,
so
yeah.
It's
still
pretty
small
data
set
in
the
scheme
of
things,
but
the
impedance
one
that
we'll
be
doing
next.
All
the
stations
worked
out
pretty
well
for
that.
H
So
I'll
get
a
better
idea
on
how
adding
more
data
is
gonna
slow
down
the
inversion
or
not,
and
oh
yeah
Joe
I-
might
bug
you
this
week
about
changing
over
the
MT
gas
stuff
to
what
you
have
now
for
simulation.
It
should
be
pretty
easy.
I
just
have
to
pretty
much
move
all
that
code
into
the
desk
folder
and
just
make
up.
H
D
H
I
got
into
the
DC
stuff
that
you
did
last
week
a
little
bit.
I
think
I
see
what
you're
doing
so.
Yeah
I
might
bug
you
here
and
there,
but
sure
you'd
be
good.
H
D
So
last
week
at
then,
it's
good
writing
up
a
last
bit
of
a
project
I've
been
working
on,
but
I
also
started
looked
into
again.
The
serialization
stuff
clones
ian
has
been
working
on
trying
to
figure
it
out
and
write
into
it
a
little
bit
of
a
problem,
and
it
looks
like
it's
all.
Gonna
take
a
little
bit
more
work
than
we
first
thought
to
get
all
the
surveys
and
simulations
serializing
properly
so
or
pop
I
think
it's
best
to
not
worry
about
it.
D
At
this
point
for
the
simulation
pull
requests
and
work
out
afterwards,
we
have
a.
We
have
a
path
forward
to
fix
the
issues,
but
essentially
what
was
happening
is
that
every
single
class
that
inherited
from
has
properties
like
has
a
properties
has
properties
is
also
was
all
stored
on
the
same
kind
of
list
of
classes
like
same
registry.
This
registry
that
lived
inside
that
has
properties
class
and
it's
the
way
when
he
would
deserialize
something
good
to
figure
out
what
class.
D
To
put
it
in
is
that
it
would
look
at
this
list
find
the
name
of
the
class,
so
the
class
would
let's
say
the
classes
called
dipole
white.
At
least
you
know,
iam
static
receivers
dipole,
so
the
name
of
the
class
is
dipole.
So
it'd
look
up
dipole
in
this
registry
that
had
a
list
of
everything
I
and
it
would
try
to
pick
out
the
class.
Well,
there's
a
few
different
classes
and
sin
that
are
named
dipole
but
they're.
D
C
D
This
so
the
basically
the
path
forward
is
that
we
need
to
purposely
redefine
the
registry
on
each
class
that
we
intend.
So
it's
like.
Okay,
we
only
intend
for
these
classes
like
that,
to
these
classes
to
be
together
so
that
the
base
has
to
have
its
own
registry,
but
that'll
take
a
little
bit
more
work,
moving
forward
to
get
everything
Alex
grab,
to
tease
out
a
few
things,
especially
with
the
natural
source
surveys
that
on
here
come
a
base,
a
base
frequency
domain
things,
even
though
they
they
need
to
be.
D
The
simulation
expects
them
to
be
only
one,
only
three
specific
classes,
probably
better,
to
add.
In
fact,
a
map
based
natural
source
source
class
just
to
have
a
common.
So
what
those
sources
all
have
a
common
thing:
it
doesn't
have
those
kind
of
classes,
don't
have
to
do
anything,
they
can
just
be
really
big
into
the
empty,
except
for
redefining
this
registry
and
then,
but
we
have
a
path
forward
for
that
now,
but,
like
I
said,
it's
gonna
we're
gonna
leave
that
until
after
the
simulation
clerk
has
good
skin.
D
F
A
D
B
G
B
B
D
E
Yeah
I
need
to
get
going
and
portion
here
pretty
much
right
after
this,
but
so
I
two
things
one
is
that
I
got
more
emails
from
scg
about
our
workshop
in
October
and
so
they're
asking
that
we
really
set
all
speakers
we
don't
have
to
have
speakers
would
basically
need
to
have
an
outline
by
the
end
of
July
exactly
what
it
is.
It's
a
bit
Fe
right
because
I
was
talking
to
judge
he's
like
Irene
izing,
one
for
machine
learning
and
there's
so
much
chatter
on
the
mailing
list
that
people
are
not
gonna.
E
E
A
A
E
I
E
The
other
thing
right
is
that
the
SE
g1
s
e-g,
is
not
supporting
us
for
enter.
Now
we
have
to
bring
our
own
I.
Guess.
I
didn't
talk
about
this,
but
I
I
emailed
them,
they
I
emailed
us
and
said:
would
you
guys
need
for
the
workshop
and
I
said
well,
doesn't
matter
as
long
as
what
I
meant
there
now
and
she
said
no
we're
not
providing
any
of
this,
so
we
need
to
bring
out
on
internet
and
then,
if
people,
if
our
speakers
are
online,
then
we
you
know
it's
so
super
s.
E-G.
E
The
other
point
is
that
I'm
working
with
with
Sarah
again
on
RDC
agriculture,
the
agriculture
problem
thing
and
for
some
reason,
and
that's
a
question
for
a
SAG
yep
Erin
resistivities
for
the
short
dipoles
are
way
off
right.
So
we
do
a
forward
summation
with
a
with
a
uniform
half
space.
It's
just
flat.
You
know
we
referred
it
to
just
go
flat
and
even
down
after
we
calculate
their
apparent
resistivities
as
soon
as
the
dipoles
get
like.
E
C
E
C
Yeah,
because,
if
you're
actually
using
problem,
C
seen
to
decode
it
like
it's,
so
it's
actually
quite
different
I
think
I
love
this,
like
a
boundary
condition
set.
We
made
it
for
tensor
mesh,
wouldn't
work
so
because
we
haven't
designed
for
quadtree
mesh.
Okay,
I
think
that
problem
and
should
be
fine,
because
that's
what
we're
using
for
gwv
codes
and
stuff
unless
we
are
using
Popo
so
yeah,
it's
a
double
check.
What
problem
you're
using
and
you're.
C
For
like
a
tree
mesh
problems,
you
see,
yeah
I
think
don't
use
problems,
it's
because
it
may
work,
but
you
probably
need
to
pet
a
lot
because
actually
you
just
use
it
you're
a
army,
tradition,
so
yeah
and
then
yeah.
You
probably
need
to
kind
of
check
which
one
do
kitchens
used
for.
If
you
were
using
problems
to
see
okay.
C
A
You
run
off
I
just
had
a
couple,
quick
questions
actually
about
the
transform
tutorial
and
just
wanted
to
put
some
stuff
on
on
your
radar.
I
started
in
the
outline
document
and
I
put
a
link
just
in
the
in
the
notes
here,
so
be
great
to
get
your
thoughts
and
input
and
from
anyone
else
here.
Who's
interested
in
being
involved
in
that
I've
got
a
couple
questions
so
with
respect
to
a
data
set.
A
If
there's
something
that
we
want
to
pick
and
if
we
can
settle
on
that
in
the
next
week
or
so,
because
it
would
be
great
to
sort
of
get
playing
with
that
and
make
sure
we
don't
run
into
any
significant
issues
so
that
we
acted
like
triage
later
on.
So
if
and
that's
a
discussion
we
can
can
continue
to
have
afterwards
and
then
the
other
questions
I
have
are.
A
But
we
do
need
to
let
Matt
know
fairly
soon
because
he's
actually
gonna
ship
mics
to
all
the
presenters,
so
the
audios
good
and
then
even
if
you
don't
want
to
be
necessarily
online
if
somebody's
willing
to
help
with
slack.
So
that's
gonna
be
the
main
way
that
we
communicate
with
participants.
A
They're
gonna
have
a
dedicated
slack
channel,
and
so,
if
somebody
would
be
willing
to
just
like
monitor
and
sort
of
help,
ask
questions
or
help
raise
questions
for
myself
and
whoever
else
is
presenting
to
make
sure
that
we
don't
lose
track
of
stuff
because
it's
hard
to
man
all
of
those
things
at
once,
yeah
so
I,
don't
know.
If
that's
something
you
want
to
discuss
now
or
we
can,
we
can
discuss
that
later
to
Dom
I
think
you
need
to
run
yeah.
A
Thanks
for
coming
is
anyone
else
well,
while
we're
on
this
topic
with
respect
to
the
data
set,
so
the
ones
that
we
had
mentioned
were
Ragland
or
century
I,
don't
know
Doug
if
you
have
thoughts
on
those
or
if
anyone
else
has
thoughts
on
like
what
would
be
basically
like,
we
need
something
that
is
simple
enough
that
you
know
folks
can
run
this
on
their
on
their
laptops,
hopefully,
and
like
the
inversion
shouldn't
be
hopefully
too
too
complicated
like
requiring
too
much
tuning,
but
at
the
same
breath
we
you
know
likes
showing
some
of
these
realistic
things
is,
is
helpful.
A
G
G
A
A
Sorry,
it's
taken
me
so
long,
but
I've
been
playing
around
actually
a
lot
with
the
Autry
machines
for
the
frequency
domain
code
and
really
actually
trying
to
get
a
sense
of
where
stuff
works
well
and
where
solutions
are
breaking
down.
I'm
still
still
working
on
basically
trying
to
team
that
to
a
place
that
I'm
I'm
happy
with,
but
it's
been
great
playing
around
playing
around
with
those
brushes,
so
yeah.
A
B
I
Writing
reviews
just
like
I
was
a
frightening
ya,
know
I
told
well
of
you.
I
will
just
impose
a
paper
for
you
for
a
sec,
I
rewind
a
lot
of
the
exact
purpose,
some
sort
of
geological
constraint
so
I
will
add
that
but
yeah
nothing
really
on
really
on
the
cold
days,
like
once.
All
that
right
thing
is
done
like
like
I'm,
going
to
need
to
start
working
on
bringing
all
of
that
in
the
huge
project
in
a
huge
yeah.
A
Well,
thanks:
everyone
for
coming,
if
you
do
have
thoughts
on
the
tutorial
I
would
certainly
appreciate
any
input
and,
if
you're
willing
to
either
be
like
a
moderator
or
a
co-presenter
yeah,
just
just
let
me
know
we
can
get
get
all
that
connected.
Yeah.
H
I
could
probably
help
with
the
slack
side
of
things
like
I.
Don't
know
not
a
very
good
for
her
I
haven't
presented
a
lot
so
I'm,
not
sure
if
I'm
comfortable
that
helped
into
there,
but
whatever
little
things
I
can
certainly
add
in
though.
A
That
would
be
awesome,
I
mean
having
somebody
monitoring.
The
slack
would
just
like
that'll,
be
a
big
relief
for
for
presenting.
So
that
would
be
awesome
if
your
game
to
do
that.
Yep
Richard
thanks
John,
okay!
Well,
if
there
is
nothing
else,
then
we'll
see
you
online
and
even
video
next
week.