What's wrong with my scipt ?

[Yzxastr]

I try to add a 'sampler.emcee' to run the MCMC, but I get some error, and I don't know how to handle it. So, I want to ask how can I do to deal with this error?

This is my scipt:

b.add_solver('sampler.emcee', solver='emcee_solver',overwrite=True)

b.add_distribution({'q': phoebe.uniform(0.05,0.07),
                    'fillout_factor': phoebe.uniform(0.20,0.30),
                    'incl@binary': phoebe.uniform(70,80),
                    'teffratio': phoebe.uniform(0.9,0.98),
                    'pbdic': phoebe.uniform(0.05,0.07)},  # Custom Parameters( L2/L1:Ratio of luminosity )                
                   distribution='param_bounds2',overwrite_all=True)
b.set_value('priors@emcee_solver',['param_bounds2'])

b.set_value('priors@emcee_solver',['param_bounds'])
b.set_value('init_from', ['param_bounds2'])
b.set_value('nwalkers@emcee_solver@solver', value=20)
b.set_value('niters@emcee_solver@solver', value=500)
b.set_value('compute', solver='emcee_solver', value='phoebe01')
b.run_solver('emcee_solver', solution='emcee_sol',overwrite=True)

This is error:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Input In [25], in <cell line: 1>()
----> 1 b.run_solver('emcee_solver', solution='emcee_sol',overwrite=True)

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\parameters\parameters.py:424, in send_if_client.<locals>._send_if_client(self, *args, **kwargs)
    421     return ret_
    423 else:
--> 424     return fctn(self, *args, **kwargs)

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\frontend\bundle.py:13649, in Bundle.run_solver(self, solver, solution, detach, return_changes, **kwargs)
  13645     return job_param
  13648 solver_class = getattr(_solverbackends, '{}Backend'.format(solver_ps.kind.title()))
> 13649 params = solver_class().run(self, solver_ps.solver, compute, solution=solution, **{k:v for k,v in kwargs.items() if k not in ['compute']})
  13650 metawargs = {'context': 'solution',
  13651              'solver': solver_ps.solver,
  13652              'compute': compute,
  13653              'kind': solver_ps.kind,
  13654              'solution': solution}
  13657 comment_param = StringParameter(qualifier='comments', value=kwargs.get('comments', solver_ps.get_value(qualifier='comments', default='', **_skip_filter_checks)), description='User-provided comments for this solution.  Feel free to place any notes here.')

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\solverbackends\solverbackends.py:555, in BaseSolverBackend.run(self, b, solver, compute, **kwargs)
    552     rpacketlists_per_worker = [self.run_worker(**packet)]
    554 else:
--> 555     rpacketlists_per_worker = [self.run_worker(**packet)]
    557 logger.debug("rank:{}/{} calling _fill_solution".format(mpi.myrank, mpi.nprocs))
    558 return self._fill_solution(solution_ps, rpacketlists_per_worker)

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\solverbackends\solverbackends.py:1539, in EmceeBackend.run_worker(self, b, solver, compute, **kwargs)
   1536 expose_failed = kwargs.get('expose_failed')
   1538 logger.info("initializing {} walker positions".format(nwalkers))
-> 1539 dc, params_uniqueids, p0 = b.sample_distribution_collection(distribution=init_from,
   1540                                                             combine=init_from_combine,
   1541                                                             include_constrained=False,
   1542                                                             require_limits='limits' in init_from_requires,
   1543                                                             require_checks=compute if 'checks' in init_from_requires else False,
   1544                                                             require_compute=compute if 'compute' in init_from_requires else False,
   1545                                                             require_priors='priors@{}'.format(solver) if 'priors' in init_from_requires else False,
   1546                                                             sample_size=nwalkers,
   1547                                                             progressbar=kwargs.get('progressbar', False),
   1548                                                             return_dc_uniqueids_array=True,
   1549                                                             pool=pool
   1550                                                             )
   1552 params_uniqueids_and_indices = [_extract_index_from_string(uid) for uid in params_uniqueids]
   1553 params_twigs = [_to_twig_with_index(b.get_parameter(uniqueid=uniqueid, **_skip_filter_checks).twig, index) for uniqueid, index in params_uniqueids_and_indices]

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\frontend\bundle.py:8951, in Bundle.sample_distribution_collection(self, twig, sample_size, as_quantity, set_value, keys, parameters, **kwargs)
   8948         logger.warning("all distributions are delta, using sample_size=1 instead of sample_size={}".format(sample_size))
   8949         N = 1
-> 8951 sampled_values = dc.sample(size=sample_size).T
   8953 if require_checks or require_compute or (require_limits and not require_limits_success) or (require_priors and not require_priors_success):
   8954     def _progress(*args):

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\dependencies\distl\distl.py:4280, in DistributionCollection.sample(self, *args, **kwargs)
   4278 if kwargs.get('as_quantity', False):
   4279     array_kwargs['dtype'] = _np.dtype(object)
-> 4280 samples = _np.asarray([dist.sample(*args, seed=seeds, cache_sample=cache_sample, **sample_kwargs) for dist in self.dists], **array_kwargs).T
   4282 if cache_sample:
   4283     self._cached_sample = samples

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\dependencies\distl\distl.py:4280, in <listcomp>(.0)
   4278 if kwargs.get('as_quantity', False):
   4279     array_kwargs['dtype'] = _np.dtype(object)
-> 4280 samples = _np.asarray([dist.sample(*args, seed=seeds, cache_sample=cache_sample, **sample_kwargs) for dist in self.dists], **array_kwargs).T
   4282 if cache_sample:
   4283     self._cached_sample = samples

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\dependencies\distl\distl.py:4845, in Composite.sample(self, size, unit, as_quantity, wrap_at, seed, as_univariate, cache_sample)
   4842     for d in self.dists:
   4843         d.clear_cached_sample()
-> 4845     return super(Composite, self).sample(size=size, unit=unit, as_quantity=as_quantity, wrap_at=wrap_at, seed=seed, cache_sample=cache_sample)
   4846 else:
   4847     # NOTE: even though in these cases we sample from the underlying children
   4848     # (and therefore can account for covariances from multivariate children),
   4849     # calls to pdf/cdf/ppf will still need to merge and interpolate
   4850     # and will ignore these covariances.
   4852     sample = self._sample_from_children(self.math, self.dists, size=size, seed=seed, cache_sample=cache_sample)

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\dependencies\distl\distl.py:2753, in BaseUnivariateDistribution.sample(self, size, unit, as_quantity, wrap_at, seed, cache_sample)
   2750     _np.random.seed(seed)
   2752 qs = _np.random.random(size=size)
-> 2753 sample = self.ppf(qs, wrap_at=False)
   2754 if cache_sample:
   2755     self._cached_sample = sample

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\dependencies\distl\distl.py:2262, in BaseUnivariateDistribution.ppf(self, q, unit, as_quantity, wrap_at)
   2226 """
   2227 Expose the percent point function (ppf; iverse of cdf - percentiles) at
   2228 values of `q`.
   (...)
   2259 * (float or array) ppf values of the same type/shape as `q`
   2260 """
   2261 try:
-> 2262     ppf = self.dist_constructor_object.ppf(q)
   2263 except AttributeError:
   2264     raise NotImplementedError("{} does not support ppf".format(self.__class__.__name__))

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\dependencies\distl\distl.py:1082, in BaseDistribution.dist_constructor_object(self)
   1068 """
   1069 Return the instantiated underlying distribution constructor (often the
   1070 scipy.stats random variable object).
   (...)
   1079 * object
   1080 """
   1081 if self._dist_constructor_object_cache is None:
-> 1082     self._dist_constructor_object_cache = self.dist_constructor_func(*self.dist_constructor_args)
   1084 return self._dist_constructor_object_cache

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\phoebe\dependencies\distl\distl.py:4734, in Composite.dist_constructor_args(self)
   4730 pdf_call = _stats_custom.interpolate_callable(x, pdf)
   4732 if cdf is None:
   4733     # print("*** integrating to compute cdf over spline (this may be slow... we'll eventually cache this so it only needs to be done once until an attribute is changed)")
-> 4734     spline = _interpolate.UnivariateSpline(x, pdf, k=1, s=0)
   4735     def _spline_int_single(xi):
   4736         return spline.integral(x[0], xi)

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\scipy\interpolate\_fitpack2.py:199, in UnivariateSpline.__init__(self, x, y, w, bbox, k, s, ext, check_finite)
    196 def __init__(self, x, y, w=None, bbox=[None]*2, k=3, s=None,
    197              ext=0, check_finite=False):
--> 199     x, y, w, bbox, self.ext = self.validate_input(x, y, w, bbox, k, s, ext,
    200                                                   check_finite)
    202     # _data == x,y,w,xb,xe,k,s,n,t,c,fp,fpint,nrdata,ier
    203     data = dfitpack.fpcurf0(x, y, k, w=w, xb=bbox[0],
    204                             xe=bbox[1], s=s)

File C:\app\AND\envs\name_of_phoebe_environment\lib\site-packages\scipy\interpolate\_fitpack2.py:227, in UnivariateSpline.validate_input(x, y, w, bbox, k, s, ext, check_finite)
    225 else:
    226     if not np.all(diff(x) > 0.0):
--> 227         raise ValueError("x must be strictly increasing if s = 0")
    228 if x.size != y.size:
    229     raise ValueError("x and y should have a same length")

ValueError: x must be strictly increasing if s = 0

​

[kecnry]

The sampling of the distribution collection is failing, although I'm not entirely sure why since the distributions you show seem reasonably defined. Can you please try removing distributions one at a time from the add_distribution call to see if a single one of them is to blame? It also looks like you have param_bounds defined elsewhere and setting that as your priors - if you remove the priors does the error still get raised? How is param_bounds defined?

[YmcGW]

First, I'm sorry I forgot to remove the extra code, 'b.set_value('priors@emcee_solver',['param_bounds'])', and I actually only used 'b.set_value('priors@emcee_solver',['param_bounds2'])' in my scipt.
But, I also get same error when I add this code 'b.set_value('priors@emcee_solver',['param_bounds2'])' in my scipt, and then I try to remove it from my scipt, my scipt runs successfully. So, I want to know why that is.

[YmcGW]

Is it because the version is updated, but I am learning the tutorials of the old version?

[kecnry]

Ok, if I understood you correctly, that narrows it down to applying the priors being at fault. Have you tried removing parameters from the distribution collection one at a time to see if a single one (perhaps the custom parameter) is to blame? If that doesn't narrow it down, can you either post the snippet used to create that parameter, a full script to reproduce, or send me a copy of your bundle (b.save('filename.bundle'])) by email?

You may be able to get around the bug by removing 'priors' from init_from_requires, but I'd still like to understand what is actually going on to try to fix the underlying bug.