# Dimension error when using pm.Potential

**URL:** <https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649>\
**Category:** v5\
**Tags:** modeling\
**Created:** [August 3, 2023, 1:19pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649 "2023-08-03T13:19:25Z")\
**Posts on this page:** 10\
**Page:** 1

<div class="post-metadata">

**Author:** ![BlueMan66](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/blueman66/32/2924_2.png) [@BlueMan66](https://discourse.pymc.io/u/BlueMan66)\
**Post date:** [August 3, 2023, 1:19pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/1 "2023-08-03T13:19:25Z")

</div>

Hi! I’m trying to calibrate a physical model using PyMC. I’ve gone with the approach of wrapping my jax model in a PyTensor Op. This works fine with the Metropolis sampler which is finding some issues when sampling. Since I want to take advantage of NUTS, I implemented a grad method and a grad Op for my model:

```auto
@jax.jit
def impedance_logp(observed, theta, ws):
    """Compute the marginal log-likelihood of a single HMM process."""

    (h1, h2, sigma, mu1, mu2) = theta[0]
    model = impedance_operation(ws, h1, h2, mu1, mu2)
    logp_impedance = -(0.5 / sigma**2) * jnp.sum((observed - model)** 2)

    return jnp.array(logp_impedance)

class ImpedanceOp(Op):
    itypes = [pt.dvector] # expects a vector of parameter values when called
    otypes = [pt.dscalar] # outputs a single scalar value (the log likelihood)

    def __init__ (self, ws, data, **kwargs):
        super(). __init__ (**kwargs)
        self.ws = ws
        self.data = data
        self.logpgrad = ImpedanceGrad(self.data, self.ws)

    def impedance_logp(self, theta):
        return impedance_logp(self.data, theta, self.ws)

    def perform(self, node, inputs, outputs):
        # (h,) = inputs
        result = self.impedance_logp(inputs)
        outputs[0][0] = np.asarray(result, dtype=node.outputs[0].dtype)

    def grad(self, inputs, output_gradients):
        grads = self.logpgrad(*inputs)
        output_gradient = output_gradients[0]
        return [output_gradient * grads]

class ImpedanceGrad(Op):
    itypes = [pt.dvector]
    otypes = [pt.dvector]

    def __init__ (self, data, ws) -> None:
        super(). __init__ ()
        self.data = data
        self.ws = ws

    def impedance_logp(self, theta):
        return impedance_logp(self.data, theta, self.ws)

    def impedance_logp_grad(self, theta):
        return jax.grad(self.impedance_logp)(theta)

    def perform(self, node, inputs, outputs):
        grads = self.impedance_logp_grad(inputs)
        outputs[0][0] = np.asarray(grads, dtype=node.outputs[0].dtype)

```

And here is the model I’m trying to sample:

```auto
with pm.Model() as impedance_calibration:
    h1 = pm.HalfNormal("h1", sigma=1)
    h2 = pm.HalfNormal("h2", sigma=1)
    mu1 = pm.HalfNormal("mu1", sigma=1)
    mu2 = pm.HalfNormal("mu2", sigma=1)
    # sigma = pm.TruncatedNormal("sigma", sigma=1, mu=5, lower=0, upper=10)
    sigma = pm.HalfNormal("sigma", sigma=1)
    theta = pt.as_tensor_variable([h1, h2, sigma, mu1, mu2])
    likelihood = pm.Potential(
        "likelihood",
        impedance_mean(theta),
    )

    if not os.path.exists(TRACE_PATH):
        # step = pm.Metropolis()
        # trace = pm.sample(step=step, tune=10000, draws=5000)
        trace = pm.sample(tune=10000, draws=5000)
        # trace = pm.sample_smc(draws=5000, chains=4, parallel=True)
        trace.to_netcdf(TRACE_PATH)
    else:
        trace = az.from_netcdf(TRACE_PATH)

```

Problem is that sampling with nuts yields the following error:

```auto
pymc.sampling.parallel.ParallelSamplingError: Chain 0 failed with: Expected 1 dimensions input
Apply node that caused the error: Subtensor{i}(ImpedanceGrad.0, 0)
Toposort index: 10
Inputs types: [TensorType(float64, shape=(None,)), ScalarType(uint8)]
Inputs shapes: [(1, 5), ()]
Inputs strides: [(40, 8), ()]
Inputs values: [array([[ 4.21017600e+08, -1.44086746e+09, 2.44285594e+09,
        -1.65140378e+09, -8.00240256e+08]]), 0]
Outputs clients: [[Composite{(((i0 + i1) * i2) + i3)}(Composite{...}.2, Subtensor{i}.0, h1, (d __logp/dh1_log___ logprob){1.0})]]

```

Which I need some help to interpret exactly which of the variables is causing trouble  
Thanks for the help!

---

<div class="post-metadata">

**Author:** ![BlueMan66](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/blueman66/32/2924_2.png) [@BlueMan66](https://discourse.pymc.io/u/BlueMan66)\
**Post date:** [August 3, 2023, 2:10pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/2 "2023-08-03T14:10:51Z")

</div>

pm.sample is passing an additional integer parameter to my pytensor grad which is causing the issue

```auto
  File "/home/blueman69/git/vibra-analyzer/impedance_diana.py", line 84, in <module>
    trace = pm.sample(tune=10000, draws=5000, chains=1, cores=1)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pymc/sampling/mcmc.py", line 785, in sample
    _sample_many(**sample_args)
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pymc/sampling/mcmc.py", line 927, in _sample_many
    _sample(
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pymc/sampling/mcmc.py", line 1000, in _sample
    for it, diverging in enumerate(sampling):
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/fastprogress/fastprogress.py", line 50, in __iter__
    raise e
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/fastprogress/fastprogress.py", line 41, in __iter__
    for i,o in enumerate(self.gen):
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pymc/sampling/mcmc.py", line 1068, in _iter_sample
    point, stats = step.step(point)
                   ^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pymc/step_methods/arraystep.py", line 174, in step
    return super().step(point)
           ^^^^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pymc/step_methods/arraystep.py", line 100, in step
    apoint, stats = self.astep(q)
                    ^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pymc/step_methods/hmc/base_hmc.py", line 168, in astep
    start = self.integrator.compute_state(q0, p0)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pymc/step_methods/hmc/integration.py", line 56, in compute_state
    logp, dlogp = self._logp_dlogp_func(q)
                  ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pymc/model.py", line 373, in __call__
    cost, *grads = self._pytensor_function(*grad_vars)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

```

---

<div class="post-metadata">

**Author:** ![ricardoV94](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/ricardov94/32/5775_2.png) [@ricardoV94](https://discourse.pymc.io/u/ricardoV94)\
**Post date:** [August 3, 2023, 2:40pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/3 "2023-08-03T14:40:12Z")

</div>

Have you tried the utility `verify_grad`? Should help you debug the gradient with less layers in between. It’s suggested throughout this blogpost in case it helps: [How to use JAX ODEs and Neural Networks in PyMC - PyMC Labs](https://www.pymc-labs.io/blog-posts/jax-functions-in-pymc-3-quick-examples/)

---

<div class="post-metadata">

**Author:** ![BlueMan66](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/blueman66/32/2924_2.png) [@BlueMan66](https://discourse.pymc.io/u/BlueMan66)\
**Post date:** [August 3, 2023, 3:49pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/4 "2023-08-03T15:49:40Z")

</div>

Thanks for the reply! That utility has indeed been helpful. However I’m finding the issue that I can’t get the correct inputs to my model:

```auto
TypeError: impedance_matrix() missing 4 required positional arguments: 'h1', 'h2', 'mu1', and 'mu2'
Apply node that caused the error: ImpedanceGrad(input 0)
Toposort index: 0
Inputs types: [TensorType(float64, shape=(5,))]
Inputs shapes: [(5,)]
Inputs strides: [(8,)]
Inputs values: [array([0., 1., 2., 3., 4.])]
Outputs clients: [[Mul(ExpandDims{axis=0}.0, ImpedanceGrad.0)]]

```

During execution, empty values will be pased to my gradfun resulting in the following:

```auto
VALUE PASSED=> [1.73490671 1.44366054 0.82801176 0.75424598]
VALUE PASSED=> [1.73490671 1.44366054 0.82801176 0.75424598]
VALUE PASSED=> [1.73490671 1.44366054 0.82801176 0.75424598]
VALUE PASSED=> []

```

I’m getting 4 values since I dropped one when calculating my model (the last one was the standard deviation). When the empty value gets passed I get an error and the code crashes

---

<div class="post-metadata">

**Author:** ![ricardoV94](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/ricardov94/32/5775_2.png) [@ricardoV94](https://discourse.pymc.io/u/ricardoV94)\
**Post date:** [August 3, 2023, 4:51pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/5 "2023-08-03T16:51:22Z")

</div>

Difficult to say without being able to run the code myself. You can consider not packing the 5 inputs in a vector/tuple and instead make you Op have 5 separate inputs.

Maybe that will help you pinpoint the source of the issue?

---

<div class="post-metadata">

**Author:** ![BlueMan66](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/blueman66/32/2924_2.png) [@BlueMan66](https://discourse.pymc.io/u/BlueMan66)\
**Post date:** [August 3, 2023, 5:51pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/6 "2023-08-03T17:51:55Z")

</div>

I re wrote the function to add those changes. Here are the files I use to check the gradient:  
[impedance\_matrix\_new.py](https://discourse.pymc.io/uploads/short-url/mpspZYmdqyELW4ms7SE9dZPZyu8.py) (5.4 KB)  
[test\_gradient.py](https://discourse.pymc.io/uploads/short-url/1w2IRp3IMCg8k8zOIyrfVSoZe62.py) (1.3 KB)  
This now fails with:

```auto
Traceback (most recent call last):
  File "/home/blueman69/git/vibra-analyzer/test_gradient.py", line 36, in <module>
    pytensor.gradient.verify_grad(
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pytensor/gradient.py", line 1838, in verify_grad
    symbolic_grad = grad(cost, tensor_pt, disconnected_inputs="ignore")
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pytensor/gradient.py", line 617, in grad _rval: Sequence[Variable] = _populate_grad_dict(
                                ^^^^^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pytensor/gradient.py", line 1420, in _populate_grad_dict
    rval = [access_grad_cache(elem) for elem in wrt]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pytensor/gradient.py", line 1420, in <listcomp>
    rval = [access_grad_cache(elem) for elem in wrt]
            ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pytensor/gradient.py", line 1375, in access_grad_cache
    term = access_term_cache(node)[idx]
           ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/blueman69/mambaforge/envs/pymc_env/lib/python3.11/site-packages/pytensor/gradient.py", line 1213, in access_term_cache
    raise ValueError(
ValueError: ImpedanceOp returned the wrong number of gradient terms.

```

If you can check this it will help me a lot!

---

<div class="post-metadata">

**Author:** ![ricardoV94](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/ricardov94/32/5775_2.png) [@ricardoV94](https://discourse.pymc.io/u/ricardoV94)\
**Post date:** [August 3, 2023, 6:07pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/7 "2023-08-03T18:07:57Z")

</div>

Your Grad Op will now have to return 5 outputs instead of 1 (one for each input). I haven’t looked at the code, but I imagine that’s the problem you see now from the error message.

---

<div class="post-metadata">

**Author:** ![BlueMan66](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/blueman66/32/2924_2.png) [@BlueMan66](https://discourse.pymc.io/u/BlueMan66)\
**Post date:** [August 3, 2023, 6:41pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/8 "2023-08-03T18:41:52Z")

</div>

That indeed was the problem, after correcting the make\_node method of the grad op to the following:

```auto
def make_node(self, h1, h2, mu1, mu2, sigma, gz) -> Apply:
     inputs = [
         pt.as_tensor_variable(h1),
         pt.as_tensor_variable(h2),
         pt.as_tensor_variable(mu1),
         pt.as_tensor_variable(mu2),
         pt.as_tensor_variable(sigma),
         pt.as_tensor_variable(gz),
     ]
     outputs = [
         inputs[0].type(),
         inputs[0].type(),
         inputs[0].type(),
         inputs[0].type(),
         inputs[0].type(),
     ]
     return Apply(self, inputs, outputs)

```

I get the same error.  
The following image is the debugger output. It seems to be passing a list of a list of vectors instead of a list of vectors as it should. I don’t know how to correct this.

 ![image](https://canada1.discourse-cdn.com/flex036/uploads/pymc3/original/2X/b/b155d6a51e37d3844665294e1acb0b052412b695.png)

---

<div class="post-metadata">

**Author:** ![BlueMan66](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/blueman66/32/2924_2.png) [@BlueMan66](https://discourse.pymc.io/u/BlueMan66)\
**Post date:** [August 4, 2023, 1:29pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/9 "2023-08-04T13:29:46Z")

</div>

Bump cause of urgency!

---

<div class="post-metadata">

**Author:** ![ricardoV94](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/ricardov94/32/5775_2.png) [@ricardoV94](https://discourse.pymc.io/u/ricardoV94)\
**Post date:** [August 4, 2023, 1:37pm UTC](https://discourse.pymc.io/t/dimension-error-when-using-pm-potential/12649/10 "2023-08-04T13:37:02Z")

</div>

You are still not returning the right outputs from the gradient. Here is a simple Op which multiplies two inputs and uses another Op to compute the gradients

```python
import numpy as np

import pytensor
import pytensor.tensor as pt
from pytensor.graph.op import Op
from pytensor.graph.basic import Apply

class MulScalars(Op):

    def make_node(self, x, y):
        inputs = [
            pt.as_tensor(x),
            pt.as_tensor(y),
        ]
        # Todo: Check inputs have the right type
        outputs = [inputs[0].type()]

        return Apply(self, inputs, outputs)

    def perform(self, node, inputs, output_storage):
        [x, y] = inputs
        result = x * y
        output_storage[0][0] = np.asarray(result)

    def grad(self, inputs, output_grads):
        [x, y] = inputs
        [g_output] = output_grads
        [grad_x, grad_y] = grad_mul_scalars(x, y, g_output)
        return [grad_x, grad_y]

class GradMulScalars(Op):

    def make_node(self, x, y, g_output):
        inputs = [
            pt.as_tensor(x), 
            pt.as_tensor(y),
            pt.as_tensor(g_output),
        ]
        # TODO: Check inputs have the right type
        outputs = [
            inputs[0].type(), 
            inputs[1].type(),
        ]
        return Apply(self, inputs, outputs)

    def perform(self, node, inputs, output_storage):
        [x, y, g_output] = inputs
        res_x = y * g_output
        res_y = x * g_output
        output_storage[0][0] = res_x
        output_storage[1][0] = res_y

mul_scalars = MulScalars()
grad_mul_scalars = GradMulScalars()

pytensor.gradient.verify_grad(
    mul_scalars, 
    pt=(np.array(5.0), np.array(2.0)), 
    rng=np.random.default_rng(),
)

```

If this is still unclear I suggest checking these other two guides as well:

[https://pytensor.readthedocs.io/en/latest/extending/creating\_an\_op.html](https://pytensor.readthedocs.io/en/latest/extending/creating_an_op.html)

> **[How to wrap a JAX function for use in PyMC](https://www.pymc.io/projects/examples/en/latest/case_studies/wrapping_jax_function.html)**
>
> Extra dependencies install instructions In order to run this notebook (either locally or on binder) you won’t only need a working PyMC installation with all optional dependencies, but also to insta...
