Pm.sample gets stuck after init with cores > 1

Questions
17

FWIW, having the exact same issue on MacOS.

Defined my logistic regression model and it works with cores = 1, but as soon as I try to increase the cores it just never gets started. Here’s where it hangs:

Auto-assigning NUTS sampler...
Initializing NUTS using adapt_diag...
Multiprocess sampling (2 chains in 4 jobs)
NUTS: [X1, X2, X3, X4, Intercept]
Sampling 2 chains, 0 divergences:   0%|          | 0/4000 [00:00<?, ?draws/s]

And this is the stack trace when I stop it after > 30 mins:

---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
~/opt/anaconda3/lib/python3.7/site-packages/pymc3/sampling.py in _mp_sample(draws, tune, step, chains, cores, chain, random_seed, start, progressbar, trace, model, **kwargs)
   1058             with sampler:
-> 1059                 for draw in sampler:
   1060                     trace = traces[draw.chain - chain]

~/opt/anaconda3/lib/python3.7/site-packages/pymc3/parallel_sampling.py in __iter__(self)
    393         while self._active:
--> 394             draw = ProcessAdapter.recv_draw(self._active)
    395             proc, is_last, draw, tuning, stats, warns = draw

~/opt/anaconda3/lib/python3.7/site-packages/pymc3/parallel_sampling.py in recv_draw(processes, timeout)
    283         pipes = [proc._msg_pipe for proc in processes]
--> 284         ready = multiprocessing.connection.wait(pipes)
    285         if not ready:

~/opt/anaconda3/lib/python3.7/multiprocessing/connection.py in wait(object_list, timeout)
    919             while True:
--> 920                 ready = selector.select(timeout)
    921                 if ready:

~/opt/anaconda3/lib/python3.7/selectors.py in select(self, timeout)
    414         try:
--> 415             fd_event_list = self._selector.poll(timeout)
    416         except InterruptedError:

KeyboardInterrupt: 

During handling of the above exception, another exception occurred:

ValueError                                Traceback (most recent call last)
<ipython-input-126-91d098b82ba5> in <module>
      1 with logistic_model:
----> 2     pred_trace = pm.sample(draws=1000, tune=1000, chains=2, cores=4, init='adapt_diag')

~/opt/anaconda3/lib/python3.7/site-packages/pymc3/sampling.py in sample(draws, step, init, n_init, start, trace, chain_idx, chains, cores, tune, progressbar, model, random_seed, discard_tuned_samples, compute_convergence_checks, **kwargs)
    467         _print_step_hierarchy(step)
    468         try:
--> 469             trace = _mp_sample(**sample_args)
    470         except pickle.PickleError:
    471             _log.warning("Could not pickle model, sampling singlethreaded.")

~/opt/anaconda3/lib/python3.7/site-packages/pymc3/sampling.py in _mp_sample(draws, tune, step, chains, cores, chain, random_seed, start, progressbar, trace, model, **kwargs)
   1078         return MultiTrace(traces)
   1079     except KeyboardInterrupt:
-> 1080         traces, length = _choose_chains(traces, tune)
   1081         return MultiTrace(traces)[:length]
   1082     finally:

~/opt/anaconda3/lib/python3.7/site-packages/pymc3/sampling.py in _choose_chains(traces, tune)
   1094     lengths = [max(0, len(trace) - tune) for trace in traces]
   1095     if not sum(lengths):
-> 1096         raise ValueError("Not enough samples to build a trace.")
   1097 
   1098     idxs = np.argsort(lengths)[::-1]

ValueError: Not enough samples to build a trace.