Ref/tdem testing #620

lheagy · 2017-06-24T18:09:27Z

clean up in the time domain testing
efficiency fix in the Problem class. Now we check if the model has truly updated before clearing matrices:

rowanc1 · 2017-06-24T18:19:36Z

SimPEG/Problem.py

-            if hasattr(self, prop):
-                delattr(self, prop)
+    def _on_model_update(self, change):
+        if self.model is None or not np.allclose(self.model, change['value']):


I wonder if an early exit of self.model is change['value'] would be faster. If the model is large a point wise comparison is going to be slow. If they are pointing to the same block in memory, then we can skip that.

rowanc1 · 2017-06-24T18:22:32Z

tests/em/tdem/test_TDEM_DerivAdjoint_RawWaveform.py


+# class TDEM_DerivTests(unittest.TestCase):


Would be nice to have these tests perhaps tested on dev-->master? Just so that we know that it is still working, because they are useful for debugging.

right now the TDEM tests are still timing out, so this would first take some re-organizing of tests

I added this back on the test_TDEM_DerivAdjoint

codecov · 2017-06-24T18:48:52Z

Codecov Report

Merging #620 into em/dev will decrease coverage by <.01%.
The diff coverage is 80%.

@@            Coverage Diff             @@
##           em/dev     #620      +/-   ##
==========================================
- Coverage   74.33%   74.32%   -0.01%     
==========================================
  Files         105      105              
  Lines       13956    13960       +4     
==========================================
+ Hits        10374    10376       +2     
- Misses       3582     3584       +2

Impacted Files	Coverage Δ
SimPEG/Utils/modelutils.py	`59.45% <100%> (ø)`	⬆️
SimPEG/Problem.py	`83.03% <75%> (-0.3%)`	⬇️
SimPEG/EM/TDEM/ProblemTDEM.py	`93.56% <0%> (-0.23%)`	⬇️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1927828...60f0f29. Read the comment docs.

rowanc1 · 2017-06-24T19:51:06Z

SimPEG/Problem.py

-            if hasattr(self, prop):
-                delattr(self, prop)
+    def _on_model_update(self, change):
+        if self.model is change['value']:


change['prev'] is change['value'] the model will always be the value!!

rowanc1

I find the if (not this and not that) a bit unclear as to why you would skip the check. If you are checking a corner case, then I would say err on the side of clearing model dependence and just remove this logic?

However, if your check on undefined and None is to make sure that these don't fail the np.allclose then I would suggest changing that to

isinstance(change['previous'], np.ndarray) and 
isinstance(change['value'], np.ndarray) and 
np.allclose(change['previous'], change['value'])

Which is checking what you actually want...!

So something like:

if change['previous'] is change['value']:
    return  # model has not changed
if (
    isinstance(change['previous'], np.ndarray) and 
    isinstance(change['value'], np.ndarray) and 
    np.allclose(change['previous'], change['value'])
):
    return  # model has not *really* changed

I think your current code would break by setting model to properties.undefined, for example.

rowanc1 · 2017-06-24T21:18:17Z

SimPEG/Problem.py

+        if change['previous'] is change['value']:
+            return
+        if (
+            not isinstance(change['previous'], properties.utils.Sentinel) and


I think that you can say change['previous'] is not properties.undefined there is only ever one instance of the sentinel hanging around, and it is much more clear what you care looking to do.

perfect, thanks @rowanc1

…not clearing property matrices

…s in test_TDEM_DerivAdjoint

lheagy · 2017-06-25T21:04:45Z

tests/em/tdem/test_TDEM_MultiSrc_DerivAdjoint.py

@@ -1,160 +0,0 @@
-from __future__ import division, print_function


I have updated test_TDEM_DerivAdjoint to test multiple sources, so this test is now redundant

lheagy · 2017-06-25T21:05:28Z

tests/em/static/test_DC_jvecjtvecadj.py

@@ -130,7 +130,8 @@ def test_misfit(self):
                self.survey.dpred(m), lambda mx: self.p.Jvec(self.m0, mx)
            ],
            self.m0,
-            plotIt=False
+            plotIt=False,
+            num=3


@sgkang: I fixed the number of iterations this runs. If we let it go too far it fails b/c we hit machine precision.

lheagy · 2017-06-25T22:16:25Z

examples/04-grav/plot_laguna_del_maule_inversion.py

-    download([url+f for f in cloudfiles], folder=basePath, overwrite=True)
+    # directory where the downloaded files are
+
+    url = "https://storage.googleapis.com/simpeg/Chile_GRAV_4_Miller/Chile_GRAV_4_Miller.tar.gz"


@fourndo, @craigm, I have zipped up the files into a .tar.gz (the originals are still in the same place, but downloading a .tar.gz is much faster than 4 independent files)

lheagy · 2017-06-25T22:17:20Z

examples/04-grav/plot_laguna_del_maule_inversion.py

+
+    url = "https://storage.googleapis.com/simpeg/Chile_GRAV_4_Miller/Chile_GRAV_4_Miller.tar.gz"
+    downloads = download(url, overwrite=True)
+    basePath = downloads.split(".")[0]


I have also downloaded to the current directory (a folder called Chile_GRAV_4_Miller will be created). This makes it easier to work with different systems and to keep track of the files (esp on travis)

cache docs examples

…reated when tar unzipped

lheagy · 2017-06-26T22:11:36Z

examples/00_published/plot_booky_1D_time_freq_inv.py

@@ -392,7 +392,7 @@ def run(plotIt=True, saveFig=False, cleanup=True):

    ax0.semilogx(sigma_re, z, 'k', lw=2, label="RESOLVE")
    ax0.semilogx(sigma_sky, z, 'b', lw=2, label="SkyTEM")
-    ax0.set_ylim(-100, 0)
+    ax0.set_ylim(-50, 0)


this is the range we used in the paper

lheagy · 2017-06-26T22:21:54Z

examples/04-grav/plot_laguna_del_maule_inversion.py

-    # Download to Downloads/SimPEGtemp
-    basePath = os.path.expanduser('~/Downloads/simpegtemp')
-    download([url+f for f in cloudfiles], folder=basePath, overwrite=True)
+    url = "https://storage.googleapis.com/simpeg/Chile_GRAV_4_Miller/Chile_GRAV_4_Miller.tar.gz"


@fourndo are you good with this?

i'm good with that.

… nan. Thumbs up of the merge - DF

lheagy · 2017-06-27T15:52:24Z

SimPEG/Utils/modelutils.py

@@ -2,7 +2,7 @@
 import numpy as np
 from scipy.interpolate import griddata, interp1d

-def surface2ind_topo(mesh, topo, gridLoc='CC', method='cubic', fill_value=np.nan):
+def surface2ind_topo(mesh, topo, gridLoc='CC', method='nearest', fill_value=np.nan):


thanks @fourndo!

lheagy · 2017-06-28T02:03:05Z

Thanks for your input all!

lheagy added 4 commits June 23, 2017 16:24

use setup class to define and store fields to speed up testing

d71d013

cleanup of the raw waveform tests as well

35f204b

remove fdem inversion example in favor of the published one

68dc071

check if the model has truly been updated before clearing matrices

604b07f

rowanc1 reviewed Jun 24, 2017

View reviewed changes

early exit if the model is change['value']

fc7b423

rowanc1 reviewed Jun 24, 2017

View reviewed changes

lheagy added 2 commits June 24, 2017 13:07

clean up logic for model observer

3ade675

limit the number of steps taken in the sensitivity test for DC

606b503

rowanc1 requested changes Jun 24, 2017

View reviewed changes

lheagy added 5 commits June 24, 2017 17:51

update logic so it is more transparent when we are early exiting and …

1e5d502

…not clearing property matrices

extend timing slightly for docs build

ca291a1

reverse the travis wait addition

1a99ecb

try using travis wait only on docs

bcc048f

remove test_TDEM_MltiSrcDerivAdjoint and instead test multiple source…

08e89b4

…s in test_TDEM_DerivAdjoint

lheagy commented Jun 25, 2017

View reviewed changes

lheagy added 3 commits June 25, 2017 14:59

download laguna del maule data as a .tar.gz

8a74a56

remove printing comments around active cells

43513dd

restore commented-out tdem test

7683e8e

lheagy commented Jun 25, 2017

View reviewed changes

lheagy added 6 commits June 25, 2017 15:31

shrink the size of the mesh being used in the TDEM tests

b83926c

upload half the examples via travis

4e4eed2

comment out celia figure for now

d23da2d

print pwd contents, travis

8537633

upload examples directory as well

8332c9c

simplify download locations

8734b83

lheagy added 20 commits June 25, 2017 20:14

travis typo in path to where examples are

791239d

remove the zip file as well in grad laguna example

3ee08a7

be verbose in unpacking of google files

d75a9b9

temp remove some examples so travis re-uploads docs and examples

11b4fe3

bring back em examples for upload

aecce04

add dc back

d0ec1af

tmp remove maps, grav

f9f9a31

just build docs, don't nosetest

d86d1a8

upload artifacts created on travis for half of the examples

b5afbf4

simplify unpacking of docs archives

d7d5210

cd to the right place

61d7490

add back examples, use make html-noplot

16c0395

more pathing cleanups

cf48e80

try make html

e2c0e0b

restore all tests, restore use of nitpick on 2.7

e231b6d

remove aux bookpurnong files

6e587a0

restore python 3 tests

a012d84

use noplot for python 3 test

e4ac2ae

Merge pull request #621 from simpeg/tmp/upload-half-examples

4087241

cache docs examples

update ylim to be consistent with publication, remove auxilary file c…

bdb4f2a

…reated when tar unzipped

lheagy commented Jun 26, 2017

View reviewed changes

Change default interpolation to "nearest" to avoid logical error with…

4890374

… nan. Thumbs up of the merge - DF

lheagy commented Jun 27, 2017

View reviewed changes

thast added 2 commits June 27, 2017 13:52

add documentation to surface2ind_topo

dc4ac6c

add documentation surface2ind_topo

60f0f29

lheagy merged commit 1aab2bf into em/dev Jun 28, 2017

lheagy deleted the ref/tdem-testing branch June 28, 2017 02:03

lheagy mentioned this pull request Jun 28, 2017

em paper #622

Merged

2 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Ref/tdem testing #620

Ref/tdem testing #620

lheagy commented Jun 24, 2017

rowanc1 Jun 24, 2017

rowanc1 Jun 24, 2017

lheagy Jun 25, 2017

lheagy Jun 26, 2017

codecov bot commented Jun 24, 2017 •

edited

rowanc1 Jun 24, 2017

rowanc1 left a comment

rowanc1 Jun 24, 2017 •

edited

lheagy Jun 25, 2017

lheagy Jun 25, 2017

lheagy Jun 25, 2017

lheagy Jun 25, 2017

lheagy Jun 25, 2017

lheagy Jun 26, 2017

lheagy Jun 26, 2017

craigmillernz Jun 27, 2017

lheagy Jun 27, 2017

lheagy commented Jun 28, 2017

		@@ -1,160 +0,0 @@
		from __future__ import division, print_function

Ref/tdem testing #620

Ref/tdem testing #620

Conversation

lheagy commented Jun 24, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov bot commented Jun 24, 2017 • edited

Codecov Report

Choose a reason for hiding this comment

rowanc1 left a comment

Choose a reason for hiding this comment

rowanc1 Jun 24, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lheagy commented Jun 28, 2017

codecov bot commented Jun 24, 2017 •

edited

rowanc1 Jun 24, 2017 •

edited