AssertionError in PyMC3 code using Deterministic distributions

Questions
8

Thanks a lot, it works indeed at the same speed as the original code (reassuring for PyMC!) with this implementation of of the log likelihood, I was not quite sure of how to implement that, even though I had seen it mentioned.

I had found this the issue #1925 in Github also, but even though I tried yesterday and today to implement the SMC sampler, I couldn’t get it to work with my example.

I have found this post but when I run it, I get this Error :

recover_existing_results() missing 1 required positional argument: ‘n_jobs’

I checked the Github and it seems the SMC sampler has been worked on a lot recently, so I was wondering if there could be a version difference between the PyMC used on the example and the one in the above mentioned post?

Is there a way to execute the test_smc.py present in the test folder (I have the last version of PyMC3, the 3.1) ? Or any other working example?
Because I tried some other ways with my example code here, like:

import pymc3 as pm
from pymc3.step_methods import smc

with basic_model:
    mtrace = smc.sample_smc(1000)

or

import pymc3 as pm
from pymc3.step_methods import smc

with basic_model:
    step=pm.smc()
    mtrace = pm.sample(1000, step=step)

but none of these worked, I got TypeError: Argument homepath should be path to result_directory. and TypeError: 'module' object is not callable for the second case.
I even tried to put new code from the PyMC/step_methods/smc.py file on the Github in my version, but it didn’t work either.