Error with KerasLayer and Amplitude Embedding

Laia_Domingo · January 18, 2024, 4:01pm

Hi there! I’m getting an error when trying to create a Keras layer with AmplitudeEmbedding. It has to do with casting the input to complex128. Here’s my code:

import pennylane as qml
from pennylane import numpy as qnp
from pennylane.transforms import richardson_extrapolate, fold_global
import numpy as np
import tensorflow as tf

# Hyperparameters of the circui
nqbits=4
depth=1

# Device definition
dev_ideal = qml.device('default.mixed', wires=nqbits)
dev_mixed = qml.transforms.insert(dev_ideal, qml.DepolarizingChannel, 0.1) # Adding noise

@qml.transforms.mitigate_with_zne([1, 2, 3], fold_global, richardson_extrapolate) # Adding error mitigation
@qml.qnode(dev_mixed)
def mitigated_qnode(inputs, weights):
    qml.AmplitudeEmbedding(features=inputs, wires=range(nqbits),normalize=True)
    qml.templates.StronglyEntanglingLayers(weights, wires=range(nqbits))
    return [qml.expval(qml.PauliZ(wires=i)) for i in range(nqbits)]

# Create keras model
weight_shapes = {"weights": (depth, nqbits,3)}
qlayer = qml.qnn.KerasLayer(mitigated_qnode, weight_shapes, output_dim=nqbits)

# Make a prediction
inputs = qnp.random.normal(2, 4,(1, 2**nqbits), requires_grad=False)
qlayer(inputs)

And here’s the full error message below:

ValueError                                Traceback (most recent call last)
<ipython-input-32-37ce10b2c2b7> in <cell line: 29>()
     27 # Make a prediction
     28 inputs = qnp.random.normal(2, 4,(1, 2**nqbits), requires_grad=False)
---> 29 qlayer(inputs)

15 frames
/usr/local/lib/python3.10/dist-packages/pennylane/math/single_dispatch.py in <lambda>(x, **kwargs)
    266 
    267 ar.register_function(
--> 268     "tensorflow", "asarray", lambda x, **kwargs: _i("tf").convert_to_tensor(x, **kwargs)
    269 )
    270 ar.register_function(

ValueError: Exception encountered when calling layer "keras_layer_7" (type KerasLayer).

Tensor conversion requested dtype complex128 for Tensor with dtype float32: <tf.Tensor: shape=(16,), dtype=float32, numpy=
array([ 0.32765773,  0.04465814, -0.05177495, -0.19328678,  0.28668392,
        0.40516296, -0.09485355, -0.10624942,  0.05358212,  0.16164567,
       -0.00984542,  0.45717812,  0.21692467,  0.3813239 , -0.3816654 ,
        0.08793645], dtype=float32)>

Call arguments received:
  • inputs=tf.Tensor(shape=(1, 16), dtype=float32)

I’m using Google Colab to test this, here’s the output of qml.about:

Name: PennyLane
Version: 0.34.0
Summary: PennyLane is a Python quantum machine learning library by Xanadu Inc.
Home-page: https://github.com/PennyLaneAI/pennylane
Author: 
Author-email: 
License: Apache License 2.0
Location: /usr/local/lib/python3.10/dist-packages
Requires: appdirs, autograd, autoray, cachetools, networkx, numpy, pennylane-lightning, requests, rustworkx, scipy, semantic-version, toml, typing-extensions
Required-by: PennyLane-Lightning

Platform info:           Linux-6.1.58+-x86_64-with-glibc2.35
Python version:          3.10.12
Numpy version:           1.23.5
Scipy version:           1.11.4
Installed devices:
- lightning.qubit (PennyLane-Lightning-0.34.0)
- default.gaussian (PennyLane-0.34.0)
- default.mixed (PennyLane-0.34.0)
- default.qubit (PennyLane-0.34.0)
- default.qubit.autograd (PennyLane-0.34.0)
- default.qubit.jax (PennyLane-0.34.0)
- default.qubit.legacy (PennyLane-0.34.0)
- default.qubit.tf (PennyLane-0.34.0)
- default.qubit.torch (PennyLane-0.34.0)
- default.qutrit (PennyLane-0.34.0)
- null.qubit (PennyLane-0.34.0)

Thanks!!

isaacdevlugt · January 19, 2024, 10:51pm

Hey @Laia_Domingo,

Thanks! I was able to replicate your error and was a little perplexed. I tried running your code without ZNE stuff and with regular default.qubit and it worked, but simply using default.mixed instead is causing problems. You’ve found a bug!

I made an issue on our repository: [BUG] Tensorflow type promotion error with `default.mixed` · Issue #5085 · PennyLaneAI/pennylane · GitHub. We will update you here when there’s a fix!

Laia_Domingo · January 22, 2024, 8:04am

Hi @isaacdevlugt,

Thank you for your help!

Laia_Domingo · January 22, 2024, 10:39am

Hi again!

Unable to create a Keras layer with ZNE, I attempted to create a Torch layer instead, encountering a new issue. The code functions correctly when the batch size is set to 1, but an error arises when I increase the batch size beyond 1. While such errors are typically associated with qnodes not utilizing input parameters, this is not the source of the problem in this scenario. Interestingly, removing the ZNE component resolves the error. Here’s my code:

import pennylane as qml
from pennylane import numpy as qnp
from pennylane.transforms import richardson_extrapolate, fold_global
import numpy as np
import torch

# Hyperparameters of the circui
nqbits=4
depth=1
batch_size = 2

# Device definition
dev_ideal = qml.device('default.mixed', wires=nqbits)
dev_mixed = qml.transforms.insert(dev_ideal, qml.DepolarizingChannel, 0.1) # Adding noise


@qml.transforms.mitigate_with_zne([1, 2, 3], fold_global, richardson_extrapolate) # Adding error mitigation
@qml.qnode(dev_mixed)
def mitigated_qnode(inputs, weights):
    qml.AmplitudeEmbedding(features=inputs, wires=range(nqbits),normalize=True)
    qml.templates.StronglyEntanglingLayers(weights, wires=range(nqbits))
    return [qml.expval(qml.PauliZ(wires=i)) for i in range(nqbits)]

# Creating pytorch quantum layer
weight_shapes = {"weights": (depth, nqbits,3)}
qlayer = qml.qnn.TorchLayer(mitigated_qnode, weight_shapes)

# Running layer with batch_size >1
inputs = torch.tensor(qnp.random.normal(2, 4,(batch_size, 2**nqbits)))
qlayer(inputs)

And here’s the error I get:

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
Cell In[46], line 30
     28 # Running layer with batch_size >1
     29 inputs = torch.tensor(qnp.random.normal(2, 4,(batch_size, 2**nqbits)))
---> 30 qlayer(inputs)

File c:\Users\laiad\anaconda3\envs\Quantum4\Lib\site-packages\torch\nn\modules\module.py:1501, in Module._call_impl(self, *args, **kwargs)
   1496 # If we don't have any hooks, we want to skip the rest of the logic in
   1497 # this function, and just call forward.
   1498 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   1499         or _global_backward_pre_hooks or _global_backward_hooks
   1500         or _global_forward_hooks or _global_forward_pre_hooks):
-> 1501     return forward_call(*args, **kwargs)
   1502 # Do not call functions when jit is used
   1503 full_backward_hooks, non_full_backward_hooks = [], []

File c:\Users\laiad\anaconda3\envs\Quantum4\Lib\site-packages\pennylane\qnn\torch.py:402, in TorchLayer.forward(self, inputs)
    399     inputs = torch.reshape(inputs, (-1, inputs.shape[-1]))
    401 # calculate the forward pass as usual
--> 402 results = self._evaluate_qnode(inputs)
    404 # reshape to the correct number of batch dims
    405 if has_batch_dim:

File c:\Users\laiad\anaconda3\envs\Quantum4\Lib\site-packages\pennylane\qnn\torch.py:423, in TorchLayer._evaluate_qnode(self, x)
...
-> 1100     return _VF.tensordot(a, b, dims_a, dims_b)  # type: ignore[attr-defined]
   1101 else:
   1102     return _VF.tensordot(a, b, dims_a, dims_b, out=out)

RuntimeError: contracted dimensions need to match, but first has size 3 in dim -1 and second has size 4 in dim -2

Any idea of what is happening here? Thanks!!

isaacdevlugt · January 23, 2024, 11:09pm

Hey @Laia_Domingo,

Everything works fine without the error mitigation transform, but when it’s introduced then we get problems. I’ll get back to you on this

isaacdevlugt · January 24, 2024, 3:07pm

Hey @Laia_Domingo,

If you add the qml.transforms.broadcast_expand transform it should work.

# Hyperparameters of the circui
nqbits=4
depth=1
batch_size = 2

# Device definition
dev_ideal = qml.device('default.mixed', wires=nqbits)
dev_mixed = qml.transforms.insert(dev_ideal, qml.DepolarizingChannel, 0.1) # Adding noise


#@qml.qnode(dev_ideal)
@qml.transforms.mitigate_with_zne([1, 2, 3], fold_global, richardson_extrapolate) # Adding error mitigation
@qml.transforms.broadcast_expand
@qml.qnode(dev_mixed)
def mitigated_qnode(inputs, weights):

    qml.AmplitudeEmbedding(features=inputs, wires=range(nqbits),normalize=True)
    qml.templates.StronglyEntanglingLayers(weights, wires=range(nqbits))

    return [qml.expval(qml.PauliZ(wires=i)) for i in range(nqbits)]

# Creating pytorch quantum layer
weight_shapes = {"weights": (depth, nqbits, 3)}
qlayer = qml.qnn.TorchLayer(mitigated_qnode, weight_shapes)

# Running layer with batch_size >1
inputs = torch.tensor(qnp.random.normal(2, 4,(batch_size, 2**nqbits)))
qlayer(inputs)

The problem is that error mitigation doesn’t support broadcasting — a bug! 2 bugs in one forum post… excellent work ! I made a bug report here and we’ll make sure to update you with progress when it happens.

github.com/PennyLaneAI/pennylane

[BUG] Error mitigation transform does not support broadcasting

opened 03:07PM - 24 Jan 24 UTC

isaacdevlugt

bug

### Expected behavior A QNode transformed with `mitigate_with_zne` should accep…t parameters with a leading dimension. ### Actual behavior A QNode transformed with `mitigate_with_zne` does not accept parameters with a leading dimension. Adding `qml.transforms.broadcast_expand` fixes it. ### Additional information Reported here: https://discuss.pennylane.ai/t/error-with-keraslayer-and-amplitude-embedding/3924 Please update this forum post when there is progress made. ### Source code ```shell import pennylane as qml from pennylane import numpy as np from pennylane.transforms import richardson_extrapolate, fold_global nqbits=4 batch_size = 2 dev_ideal = qml.device('default.mixed', wires=nqbits) dev_mixed = qml.transforms.insert(dev_ideal, qml.DepolarizingChannel, 0.1) # Adding noise #@qml.transforms.broadcast_expand @qml.qnode(dev_mixed) def original_qnode(inputs): qml.AmplitudeEmbedding(features=inputs, wires=range(nqbits), normalize=True) return [qml.expval(qml.PauliZ(wires=i)) for i in range(nqbits)] mitigated_qnode = qml.transforms.mitigate_with_zne(original_qnode, [1,2,3], fold_global, richardson_extrapolate) inputs = np.random.uniform(0, 1, size=(batch_size, 2**nqbits)) mitigated_qnode(inputs) ``` ### Tracebacks ```shell --------------------------------------------------------------------------- ValueError Traceback (most recent call last) Cell In[16], line 20 17 mitigated_qnode = qml.transforms.mitigate_with_zne(original_qnode, [1,2,3], fold_global, richardson_extrapolate) 19 inputs = np.random.uniform(0, 1, size=(batch_size, 2**nqbits)) ---> 20 mitigated_qnode(inputs) File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/qnode.py:1039, in QNode.__call__(self, *args, **kwargs) 1034 full_transform_program._set_all_argnums( 1035 self, args, kwargs, argnums 1036 ) # pylint: disable=protected-access 1038 # pylint: disable=unexpected-keyword-arg -> 1039 res = qml.execute( 1040 (self._tape,), 1041 device=self.device, 1042 gradient_fn=self.gradient_fn, 1043 interface=self.interface, 1044 transform_program=full_transform_program, 1045 config=config, 1046 gradient_kwargs=self.gradient_kwargs, 1047 override_shots=override_shots, 1048 **self.execute_kwargs, 1049 ) 1051 res = res[0] 1053 # convert result to the interface in case the qfunc has no parameters File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/interfaces/execution.py:649, in execute(tapes, device, gradient_fn, interface, transform_program, config, grad_on_execution, gradient_kwargs, cache, cachesize, max_diff, override_shots, expand_fn, max_expansion, device_batch_transform, device_vjp) 647 if no_interface_boundary_required: 648 results = inner_execute(tapes) --> 649 return post_processing(results) 651 _grad_on_execution = False 653 if config.use_device_jacobian_product and interface in jpc_interfaces: File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/interfaces/execution.py:641, in execute.<locals>.post_processing(results) 640 def post_processing(results): --> 641 return program_post_processing(program_pre_processing(results)) File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/transforms/core/transform_program.py:86, in _apply_postprocessing_stack(results, postprocessing_stack) 63 """Applies the postprocessing and cotransform postprocessing functions in a Last-In-First-Out LIFO manner. 64 65 Args: (...) 83 84 """ 85 for postprocessing in reversed(postprocessing_stack): ---> 86 results = postprocessing(results) 87 return results File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/transforms/core/transform_program.py:56, in _batch_postprocessing(results, individual_fns, slices) 30 def _batch_postprocessing( 31 results: ResultBatch, individual_fns: List[PostProcessingFn], slices: List[slice] 32 ) -> ResultBatch: 33 """Broadcast individual post processing functions onto their respective tapes. 34 35 Args: (...) 54 55 """ ---> 56 return tuple(fn(results[sl]) for fn, sl in zip(individual_fns, slices)) File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/transforms/core/transform_program.py:56, in <genexpr>(.0) 30 def _batch_postprocessing( 31 results: ResultBatch, individual_fns: List[PostProcessingFn], slices: List[slice] 32 ) -> ResultBatch: 33 """Broadcast individual post processing functions onto their respective tapes. 34 35 Args: (...) 54 55 """ ---> 56 return tuple(fn(results[sl]) for fn, sl in zip(individual_fns, slices)) File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/transforms/mitigate.py:541, in mitigate_with_zne.<locals>.processing_fn(results) 536 for i in range(0, len(results), reps_per_factor): 537 # The stacking ensures the right interface is used 538 # averaging over axis=0 is critical because the qnode may have multiple outputs 539 results_flattened.append(mean(qml.math.stack(results[i : i + reps_per_factor]), axis=0)) --> 541 extrapolated = extrapolate(scale_factors, results_flattened, **extrapolate_kwargs) 543 extrapolated = extrapolated[0] if shape(extrapolated) == (1,) else extrapolated 545 # unstack the results in the case of multiple measurements File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/transforms/mitigate.py:320, in richardson_extrapolate(x, y) 297 def richardson_extrapolate(x, y): 298 r"""Polynomial fit where the degree of the polynomial is fixed to being equal to the length of ``x``. 299 300 In a nutshell, this function is calling :func:`~.pennylane.transforms.poly_extrapolate` with ``order = len(x)-1``. (...) 318 319 """ --> 320 return poly_extrapolate(x, y, len(x) - 1) File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/transforms/mitigate.py:293, in poly_extrapolate(x, y, order) 269 def poly_extrapolate(x, y, order): 270 r"""Extrapolator to :math:`f(0)` for polynomial fit. 271 272 The polynomial is defined as ``f(x) = p[0] * x**deg + p[1] * x**(deg-1) + ... + p[deg]`` such that ``deg = order + 1``. (...) 291 292 """ --> 293 coeff = _polyfit(x, y, order) 294 return coeff[-1] File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/transforms/mitigate.py:264, in _polyfit(x, y, order) 262 c = qml.math.linalg.pinv(qml.math.transpose(X) @ X) 263 c = c @ qml.math.transpose(X) --> 264 c = qml.math.dot(c, y) 265 c = qml.math.transpose(qml.math.transpose(c) / scale) 266 return c File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/math/multi_dispatch.py:151, in multi_dispatch.<locals>.decorator.<locals>.wrapper(*args, **kwargs) 148 interface = interface or get_interface(*dispatch_args) 149 kwargs["like"] = interface --> 151 return fn(*args, **kwargs) File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/pennylane/math/multi_dispatch.py:358, in dot(tensor1, tensor2, like) 354 return x @ y 356 return np.tensordot(x, y, axes=[[-1], [-2]], like=like) --> 358 return np.dot(x, y, like=like) File ~/.virtualenvs/pennylane-torch/lib/python3.11/site-packages/autoray/autoray.py:80, in do(fn, like, *args, **kwargs) 31 """Do function named ``fn`` on ``(*args, **kwargs)``, peforming single 32 dispatch to retrieve ``fn`` based on whichever library defines the class of 33 the ``args[0]``, or the ``like`` keyword argument if specified. (...) 77 <tf.Tensor: id=91, shape=(3, 3), dtype=float32> 78 """ 79 backend = choose_backend(fn, *args, like=like, **kwargs) ---> 80 return get_lib_fn(backend, fn)(*args, **kwargs) ValueError: shapes (3,3) and (3,4,2) not aligned: 3 (dim 1) != 4 (dim 1) ``` ### System information ```shell Name: PennyLane Version: 0.34.0 Summary: PennyLane is a Python quantum machine learning library by Xanadu Inc. Home-page: https://github.com/PennyLaneAI/pennylane Author: Author-email: License: Apache License 2.0 Location: /Users/isaac/.virtualenvs/pennylane-torch/lib/python3.11/site-packages Requires: appdirs, autograd, autoray, cachetools, networkx, numpy, pennylane-lightning, requests, rustworkx, scipy, semantic-version, toml, typing-extensions Required-by: PennyLane-Lightning Platform info: macOS-14.2.1-arm64-arm-64bit Python version: 3.11.6 Numpy version: 1.26.2 Scipy version: 1.11.4 Installed devices: - lightning.qubit (PennyLane-Lightning-0.34.0) - default.gaussian (PennyLane-0.34.0) - default.mixed (PennyLane-0.34.0) - default.qubit (PennyLane-0.34.0) - default.qubit.autograd (PennyLane-0.34.0) - default.qubit.jax (PennyLane-0.34.0) - default.qubit.legacy (PennyLane-0.34.0) - default.qubit.tf (PennyLane-0.34.0) - default.qubit.torch (PennyLane-0.34.0) - default.qutrit (PennyLane-0.34.0) - null.qubit (PennyLane-0.34.0) ``` ### Existing GitHub issues - [X] I have searched existing GitHub issues to make sure the issue does not already exist.

Astral_Cai · February 5, 2024, 5:11pm

This PR should fix it:

github.com/PennyLaneAI/pennylane

BugFix: Tensorflow type promotion error with Tensors

PennyLaneAI:master ← PennyLaneAI:tf-type-promo

opened 04:04PM - 05 Feb 24 UTC

astralcai

+19 -9

### Before submitting Please complete the following checklist when submitting… a PR: - [ ] All new features must include a unit test. If you've fixed a bug or added code that should be tested, add a test to the test directory! - [ ] All new functions and code must be clearly commented and documented. If you do make documentation changes, make sure that the docs build and render correctly by running `make docs`. - [ ] Ensure that the test suite passes, by running `make test`. - [ ] Add a new entry to the `doc/releases/changelog-dev.md` file, summarizing the change, and including a link back to the PR. - [ ] The PennyLane source code conforms to [PEP8 standards](https://www.python.org/dev/peps/pep-0008/). We check all of our code against [Pylint](https://www.pylint.org/). To lint modified files, simply `pip install pylint`, and then run `pylint pennylane/path/to/file.py`. When all the above are checked, delete everything above the dashed line and fill in the pull request template. ------------------------------------------------------------------------------------------------------------ **Context:** **Description of the Change:** **Benefits:** **Possible Drawbacks:** **Related GitHub Issues:**

Astral_Cai · February 13, 2024, 2:23pm

This PR should fix it:

github.com/PennyLaneAI/pennylane

Add broadcasting support to mitigate_with_zne

PennyLaneAI:master ← PennyLaneAI:err-mit-bc

opened 02:14PM - 13 Feb 24 UTC

astralcai

+27 -1

**Context:** A QNode transformed with `mitigate_with_zne` does not accept param…eters with a leading dimension **Description of the Change:** 1. Replace `qml.math.dot` with `qml.math.tensordot` in `_polyfit` 2. Test case added for parameter broadcasting with `mitigate_with_zne` **Benefits:** BugFix **Related GitHub Issues:** https://github.com/PennyLaneAI/pennylane/issues/5100 **Related Shortcut Story:** [sc-55298]

Topic		Replies	Views
Error creating a keras model with a quantum Circuit as a layer PennyLane Help	7	1473	November 29, 2023
Error using quantum nodes into Keras Layers using modulation dataset PennyLane Help	4	419	August 5, 2023
Getting TypeError when trying to use KerasLayer to wrap a quantum node PennyLane Help	6	1970	October 28, 2021
Issues with simple QNN KerasLayer PennyLane Help	1	307	December 11, 2023
KerasLayer error parameter dynamic: True PennyLane Help	5	832	March 14, 2024

Error with KerasLayer and Amplitude Embedding

Related Topics