Morphology Validation Use Case not working for Fast Spiking Interneurons

appukuttan-shailesh commented 5 years ago

Expected behavior

UseCase is expected to handle following cell types

D1-type Medium Spiny Neurons
D2-type Medium Spiny Neurons
Fast Spiking Interneurons

Actual Behavior (please include screenshot if possible)

Currently works fine for (1) and (2).... error when running for (3) neurite_Y_extent key error

Steps to reproduce the problem

Choose (3) as cell type and check output for soft constraints part of the Use Case

pedroernesto commented 5 years ago

I have the Python code to correct this issue, but it was not working on the Colllab, just locally. I am going to work on this in a few weeks.

alex4200 commented 5 years ago

@appukuttan-shailesh Is this ticket the same as ticket #386? Or what is the relationship?

Also, will this be fixed until the end of MS12 (end March) or only later?

appukuttan-shailesh commented 5 years ago

@alex4200 : As mentioned above, the UseCase is supposed to support multiple cell types (one of the options that a user chooses while running the UseCase). I have fixed the issues (as part of #386) such that it works for the first 2 cell types (D1 MSN, D2 MSN), but the problem with the third cell type (FS) is test implementation specific which needs to be resolved by the morpho_stats test creator (in this case @pedroernesto ). Therefore, it would be better for @pedroernesto to specify a deadline for this.

pedroernesto commented 5 years ago

Yes @alex4200, I will complete this before end of March

pedroernesto commented 5 years ago

Code completed. Issue almost done, need to test on the Collab next week.

pedroernesto commented 5 years ago

The problem with the third cell type (FSI) is solved. The source file for this use-case (BG_Morphology_Validations) runs now succesfully at: https://collab.humanbrainproject.eu/#/collab/8123/nav/78914

I have added the following functionalities to MorphoUnit:

Checks the compliance of the raw observation file with NeuroM (fst module) nomenclature
Automatic creation of configuration files for _morphstats and NeuroM, from that raw observation file
Deletes some neurite's morphometrics added by _morphstats, but not present in the observation file
Reversed to plural the number in feature names generated by _morphstats in the prediction file, to make it compatible with the observation file
Added more features that can be specified in the observation file. They can now be generated in the prediction file by means of other NeuroM modules than _morphstats, e.g. field diameter, bounding-box -X,Y,Z- extents and -largest,shortest- principal extents

alex4200 commented 5 years ago

@pedroernesto Is that notebook ready for testing? Is this the correct DEV notebook: /8123/usecases/BG_MorphoUnit_usecase.ipynb ?

pedroernesto commented 5 years ago

I think so, yes. The one on the BSP is not editable, just a copy on the user space. So I have been working on this one. Is it there another copy of the notebook on the Collab?

alex4200 commented 5 years ago

@pedroernesto I just tested the notebook located at /8123/usecases/BG_MorphoUnit_usecase.ipynb and I get an error in the second part:

Selection_664

My best guess is that you need a chance on how you create the morph_files:

model_collab_paths.append(model_instance["source"].split("http://collab.link")[1])

With that fix, the use case runs smoothly to the end.

However, when I run the use case with the following selections:

Option: D1-type Medium Spiny Neurons
Morphologies: [1,2]
model_uuid: db2cbbb4-4915-49ea-8b9a-4fa78746ab18

I get the following error:

Selection_665

Full error message:

AttributeErrorTraceback (most recent call last)
<ipython-input-34-7bce1e804a2f> in <module>()
      5                            model=morph_model, test_alias=test_alias, test_version="2.0",
      6                            storage_collab_id=collab_id, register_result=True,
----> 7                            client_obj=testLibrary, observation_path=obs_file_path)
      8 
      9 result_uuids.append(result_id)

/opt/conda/envs/python2/lib/python2.7/site-packages/hbp_validation_framework/utils.pyc in run_test(username, password, environment, model, test_instance_id, test_id, test_alias, test_version, storage_collab_id, register_result, client_obj, **params)
    362 def run_test(username="", password=None, environment="production", model="", test_instance_id="", test_id="", test_alias="", test_version="", storage_collab_id="", register_result=True, client_obj=None, **params):
    363     test_config_file = prepare_run_test_offline(username=username, password=password, environment=environment, test_instance_id=test_instance_id, test_id=test_id, test_alias=test_alias, test_version=test_version, client_obj=client_obj, **params)
--> 364     test_result_file = run_test_offline(model=model, test_config_file=test_config_file)
    365     result_id, score = upload_test_result(username=username, password=password, environment=environment, test_result_file=test_result_file, storage_collab_id=storage_collab_id, register_result=register_result, client_obj=client_obj)
    366     return result_id, score

/opt/conda/envs/python2/lib/python2.7/site-packages/hbp_validation_framework/utils.pyc in run_test_offline(model, test_config_file)
    243     # Run the test
    244     t_start = datetime.utcnow()
--> 245     score = test.judge(model, deep_error=True)
    246     t_end = datetime.utcnow()
    247 

/opt/conda/envs/python2/lib/python2.7/site-packages/sciunit/tests.pyc in judge(self, model, skip_incapable, stop_on_error, deep_error)
    233 
    234         if deep_error:
--> 235             score = self._judge(model, skip_incapable=skip_incapable)
    236         else:
    237             try:

/opt/conda/envs/python2/lib/python2.7/site-packages/sciunit/tests.pyc in _judge(self, model, skip_incapable)
    176 
    177         # 2.
--> 178         prediction = self.generate_prediction(model)
    179         self.check_prediction(prediction)
    180         self.last_model = model

/opt/conda/envs/python2/lib/python2.7/site-packages/morphounit/tests/test_NeuroM_MorphStats.pyc in generate_prediction(self, model, verbose)
    268 
    269         # Creating the prediction file with morph_stats
--> 270         self.path_test_output = model.morph_stats_output
    271         self.morp_path = model.morph_path
    272 

/opt/conda/envs/python2/lib/python2.7/site-packages/sciunit/models/base.pyc in __getattr__(self, attr)
     96             except:
     97                 raise AttributeError("Model %s has no attribute %s"
---> 98                                      % (self, attr))
     99         return result
    100 

AttributeError: Model 052913-7DL40xNJZ-rep-ax-cor has no attribute morph_stats_output

pedroernesto commented 5 years ago

In principle, the second part of the test is currently handling only FS neurons, not MS neurons. In fact, there is a cell just at the beginning to check this. Above it says 'Part 2: Soft Constraints Validation [Currently only available for Fast Spiking Interneurons]' More experimental data is needed for MS neurons

cnr-ibf-pa / hbp-bsp-issues