Segfault in basico when simulating BIOMD0000000516 after BIOMD0000000513.

[CiaranWelsh]

Hello, I've been trying to load and simulate the curated section of biomodels using basico. It seems there is a segfault when trying to simulate model 516 after 513 has already been loaded/simulated. Here's the code:

from basico import *
import os
import glob
import time
try:
    from roadrunner.tests.download_biomodels import download_biomodels
except ImportError:
    raise ImportError("Do a `pip install libroadrunner==2.2.0`")

biomodels_folder = os.path.join(os.path.dirname(__file__), "biomodels")

if not os.path.isdir(biomodels_folder):
    biomodels_folder = download_biomodels(os.path.dirname(__file__))

if not os.path.isdir(biomodels_folder):
    raise NotADirectoryError(biomodels_folder)

biomodels_files = glob.glob(os.path.join(biomodels_folder, "*.xml"))

print(len(biomodels_files))

start = time.time()
for model_file in [biomodels_files[512], biomodels_files[515]]:
    try:
        print(model_file)
        model = load_model(model_file)
        result = run_time_course(duration=100)
    except Exception:
        print(f"Model {model_file} failed")

end = time.time() - start

output:

1036
D:\roadrunner\roadrunner\docs\source\parallel\roadrunner_map\biomodels\BIOMD0000000513.xml
D:\roadrunner\roadrunner\docs\source\parallel\roadrunner_map\biomodels\BIOMD0000000516.xml

Process finished with exit code -1073740791 (0xC0000409)

[fbergmann]

thanks for letting us know, i will see what the issue would be.

[fbergmann]

Sorry for the inconvenience, for now please use the workaround:

for model_file in [biomodels_files[512], biomodels_files[515]]:
    try:
        print(model_file)
        model = load_model(model_file)
        result = run_time_course(duration=100)
        remove_datamodel(model)
        remove_user_defined_functions()
    except Exception:
        print(f"Model {model_file} failed")

until we can release a new COPASI version to fix this issue.

[CiaranWelsh]

Thanks Frank, that works now. As an FYI, basico/coapsi loaded and simulated all the models in biomodels (ignoring failures as above) in around 160 seconds. RoadRunner v2.2.0 did it in around 135, though I was able to get about 3 times performance improvement with multithreading (RoadRunnerMap). Do you know if basico/copasi will work with the multithreading module in Python, or will I run into pickling issues? Thanks

[fbergmann]

I have an example of using the multiprocessing module with basico here:

https://github.com/copasi/basico/blob/master/examples/parallel_scan.py

If you do encounter issues, please let me know.

[CiaranWelsh]

Cool. Thanks

[CiaranWelsh]

Sometimes the result = run_time_course(duration=100) line encounters numerical instability. Is there any way to detect this programatically so that I can count the failures?

[fbergmann]

currently, if an error is found, it is logged in the error category, so assuming you use this in test functions, you could use any of the test fixtures for logging:

https://testfixtures.readthedocs.io/en/latest/logging.html

i could add an option to throw an exception instead of logging.

[CiaranWelsh]

Would there be a way to retroactively check the number of errors logged? The following fails to count all models that have failed to load or simulate.

    start = time.time()
    failures = 0
    for model_file in biomodels_files:
        try:
            model = load_model(model_file)
            result = run_time_course(duration=100)
            remove_datamodel(model)
            remove_user_defined_functions()
        except Exception:
            failures += 1
            print(f"Model {model_file} failed")

    end = time.time() - start

    print(f"Took {end} seconds")
    print(f"failures: {failures}")

[CiaranWelsh]

Okay I see what you meant now. LogCapture works, thanks.

[fbergmann]

a version of the python-copasi package has been released that avoids the original segfault. closing this issue.