Data demonstration

[mnmneE]

Hi there,

after implementing the free_water task;

, I tried to show the data but I couldn't find how to do it in the terminal with both the blink and the free water tasks, I am getting an error related to plot parameters even though I don't need to plot and no instruction in the code to plot any data (please find attached a screenshot for the error);
I followed the documentation where it is mentioned that in order to show the data the user will be able to access the data from the run with :meth:.Subject.get_trial_data; can you please provide one example about it; will the data be shown automatically if I don't have the plot error; or I should explicitly use the method in the terminal ;

yet I am blocked because of the error related to the plot parameters, I tried also to show the data using HDF viewer but it didn't work;

thanks for your help.
regards

[sneakers-the-rat]

It looks like there are a few errors in that screenshot -- the first two are from not having a subject selected in the subject list in the gui (the list next to the start/stop button) when pressing start/stop button.

the second is from a misspecified PLOT dictionary in the task - this is documented here: https://docs.auto-pi-lot.com/en/main/guide/task.html#plot. In this case it's sort of odd because a preceding line checks if the PLOT dictionary is empty, which is the typical way to specify a plotless task, and returns quietly. So the dictionary must have something in it, but not have a data subdictionary. I just modified the behavior of the code here 9a8c0e6 so that if there is no data key it continues quietly.

It sounds like you are talking about two different things though, the PLOT dictionary specifies how the plot drawn in the GUI looks while the task is being performed. In other words, it maps the task data (like the target, whether the subject got the trial correct, etc) to graphical primitives like points and lines. eg. "draw the correct value as a point at 0 or 1." I haven't tried to robustly support post-collection plotting since pandas and matplotlib cover that better than I could already.

To get the trial data, yes you would use Subject.get_trial_data(). That would look like

from autopilot.core.subject import Subject

# instantiate the subject class with the subject id/name of the subject
# it should find the .h5 file automatically, but you can also pass the directory, etc
# see the docs page for more instantiation params
sub = Subject('subject_id')

# return data from the most recent step in the training protocol
last_step = sub.get_trial_data()

# get all data from all steps
all_data = sub.get_trial_data('all')

from there you could use pandas DataFrame plotting methods or whatever method you choose.

Does that help? lmk if i missed your question :)

[sneakers-the-rat]

as an aside, yes the hdf viewer is very buggy for me as well.

[mnmneE]

Thanks a lot, surly it is very helpful and informative.

[mnmneE]

I tried to import Subject from autopilot core yet it throws the following error related to task list;

ImportError : cannot import name 'TASK_LIST' from 'autopilot.tasks'
should I modify and can you please tell me what parameters may solve the issue

In advance thanks very much for your help

[mnmneE]

I just commented the import line in the core script and it worked.

[sneakers-the-rat]

try updating your autopilot version, the latest version (>0.4.0) does away with the task list concept in favor of a more general registry system, documented here :) https://docs.auto-pi-lot.com/en/latest/guide/plugins.html