# ValueError: expected an ndarray

**URL:** <https://discourse.pymc.io/t/valueerror-expected-an-ndarray/187>\
**Category:** Questions\
**Tags:** theano\
**Created:** [July 24, 2017, 6:32pm UTC](https://discourse.pymc.io/t/valueerror-expected-an-ndarray/187 "2017-07-24T18:32:08Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![zahybnaya](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/zahybnaya/32/81_2.png) [@zahybnaya](https://discourse.pymc.io/u/zahybnaya)\
**Post date:** [July 24, 2017, 6:32pm UTC](https://discourse.pymc.io/t/valueerror-expected-an-ndarray/187/1 "2017-07-24T18:32:08Z")

</div>

Hello,  
I am confused about why this happens when I try to call `sample` with a `step` argument -  
a `ValueError: expected an ndarray` is thrown.  
Here is my model followed by the traceback.

```
from pymc3 import Model, DiscreteUniform, Uniform, Discrete, Deterministic, sample, loo, Metropolis
import theano.tensor as tt
import theano
from theano.compile.ops import as_op
from rushhour import min_manhattan_distance, opt_solution_instances
from astar import LRTA, eval_theano
import numpy as np
from test import instance_set
from csv import DictReader
from random import random
from sys import argv

theano.config.exception_verbosity='high'
theano.config.optimizer='fast_compile'
# Two parameters 
try:
    subject=argv[1]
    path_file=argv[2]
except:
    print 'Args: <subject> <path_file>'

class IVS(Discrete):
    def __init__ (self,h_epsilon,learning_iter,exp_select, *args, **kwargs):
        super(IVS, self). __init__ (*args, **kwargs)#What does discrete expects?
        self.h_epsilon = h_epsilon
        self.learning_iter = learning_iter
        self.exp = exp_select

    def logp(self, value):
        return logp(self.h_epsilon,self.learning_iter,self.exp,value)

def get_instance_names_by_subject(path_file,subject):
    with open(path_file,'rb') as f:
        reader=DictReader(f)
        data=[d['instance'] for d in reader if d['subject']==subject and d['complete']=='True']
        return sorted(data)

def get_paths_by_subject(path_file,instance_names,subject,fun):
    with open(path_file,'rb') as f:
        reader=DictReader(f)
        data=[(d['instance'],fun(d)) for d in reader if d['subject']==subject and d['instance'] in instance_names]
        return np.array([fr for (x,fr) in sorted(data, key=lambda x: instance_names.index(x[0]))])

def make_model_path_length(i):
    def model_path_length(h_epsilon,learning_iter,exp):
        path,_= LRTA(i,heur=lambda x: (1+h_epsilon)*min_manhattan_distance(x),update_h=True,iters=learning_iter,exp=exp)
        return len(path)
    return model_path_length

@as_op(itypes=[tt.dscalar,tt.lscalar,tt.dscalar,tt.lvector], otypes=[tt.dscalar])
def logp(h_epsilon,learning_iter,exp,sub_path_length):
    lp=0
    trials=1;
    for i,sr in zip(instances,sub_path_length):
        m_path_f = make_model_path_length(i)
        while True:
            r=m_path_f(h_epsilon,learning_iter,exp)
            if r == sr or trials > max_trials:
                break
            trials+=1
        print '{0},{1},{2},{3},{4},{5}'.format(subject,i.name,h_epsilon,learning_iter,exp,trials)
        sample_data[(i.name,str(h_epsilon),str(learning_iter))]=trials
        lp+=sum([(1./t) for t in range(1,trials+1)])
        trials=1
    return -lp

lrta_model=Model()
with lrta_model as lrta_model:
    max_trials=20
    instance_names=get_instance_names_by_subject(path_file,subject)
    instances=np.array([i for i in instance_set if i.name in instance_names]) #is this sorted properly?
    d=get_paths_by_subject(path_file,instance_names,subject,lambda x:int(x['human_length']))
    sample_data={}
    learning_iter=DiscreteUniform('learning_iter',lower=1, upper=5)
    exp_select = Uniform('exp', lower=4., upper=20.)
    h_epsilon = Uniform('h_epsilon',lower=0.,upper=1.)
    path_length=IVS('path_length',h_epsilon,learning_iter,exp_select, observed=d)
    step1=Metropolis(vars=[h_epsilon])
    step2=Metropolis(vars=[learning_iter])
    step3=Metropolis(vars=[exp_select])
    print ' *************calling sample************'
    trace = sample(2,step=[step1,step2,step3])
    print loo(trace,pointwise=True)

```

And this is the traceback-

```
Traceback (most recent call last):
  File "lrta_pymc3.py", line 85, in <module>
trace = sample(2,step=[step1,step2,step3])
  File "/usr/local/lib/python2.7/site-packages/pymc3/sampling.py", line 278, in sample
return sample_func(**sample_args)[discard:]
  File "/usr/local/lib/python2.7/site-packages/pymc3/sampling.py", line 293, in _sample
for it, strace in enumerate(sampling):
  File "/usr/local/lib/python2.7/site-packages/tqdm/_tqdm.py", line 862, in __iter__
for obj in iterable:
  File "/usr/local/lib/python2.7/site-packages/pymc3/sampling.py", line 391, in _iter_sample
point, states = step.step(point)
  File "/usr/local/lib/python2.7/site-packages/pymc3/step_methods/compound.py", line 24, in step
point, state = method.step(point)
  File "/usr/local/lib/python2.7/site-packages/pymc3/step_methods/arraystep.py", line 152, in step
apoint, stats = self.astep(bij.map(point))
  File "/usr/local/lib/python2.7/site-packages/pymc3/step_methods/metropolis.py", line 152, in astep
accept = self.delta_logp(q, q0)
  File "/usr/local/lib/python2.7/site-packages/theano/compile/function_module.py", line 898, in __call__
storage_map=getattr(self.fn, 'storage_map', None))
  File "/usr/local/lib/python2.7/site-packages/theano/gof/link.py", line 325, in raise_with_op
reraise(exc_type, exc_value, exc_trace)
  File "/usr/local/lib/python2.7/site-packages/theano/compile/function_module.py", line 884, in __call__
self.fn() if output_subset is None else\
ValueError: expected an ndarray
Apply node that caused the error: Elemwise{add,no_inplace}(Elemwise{switch,no_inplace}.0, Elemwise{add,no_inplace}.0, Elemwise{add,no_inplace}.0, FromFunctionOp{logp}.0)
Toposort index: 65
Inputs types: [TensorType(float64, scalar), TensorType(float64, scalar), TensorType(float64, scalar), TensorType(float64, scalar)]
Inputs shapes: [(), (), (), 'No shapes']
Inputs strides: [(), (), (), 'No strides']
Inputs values: [array(-1.6094379124341003), array(-1.3862943611198906), array(-1.8960286450343948), -4.019877344877345]
Inputs type_num: [12, 12, 12, '']
Outputs clients: [[Elemwise{sub,no_inplace}(Elemwise{add,no_inplace}.0, Elemwise{add,no_inplace}.0)]]

Backtrace when the node is created(use Theano flag traceback.limit=N to make it longer):
  File "lrta_pymc3.py", line 81, in <module>
step1=Metropolis(vars=[h_epsilon])
  File "/usr/local/lib/python2.7/site-packages/pymc3/step_methods/metropolis.py", line 126, in __init__
self.delta_logp = delta_logp(model.logpt, vars, shared)
  File "/usr/local/lib/python2.7/site-packages/pymc3/memoize.py", line 16, in memoizer
cache[key] = obj(*args, **kwargs)
  File "/usr/local/lib/python2.7/site-packages/pymc3/model.py", line 447, in logpt
return tt.add(*map(tt.sum, factors))

Debugprint of the apply node:
Elemwise{add,no_inplace} [id A] <TensorType(float64, scalar)> ''
 |Elemwise{switch,no_inplace} [id B] <TensorType(float64, scalar)> ''
 | |Elemwise{mul,no_inplace} [id C] <TensorType(int8, scalar)> ''
 | | |Elemwise{mul,no_inplace} [id D] <TensorType(int8, scalar)> ''
 | | | |TensorConstant{1} [id E] <TensorType(int8, scalar)>
 | | | |Elemwise{mul,no_inplace} [id F] <TensorType(int8, scalar)> ''
 | | | |TensorConstant{1} [id E] <TensorType(int8, scalar)>
 | | | |Elemwise{ge,no_inplace} [id G] <TensorType(bool, scalar)> ''
 | | | |learning_iter_shared [id H] <TensorType(int64, scalar)>
 | | | |TensorConstant{1} [id I] <TensorType(int32, scalar)>
 | | |Elemwise{mul,no_inplace} [id J] <TensorType(int8, scalar)> ''
 | | |TensorConstant{1} [id E] <TensorType(int8, scalar)>
 | | |Elemwise{le,no_inplace} [id K] <TensorType(bool, scalar)> ''
 | | |learning_iter_shared [id H] <TensorType(int64, scalar)>
 | | |TensorConstant{5} [id L] <TensorType(int32, scalar)>
 | |TensorConstant{-1.60943791243} [id M] <TensorType(float64, scalar)>
 | |TensorConstant{-inf} [id N] <TensorType(float32, scalar)>
 |Elemwise{add,no_inplace} [id O] <TensorType(float64, scalar)> ''
 | |Elemwise{switch,no_inplace} [id P] <TensorType(float32, scalar)> ''
 | | |Elemwise{mul,no_inplace} [id Q] <TensorType(int8, scalar)> ''
 | | | |Elemwise{mul,no_inplace} [id R] <TensorType(int8, scalar)> ''
 | | | | |TensorConstant{1} [id E] <TensorType(int8, scalar)>
 | | | | |Elemwise{mul,no_inplace} [id S] <TensorType(int8, scalar)> ''
 | | | | |TensorConstant{1} [id E] <TensorType(int8, scalar)>
 | | | | |Elemwise{ge,no_inplace} [id T] <TensorType(bool, scalar)> ''
 | | | | |Elemwise{add,no_inplace} [id U] <TensorType(float64, scalar)> ''
 | | | | | |Elemwise{mul,no_inplace} [id V] <TensorType(float64, scalar)> ''
 | | | | | | |TensorConstant{16.0} [id W] <TensorType(float32, scalar)>
 | | | | | | |sigmoid [id X] <TensorType(float64, scalar)> ''
 | | | | | | |exp_interval___shared [id Y] <TensorType(float64, scalar)>
 | | | | | |TensorConstant{4.0} [id Z] <TensorType(float32, scalar)>
 | | | | |TensorConstant{4.0} [id Z] <TensorType(float32, scalar)>
 | | | |Elemwise{mul,no_inplace} [id BA] <TensorType(int8, scalar)> ''
 | | | |TensorConstant{1} [id E] <TensorType(int8, scalar)>
 | | | |Elemwise{le,no_inplace} [id BB] <TensorType(bool, scalar)> ''
 | | | |Elemwise{add,no_inplace} [id U] <TensorType(float64, scalar)> ''
 | | | |TensorConstant{20.0} [id BC] <TensorType(float32, scalar)>
 | | |TensorConstant{-2.77258872986} [id BD] <TensorType(float32, scalar)>
 | | |TensorConstant{-inf} [id N] <TensorType(float32, scalar)>
 | |Elemwise{sub,no_inplace} [id BE] <TensorType(float64, scalar)> ''
 | |Elemwise{sub,no_inplace} [id BF] <TensorType(float64, scalar)> ''
 | | |TensorConstant{2.77258872986} [id BG] <TensorType(float32, scalar)>
 | | |Elemwise{mul,no_inplace} [id BH] <TensorType(float64, scalar)> ''
 | | |TensorConstant{2} [id BI] <TensorType(int8, scalar)>
 | | |softplus [id BJ] <TensorType(float64, scalar)> ''
 | | |Elemwise{neg,no_inplace} [id BK] <TensorType(float64, scalar)> ''
 | | |exp_interval___shared [id Y] <TensorType(float64, scalar)>
 | |exp_interval___shared [id Y] <TensorType(float64, scalar)>
 |Elemwise{add,no_inplace} [id BL] <TensorType(float64, scalar)> ''
 | |Elemwise{switch,no_inplace} [id BM] <TensorType(float32, scalar)> ''
 | | |Elemwise{mul,no_inplace} [id BN] <TensorType(int8, scalar)> ''
 | | | |Elemwise{mul,no_inplace} [id BO] <TensorType(int8, scalar)> ''
 | | | | |TensorConstant{1} [id E] <TensorType(int8, scalar)>
 | | | | |Elemwise{mul,no_inplace} [id BP] <TensorType(int8, scalar)> ''
 | | | | |TensorConstant{1} [id E] <TensorType(int8, scalar)>
 | | | | |Elemwise{ge,no_inplace} [id BQ] <TensorType(bool, scalar)> ''
 | | | | |Elemwise{add,no_inplace} [id BR] <TensorType(float64, scalar)> ''
 | | | | | |Elemwise{mul,no_inplace} [id BS] <TensorType(float64, scalar)> ''
 | | | | | | |TensorConstant{1.0} [id BT] <TensorType(float32, scalar)>
 | | | | | | |sigmoid [id BU] <TensorType(float64, scalar)> ''
 | | | | | | |Subtensor{int64} [id BV] <TensorType(float64, scalar)> ''
 | | | | | | |Subtensor{int64:int64:} [id BW] <TensorType(float64, (True,))> ''
 | | | | | | | |inarray1 [id BX] <TensorType(float64, (True,))>
 | | | | | | | |Constant{0} [id BY] <int64>
 | | | | | | | |Constant{1} [id BZ] <int64>
 | | | | | | |Constant{0} [id BY] <int64>
 | | | | | |TensorConstant{0.0} [id CA] <TensorType(float32, scalar)>
 | | | | |TensorConstant{0.0} [id CA] <TensorType(float32, scalar)>
 | | | |Elemwise{mul,no_inplace} [id CB] <TensorType(int8, scalar)> ''
 | | | |TensorConstant{1} [id E] <TensorType(int8, scalar)>
 | | | |Elemwise{le,no_inplace} [id CC] <TensorType(bool, scalar)> ''
 | | | |Elemwise{add,no_inplace} [id BR] <TensorType(float64, scalar)> ''
 | | | |TensorConstant{1.0} [id BT] <TensorType(float32, scalar)>
 | | |TensorConstant{0.0} [id CA] <TensorType(float32, scalar)>
 | | |TensorConstant{-inf} [id N] <TensorType(float32, scalar)>
 | |Elemwise{sub,no_inplace} [id CD] <TensorType(float64, scalar)> ''
 | |Elemwise{sub,no_inplace} [id CE] <TensorType(float64, scalar)> ''
 | | |TensorConstant{0.0} [id CA] <TensorType(float32, scalar)>
 | | |Elemwise{mul,no_inplace} [id CF] <TensorType(float64, scalar)> ''
 | | |TensorConstant{2} [id BI] <TensorType(int8, scalar)>
 | | |softplus [id CG] <TensorType(float64, scalar)> ''
 | | |Elemwise{neg,no_inplace} [id CH] <TensorType(float64, scalar)> ''
 | | |Subtensor{int64} [id BV] <TensorType(float64, scalar)> ''
 | |Subtensor{int64} [id BV] <TensorType(float64, scalar)> ''
 |FromFunctionOp{logp} [id CI] <TensorType(float64, scalar)> ''
   |ViewOp [id CJ] <TensorType(float64, scalar)> 'h_epsilon'
   |learning_iter_shared [id H] <TensorType(int64, scalar)>
   |ViewOp [id CK] <TensorType(float64, scalar)> 'exp'
   |TensorConstant{[13 11]} [id CL] <TensorType(int64, vector)>

Storage map footprint:
 - TensorConstant{[13 11]}, Shape: (2,), ElemSize: 8 Byte(s), TotalSize: 16 Byte(s)
 - Elemwise{switch,no_inplace}.0, Shape: (), ElemSize: 8 Byte(s), TotalSize: 8.0 Byte(s)
 - Elemwise{add,no_inplace}.0, Shape: (), ElemSize: 8 Byte(s), TotalSize: 8.0 Byte(s)
 - Constant{0}, Shape: (), ElemSize: 8 Byte(s), TotalSize: 8.0 Byte(s)
 - Constant{1}, Shape: (), ElemSize: 8 Byte(s), TotalSize: 8.0 Byte(s)
 - inarray1, Input, Shape: (1,), ElemSize: 8 Byte(s), TotalSize: 8 Byte(s)
 - inarray, Input, Shape: (1,), ElemSize: 8 Byte(s), TotalSize: 8 Byte(s)
 - learning_iter_shared, Shared Input, Shape: (), ElemSize: 8 Byte(s), TotalSize: 8.0 Byte(s)
 - exp_interval___shared, Shared Input, Shape: (), ElemSize: 8 Byte(s), TotalSize: 8.0 Byte(s)
 - TensorConstant{-1.60943791243}, Shape: (), ElemSize: 8 Byte(s), TotalSize: 8.0 Byte(s)
 - Elemwise{add,no_inplace}.0, Shape: (), ElemSize: 8 Byte(s), TotalSize: 8.0 Byte(s)
 - exp, Shape: (), ElemSize: 8 Byte(s), TotalSize: 8.0 Byte(s)
 - TensorConstant{5}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{-inf}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{1}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{1.0}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{0.0}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{20.0}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{4.0}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{2.77258872986}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{-2.77258872986}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{16.0}, Shape: (), ElemSize: 4 Byte(s), TotalSize: 4.0 Byte(s)
 - TensorConstant{2}, Shape: (), ElemSize: 1 Byte(s), TotalSize: 1.0 Byte(s)
 - TensorConstant{1}, Shape: (), ElemSize: 1 Byte(s), TotalSize: 1.0 Byte(s)
 TotalSize: 146.0 Byte(s) 0.000 GB
 TotalSize inputs: 114.0 Byte(s) 0.000 GB
```

---

<div class="post-metadata">

**Author:** ![aseyboldt](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/aseyboldt/32/5795_2.png) [@aseyboldt](https://discourse.pymc.io/u/aseyboldt)\
**Post date:** [July 24, 2017, 6:40pm UTC](https://discourse.pymc.io/t/valueerror-expected-an-ndarray/187/2 "2017-07-24T18:40:56Z")

</div>

Could you try casting the return value in logp to a numpy scalar: `return np.array(-lp)`  
I think the problem is that it expects a numpy float64, but you return a python float.

---

<div class="post-metadata">

**Author:** ![zahybnaya](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/zahybnaya/32/81_2.png) [@zahybnaya](https://discourse.pymc.io/u/zahybnaya)\
**Post date:** [July 24, 2017, 7:00pm UTC](https://discourse.pymc.io/t/valueerror-expected-an-ndarray/187/3 "2017-07-24T19:00:09Z")

</div>

That was the problem! Thanks! It is working now.