search

E3SM-Project / e3sm_diags

E3SM Diagnostics package
https://e3sm-project.github.io/e3sm_diags
BSD 3-Clause "New" or "Revised" License
39 stars 32 forks source link

Taylor Diagram errors #128

Closed kaabre closed 6 years ago

kaabre commented 6 years ago

I'm encountering a few weird errors running the new version of acme_diags on Edison. When running from the command line, I get a display variable error (that Susannah suggested may go away when X windows are used; this is indeed the case when I run on nomachine). I did not have this error before the update. However, even on nomachine where I don't get the display variable error, I get empty Taylor plots, and the csvs that should populate them are also empty.

The parameter file used and the outputs are located here: https://portal.nersc.gov/project/PNNL-PJR/bren937/scratch/TEST/newvers/

The logfiles are within each respective directory: one run on nomachine, one run on command line. (Even though both were created via batch scripts to the queue using the same parameters)

(I get the same issue running on Constance, although I know that's a machine you don't support. On acme2, I don't know what's going on but the acme_diags environment won't seem to update following the instructions you gave in the release notes.)

Are there any steps I am missing getting it to work on Edison/do you have expected output on that machine?

Thanks!

zshaheen commented 6 years ago

@kaabre Regarding the display variable error, we just released a new version, v1.2.1, that fixes this. It was because we had an import matplotlib statement without matplotlib.use('Agg') following it.

If you're using the non-dev env, using just conda upgrade acme_diags should get you the new version. Otherwise if on a dev env, just pull the latest code from master, install it, and it should be good.

chengzhuzhang commented 6 years ago

To-do for @zshaheen and myself

zshaheen commented 6 years ago

@kaabre Thanks for finding this bug! It seems like Taylor diagrams don't work for model vs model comparisons, for now.

kaabre commented 6 years ago

Thanks! Good to know -- I saw the related update -- I'm trying to run it now on a model-observation comparison and see what happens.

kaabre commented 6 years ago

@zshaheen @chengzhuzhang I DO still run into the "invalid DISPLAY variable" when using the command line (rather than through nomachine).

zshaheen commented 6 years ago

@kaabre Can you confirm that you're on v1.2.1? Please try the command python -c 'import acme_diags; print(acme_diags.__version__)'. It should print out v1.2.1

(acme_diags_env_old) shaheen2@aims4:~$ python -c 'import acme_diags; print(acme_diags.__version__)'
v1.2.1
kaabre commented 6 years ago

@zshaheen Hmm no despite running the conda update/upgrade acme_diags command it's still v1.2.0. Is there something else I need to do? I am in the non-dev environment.

zshaheen commented 6 years ago

Try the following two commands. Below is also the output:

Output:

(acme_diags_env_old) shaheen2@aims4:~$ conda remove acme_diags --force
Fetching package metadata ...................

Package plan for package removal in environment /export_2/shaheen2/anaconda2/envs/acme_diags_env_old:

The following packages will be REMOVED:

    acme_diags: 1.2.1-0 acme

Proceed ([y]/n)? y

(acme_diags_env_old) shaheen2@aims4:~$ conda install acme_diags -c conda-forge -c uvcdat -c acme
Fetching package metadata ...................
Solving package specifications: .

Package plan for installation in environment /export_2/shaheen2/anaconda2/envs/acme_diags_env_old:

The following NEW packages will be INSTALLED:

    acme_diags: 1.2.1-0 acme

Proceed ([y]/n)? y

(acme_diags_env_old) shaheen2@aims4:~$ python -c 'import acme_diags; print(acme_diags.__version__)'
v1.2.1
kaabre commented 6 years ago

It still only wants to install the 1.2.0 version for some reason:

(acme_diags_env) leebrent@edison11:> conda remove acme_diags --force
Solving environment: done ==> WARNING: A newer version of conda exists. <== current version: 4.4.7 latest version: 4.4.10 Please update conda by running $ conda update -n base conda Package Plan environment location: /global/homes/l/leebrent/.conda/envs/acme_diags_env removed specs:

zshaheen commented 6 years ago

Well, I guess the only thing to do is to delete your env and start again.

  1. source deactivate acme_diags_env
  2. conda remove -n acme_diags_env --all
  3. Follow the steps under 'Latest stable release' here.

That 'should' actually do it.

kaabre commented 6 years ago

All right, back to the beginning -- and it did! Version is now 1.2.1, I've submitted a test to make sure it works as expected.

zshaheen commented 6 years ago

@kaabre Nice, this is good to know!

chengzhuzhang commented 6 years ago

@kaabre I noticed that there are no html for tables generated at: https://portal.nersc.gov/project/PNNL-PJR/bren937/scratch/TEST/newvers/ However, when I try a model vs model run myself. Tables are actually generated...Would you let me know if you still missing the tables with v1.2.1 runs? Thanks!

kaabre commented 6 years ago

Hi Jill,

This is linked to issue #123 that I already noted – changing the name of the directories affects the path of the html tables. They still exist, but the links to them through the viewer are broken. Also, this output was run before the DISPLAY variable issue was fixed in version 1.2.1

The tables for the nomachine test are located here: https://portal.nersc.gov/project/PNNL-PJR/bren937/scratch/TEST/newvers/nomachinetest/viewer/table-data/ANN_metrics_table.html – basically you have to change the path manually to see them from the viewer links.

From: Jill Chengzhu Zhang notifications@github.com Reply-To: ACME-Climate/acme_diags reply@reply.github.com Date: Thursday, February 15, 2018 at 3:30 PM To: ACME-Climate/acme_diags acme_diags@noreply.github.com Cc: "Brent, Lee A" lee.brent@pnnl.gov, Mention mention@noreply.github.com Subject: Re: [ACME-Climate/acme_diags] Taylor Diagram errors (#128)

@kaabrehttps://github.com/kaabre I noticed that there are no html for tables generated at: https://portal.nersc.gov/project/PNNL-PJR/bren937/scratch/TEST/newvers/ However, when I try a model vs model run myself. Tables are actually generated...Would you let me know if you still missing the tables with v1.2.1 runs? Thanks!

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/ACME-Climate/acme_diags/issues/128#issuecomment-366097243, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AUsXw_RHnUNByNih8hbMQ3kTPFn0HgS1ks5tVL3pgaJpZM4SDFK2.

chengzhuzhang commented 6 years ago

Thank you, Lee! @kaabre We will work on the #123 and hopefully fix the issue all together...

kaabre commented 6 years ago

Appreciated!

From: Jill Chengzhu Zhang notifications@github.com Reply-To: ACME-Climate/acme_diags reply@reply.github.com Date: Thursday, February 15, 2018 at 4:17 PM To: ACME-Climate/acme_diags acme_diags@noreply.github.com Cc: "Brent, Lee A" lee.brent@pnnl.gov, Mention mention@noreply.github.com Subject: Re: [ACME-Climate/acme_diags] Taylor Diagram errors (#128)

Thank you, Lee! @kaabrehttps://github.com/kaabre We will work on the #123https://github.com/ACME-Climate/acme_diags/issues/123 and hopefully fix the issue all together...

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/ACME-Climate/acme_diags/issues/128#issuecomment-366106266, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AUsXw329yCPxhcQ2ZjbBi5u4G_y1nyeOks5tVMj2gaJpZM4SDFK2.

zshaheen commented 6 years ago

@chengzhuzhang @kaabre Just letting you both know, there's a fix for #123 here: 587e5af0521493c4264fc0babb927f150d60d980. I just need to merge that branch into master soon. I'll let you both know when that's done.

@kaabre We plan on another release sometime in early to mid March, and that will have the fix for #123.

chengzhuzhang commented 6 years ago

Pull request #139 resolves this issue