New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
desi_proc_tilenight processes skysub when --laststeps=skysub #2015
Conversation
3004b4a
to
7a9d122
Compare
@sbailey do the changes included in this PR reflect the behavior you are expecting for the pipeline? Specifically I am referring to the dependence on the option provided to the
If this looks good then we can go ahead and test for the example cited in #2007. Thanks. |
py/desispec/scripts/submit_night.py
Outdated
@@ -150,7 +150,8 @@ def submit_night(night, proc_obstypes=None, z_submit_types=None, queue='realtime | |||
|
|||
## If laststeps not defined, default is only LASTSTEP=='all' exposures for non-tilenight runs | |||
if laststeps is None: | |||
laststeps = ['all',] | |||
if not use_tilenight: | |||
laststeps = ['all',] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
when laststeps=None
and use_tilenight=True
, this block leaves laststeps=None
. This results in line 232 not selecting any exposures to process because nothing is in None:
good_exps = np.isin(np.array(etable['LASTSTEP']).astype(str), laststeps)
For example:
$> desi_run_night --dry-run -n 20230210 --z-submit-types cumulative --all-tiles
Processing the following obstypes: ['bias', 'dark', 'arc', 'flat', 'science', 'twilight', 'sci', 'dither']
Redshift fitting with redshift group types: ['cumulative']
Processing exposures with the following LASTSTEP's: None
INFO:tableio.py:334:load_table: Found table: /global/cfs/cdirs/desi/users/sjbailey/spectro/redux/pr2015/exposure_tables/202302/exposure_table_20230210.csv
INFO:tableio.py:337:load_table: Table /global/cfs/cdirs/desi/users/sjbailey/spectro/redux/pr2015/processing_tables/processing_table_pr2015-20230210.csv not found, creating new table of type proctable
INFO:submit_night.py:247:submit_night: Submitting cumulative redshifts for 0/0 tiles for which 20230210 is the last night: []
Completed submission of exposures for night 20230210.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@sbailey thanks for catching this. The most recent commit preserves the previous (correct) assignment to good_exps when laststeps
is None
, while also passing alaststeps
value of None
to submit_tilenight_and_redshifts
.
@sbailey I think I resolved the problem you cited above and the correct options are passed to
files like
which contains the line
are produced. Can you please verify that the behavior of the pipeline is as expected (including when processing without the |
Getting closer. The scripts look ok, but tilenight is still different from non-tilenight when it runs on tiles with a mix of all and skysub exposures. Consider tile 5627 on night 20230210:
When run with When run with I tested this in $CFS/desi/users/sjbailey/spectro/redux/pr2015 with a link to daily exposure_tables and commands like
Another corner case: the above generates three run/scripts/night/20230210/tilenight*.slurm scripts, from which I submit one. However, if I add the |
e0651b2
to
d7baf6a
Compare
@sbailey thanks for the feedback on what's expected for a given set of flags. After my most recent commit, exposure 167000 is processed only to sframes when
while it is not processed at all when the
I am still looking into the edge cases you cited and will leave it up to you whether to merge this now and address those as part of the upcoming refactor or wait until those have been addressed and then merge. |
@marcelo-alvarez thanks for the updates; this looks good. I agree with spinning off the other cases to separate tickets and not letting them block this. We can sort out their level of urgency in those tickets. |
Fixes #2007 by propagating the --laststeps option through to tilenight. I have tested with:
and I obtain the correct flags in the tilenight batch script:
@sbailey please take a look and see if this is the behavior you are expecting for the pipeline, and then we can proceed with further testing. I would also appreciate an extra set of eyes, as the "laststeps" argument had to be copied recursively to a significant depth of function calls. Thanks.