Repository navigation
get_flow crashes when downloading a flow from a different sklearn version #569
Description
example code (run with sklearn 0.20 installed):
openml.flows.get_flow(6969)
When deciding whether to deserialize the model, it checks the dependencies according to a different schedule than when it actually deserialized the model. In the prior case, it just checks whether the correct library is available. In the latter case, it also checks the version number.
personally, I we should fix this before releasing a next version, as it will affect many users.
traceback:
/home/janvanrijn/anaconda3/envs/openml-python-contrib/bin/python /home/janvanrijn/projects/openml-python-contrib/testjan.py
Traceback (most recent call last):
File "/home/janvanrijn/projects/openml-python-contrib/testjan.py", line 3, in <module>
openml.flows.get_flow(6969)
File "/home/janvanrijn/projects/openml-python/openml/flows/functions.py", line 23, in get_flow
flow = OpenMLFlow._from_dict(flow_dict)
File "/home/janvanrijn/projects/openml-python/openml/flows/flow.py", line 321, in _from_dict
model = flow_to_sklearn(flow)
File "/home/janvanrijn/projects/openml-python/openml/flows/sklearn_converter.py", line 173, in flow_to_sklearn
rval = _deserialize_model(o, initialize_with_defaults)
File "/home/janvanrijn/projects/openml-python/openml/flows/sklearn_converter.py", line 442, in _deserialize_model
_check_dependencies(flow.dependencies)
File "/home/janvanrijn/projects/openml-python/openml/flows/sklearn_converter.py", line 516, in _check_dependencies
'%s not satisfied.' % dependency_string)
ValueError: Trying to deserialize a model with dependency sklearn==0.20.0 not satisfied.
Activity
janvanrijn commentedon Oct 9, 2018
Also the error message is wrong. As flow 6969 was ran with 0.18.X.
janvanrijn commentedon Oct 9, 2018
We should probably put the dependency check in a more central place. And decide on the policy.
Fix #569: crash when sklearn version does not collide (#601)
mfeurer commentedon Dec 7, 2018
Fixed via #601.