[BUG] Error in ECR gates processing in `qml.from_qiskit` method

[Tomev]

Describe the bug

When creating a pennylane circuit from a qiskit circuit using qml.from_qiskit method there's an error in the operation name processing in the converter.py file. The error parses ECR Gate operation name to Gate and states that Gate instruction is not supported by PennyLane. That's because Gate is not present in the QISKIT_OPERATION_MAP whereas ECR clearly is.

The ECR gates ends up ignored in the processed circuit.

To Reproduce
Steps to reproduce the behavior:

1. Prepare a qiskit circuit with at least one ECR gate.
2. Run qml.from_qiskit method on the circuit.
3. See error.

Expected behavior
ECR gate is processed correctly and is included in the final circuit.

Error
<...>/pennylane_qiskit/converter.py:582: UserWarning: pennylane_qiskit.converter: The Gate instruction is not supported by PennyLane, and has not been added to the template.

Hotfix
A quick hotfix for this issue is to include:

if instruction.name == "ecr":
    instruction_name = "ECRGate"

after instruction_name = getattr(...) in line 502. However, I believe that this issue can be solved in a more general and elegant way.

Desktop (please complete the following information):

• OS: Tested on Windows 10, Kali Linux and MacOS.
• Version: pennylane-qiskit 0.40.0

Additional context
It's important, because all the publicly available IBM devices uses ECR Gate as their native two-qubit gate.

I believe I can take care of this issue, but I need some time to wrap up my other, work-related stuff first.

[CatalinaAlbornoz]

Hi @Tomev,
Thank you for reporting this.

Do you have an example code that we can use to replicate the issue?

Please make sure to delete your IBM token before sharing the code.

[Tomev]

Hello @CatalinaAlbornoz,

you can run the the code below.

https://github.com/Tomev/qnn_change_detection/blob/main/Examples/19.Pennylane_QiskitCompilation.py

Cheers,
T

[CatalinaAlbornoz]

Hi @Tomev ,

Thanks for sharing your repo! Can you please confirm what version of Qiskit and Qiskit-ibm-runtime you're using?
Our team is having some trouble to replicate your issue.
Could you please post the output of qml.about() too just in case?

[Tomev]

Hello @CatalinaAlbornoz,

yes, absolutely. I'm using:

qiskit                    1.3.2                    pypi_0    pypi
qiskit-ibm-runtime        0.35.0                   pypi_0    pypi

(obrained with conda list) and the output of qml.about() is:

Name: PennyLane
Version: 0.40.0
Summary: PennyLane is a cross-platform Python library for quantum computing, quantum machine learning, and quantum chemistry. Train a quantum computer the same way as a neural network.
Home-page: https://github.com/PennyLaneAI/pennylane
Author: 
Author-email: 
License: Apache License 2.0
Location: /opt/anaconda3/envs/sipqnn/lib/python3.11/site-packages
Requires: appdirs, autograd, autoray, cachetools, diastatic-malt, networkx, numpy, packaging, pennylane-lightning, requests, rustworkx, scipy, tomlkit, typing-extensions
Required-by: PennyLane-qiskit, PennyLane_Lightning

Platform info:           macOS-15.3-arm64-arm-64bit
Python version:          3.11.0
Numpy version:           1.26.4
Scipy version:           1.15.1
Installed devices:
- default.clifford (PennyLane-0.40.0)
- default.gaussian (PennyLane-0.40.0)
- default.mixed (PennyLane-0.40.0)
- default.qubit (PennyLane-0.40.0)
- default.qutrit (PennyLane-0.40.0)
- default.qutrit.mixed (PennyLane-0.40.0)
- default.tensor (PennyLane-0.40.0)
- null.qubit (PennyLane-0.40.0)
- reference.qubit (PennyLane-0.40.0)
- qiskit.aer (PennyLane-qiskit-0.40.0)
- qiskit.basicaer (PennyLane-qiskit-0.40.0)
- qiskit.basicsim (PennyLane-qiskit-0.40.0)
- qiskit.remote (PennyLane-qiskit-0.40.0)
- lightning.qubit (PennyLane_Lightning-0.40.0)

[CatalinaAlbornoz]

Hi @Tomev,

It looks like the issue is being caused by the Qiskit and Qiskit-IBM-Runtime versions you're using. At the moment we only support:

qiskit<1.3
qiskit-ibm-runtime<=0.29

My colleague Andrija was able to run your code with no issues using the following versions:

qiskit                1.2.4
qiskit-aer            0.16.0
qiskit-ibm-provider   0.11.0
qiskit-ibm-runtime    0.29.0

Are you able to downgrade your Qiskit and runtime versions?

[Tomev]

Hello,

Thank you for your patience and support! I'll investigate this issue tomorrow and let you know!

Cheers
T

[Tomev]

Hello,

I created whole new conda environment. First I installed the packages and versions you've specified, and then everything else I needed to run the script (without specifying a version). The problem persists. I get the following warning

/opt/anaconda3/envs/pennylane_qiskit_gh_issue/lib/python3.11/site-packages/pennylane_qiskit/converter.py:582: UserWarning: pennylane_qiskit.converter: The Gate instruction is not supported by PennyLane, and has not been added to the template.

and the ECR gates are not present in the transpiled script. Do note that it's a warning that I get, not any kind of error. The script reaches the end, but with the issue.

Below I'm attaching complete result of the conda list command. Notice that the versions of the libraries you've mentioned are the same as you specified.

# Name                    Version                   Build  Channel
annotated-types           0.7.0                    pypi_0    pypi
antlr4-python3-runtime    4.13.2                   pypi_0    pypi
appdirs                   1.4.4                    pypi_0    pypi
astunparse                1.6.3                    pypi_0    pypi
autograd                  1.7.0                    pypi_0    pypi
autoray                   0.7.0                    pypi_0    pypi
bzip2                     1.0.8                h80987f9_6
ca-certificates           2025.2.25            hca03da5_0
cachetools                5.5.2                    pypi_0    pypi
certifi                   2025.1.31                pypi_0    pypi
cffi                      1.17.1                   pypi_0    pypi
charset-normalizer        3.4.1                    pypi_0    pypi
cryptography              44.0.2                   pypi_0    pypi
diastatic-malt            2.15.2                   pypi_0    pypi
dill                      0.3.9                    pypi_0    pypi
filelock                  3.17.0                   pypi_0    pypi
fsspec                    2025.2.0                 pypi_0    pypi
gast                      0.6.0                    pypi_0    pypi
ibm-cloud-sdk-core        3.22.1                   pypi_0    pypi
ibm-platform-services     0.59.1                   pypi_0    pypi
idna                      3.10                     pypi_0    pypi
jinja2                    3.1.5                    pypi_0    pypi
joblib                    1.4.2                    pypi_0    pypi
libffi                    3.4.4                hca03da5_1
markupsafe                3.0.2                    pypi_0    pypi
mpmath                    1.3.0                    pypi_0    pypi
ncurses                   6.4                  h313beb8_0
networkx                  3.4.2                    pypi_0    pypi
numpy                     2.0.2                    pypi_0    pypi
openqasm3                 1.0.1                    pypi_0    pypi
openssl                   1.1.1w               h1a28f6b_0
packaging                 24.2                     pypi_0    pypi
pbr                       6.1.1                    pypi_0    pypi
pennylane                 0.40.0                   pypi_0    pypi
pennylane-lightning       0.40.0                   pypi_0    pypi
pennylane-qiskit          0.40.0                   pypi_0    pypi
pip                       25.0            py311hca03da5_0
psutil                    7.0.0                    pypi_0    pypi
pycparser                 2.22                     pypi_0    pypi
pydantic                  2.10.6                   pypi_0    pypi
pydantic-core             2.27.2                   pypi_0    pypi
pyjwt                     2.10.1                   pypi_0    pypi
pyspnego                  0.11.2                   pypi_0    pypi
python                    3.11.0               hc0d8a6c_3
python-dateutil           2.9.0.post0              pypi_0    pypi
qiskit                    1.2.4                    pypi_0    pypi
qiskit-aer                0.16.0                   pypi_0    pypi
qiskit-ibm-provider       0.11.0                   pypi_0    pypi
qiskit-ibm-runtime        0.29.0                   pypi_0    pypi
qiskit-machine-learning   0.8.2                    pypi_0    pypi
qiskit-qasm3-import       0.5.1                    pypi_0    pypi
readline                  8.2                  h1a28f6b_0
requests                  2.32.3                   pypi_0    pypi
requests-ntlm             1.3.0                    pypi_0    pypi
rustworkx                 0.16.0                   pypi_0    pypi
scikit-learn              1.6.1                    pypi_0    pypi
scipy                     1.15.2                   pypi_0    pypi
scipy-openblas32          0.3.29.0.0               pypi_0    pypi
setuptools                75.8.0          py311hca03da5_0
six                       1.17.0                   pypi_0    pypi
sqlite                    3.45.3               h80987f9_0
stevedore                 5.4.1                    pypi_0    pypi
symengine                 0.13.0                   pypi_0    pypi
sympy                     1.12.1                   pypi_0    pypi
termcolor                 2.5.0                    pypi_0    pypi
threadpoolctl             3.5.0                    pypi_0    pypi
tk                        8.6.14               h6ba3021_0
tomlkit                   0.13.2                   pypi_0    pypi
torch                     2.6.0                    pypi_0    pypi
typing-extensions         4.12.2                   pypi_0    pypi
tzdata                    2025a                h04d1e81_0
urllib3                   2.3.0                    pypi_0    pypi
websocket-client          1.8.0                    pypi_0    pypi
websockets                15.0                     pypi_0    pypi
wheel                     0.45.1          py311hca03da5_0
xz                        5.6.4                h80987f9_1
zlib                      1.2.13               h18a0788_1

Can you confirm that you were able to run the script without issue? I'm specifically interested if the transpiled scripts you get have ECR gates.

Cheers
T

[CatalinaAlbornoz]

Hi @Tomev, I'm able to replicate your issue.

Let me check with the team to see what we can do about it.

[CatalinaAlbornoz]

Hi @Tomev ,

I think the issue is related to the transpilation and parsing.

Here are two Minimal Reproducible Examples where I use the ECR gate from the qiskit.circuit.library.

In the first example I do just one pass of transpilation with the pass manager, and PennyLane is able to capture the ECRGate. In the second example I add the conversion to QASM and parsing (which is what you had in your code) and in this case PennyLane fails to add the ECR gate.
At the end I show this with a slightly modified version of your code. This shows that if you can avoid the conversion to QASM and parsing you may be able to solve your issue.

Example that works

# MWE simple
import pennylane as qml
from qiskit import QuantumCircuit
from qiskit.circuit.library import ECRGate
from qiskit_ibm_runtime.fake_provider import FakeBrisbane
from qiskit.transpiler.preset_passmanagers import generate_preset_pass_manager

# Create the Qiskit circuit
gate = ECRGate()
circuit = QuantumCircuit(2)
circuit.append(gate, [0, 1])
print("Qiskit circuit \n")
print(circuit.draw())

ansatz = circuit

# Ansatz transpilation
simulator = FakeBrisbane()
pm = generate_preset_pass_manager(optimization_level=2, backend=simulator)
transpiled_ansatz = pm.run(ansatz)
print("\n transpiled_ansatz \n")
print(transpiled_ansatz)

# Pennylane
dev = qml.device("default.qubit")

@qml.qnode(dev)
def pennylane_circuit():
    qml.from_qiskit(transpiled_ansatz)()
    return qml.expval(qml.Z(0))

print("\n pennylane_circuit \n")
qml.draw_mpl(pennylane_circuit)();

Example that doesn’t work (because of the conversion to QASM and parsing)

# MWE with QASM conversion and parsing
import pennylane as qml
from qiskit import QuantumCircuit
from qiskit.circuit.library import ECRGate
from qiskit.qasm3 import dumps
from qiskit_ibm_runtime.fake_provider import FakeBrisbane
from qiskit_qasm3_import import parse
from qiskit.transpiler.preset_passmanagers import generate_preset_pass_manager

# Create the Qiskit circuit
gate = ECRGate()
circuit = QuantumCircuit(2)
circuit.append(gate, [0, 1])
print("Qiskit circuit \n")
print(circuit.draw())

ansatz = circuit

# Ansatz transpilation
simulator = FakeBrisbane()
pm = generate_preset_pass_manager(optimization_level=2, backend=simulator)
transpiled_ansatz = pm.run(ansatz)
transpiled_ansatz_qasm = dumps(transpiled_ansatz)
transpiled_ansatz_parsed = parse(transpiled_ansatz_qasm)
print("\n transpiled_ansatz_parsed \n")
print(transpiled_ansatz_parsed)

# Pennylane transpilaiton
# There's bug when compiling ECR gates via pennylane_qiskit.converter.
dev = qml.device("default.qubit")

@qml.qnode(dev)
def pennylane_circuit():
    qml.from_qiskit(transpiled_ansatz_parsed)(wires=range(21))
    return qml.expval(qml.Z(0))

print("\n pennylane_circuit \n")
qml.draw_mpl(pennylane_circuit)();

Below is a simplified version of your code that shows this behaviour too. If you look at the print for the pennylane_circuit you'll see the ECR gates.

# MWE v0-1
def layout_to_map(layout, quantum_register) -> Dict[int, int]:
    # Get simple qubit map from circuit layout.
    # The map is a dict in form {virtual -> physical}.
    # We want to act on the physical qubits.
    initial_layout = layout.initial_layout.get_virtual_bits()
    # print(test)

    # print(initial_layout)

    qubit_map = {}

    for k in quantum_register:
        qubit_map[k._index] = initial_layout[k]

    return qubit_map

# https://qiskit-community.github.io/qiskit-machine-learning/tutorials/01_neural_networks.html
# Example implementation.

# Config
n_qubits = 4
n_wires = 2 * n_qubits

# Ansatz preparation

# We want to start with ansatz, because it's more complicated than our selected
# feature map (we use simple AngleEmbedding, which is a layer of single qubit
# rotations). This way, we can transpile it, get the final layout of our circuit
# and prepare the feature map and observable according to the layout obtained by
# ansatz transpilation.

ansatz = PauliTwoDesign(n_wires, reps=1)

# Ansatz transpilation

# backend = provider.get_backend('ibm_brisbane')
simulator = FakeBrisbane()

pm = generate_preset_pass_manager(optimization_level=2, backend=simulator)
transpiled_ansatz = pm.run(ansatz)

qubit_map = layout_to_map(transpiled_ansatz.layout, ansatz.qubits)
#print(f"Permutation map: {qubit_map}")

#transpiled_ansatz_qasm = dumps(transpiled_ansatz)
#transpiled_ansatz = parse(transpiled_ansatz_qasm)

print("transpiled_ansatz \n",transpiled_ansatz)

# Feature map creation
#print(f"Number of qubit in transpiled circuits: {transpiled_ansatz.num_qubits}")
feature_map = QuantumCircuit(transpiled_ansatz.num_qubits)

data = ParameterVector("inputs", n_wires)

for i, k in enumerate(qubit_map.keys()):
    feature_map.ry(data[i], qubit_map[k])

# Transpilation is needed, to ensure proper gates are used.
# Ensure optimization level is low, so that the qubits are not permuted.
pm = generate_preset_pass_manager(optimization_level=0, backend=simulator)
transpiled_feature_map = pm.run(feature_map)
#transpiled_feature_map_qasm = dumps(transpiled_feature_map)
#transpiled_feature_map = parse(transpiled_feature_map_qasm)

print("transpiled_feature_map \n",transpiled_feature_map)

# Pennylane transpilaiton
# There's bug when compiling ECR gates via pennylane_qiskit.converter.
# TODO: Analyze and submit it later. It can be easily fixed.
dev = qml.device("default.qubit")  # , wires=qubit_map.values())

@qml.qnode(dev)
def pennylane_circuit(inputs, weights):
    qml.from_qiskit(transpiled_feature_map)(inputs)
    qml.from_qiskit(transpiled_ansatz)(weights)
    return qml.expval(qml.Z(qubit_map[1]))

data = [i for i in range(8)]
weights = [i for i in range(16)]

print(pennylane_circuit(data, weights))

drawer = qml.draw(pennylane_circuit)
print("qml.draw(pennylane_circuit) \n",drawer(data, weights))
drawer = qml.draw(qml.from_qiskit(transpiled_feature_map))
print("qml.draw(qml.from_qiskit(transpiled_feature_map) \n",drawer(data))


# How do I know the order of the weights and inputs?
# print(qnn_circuit)

[Tomev]

Hi @CatalinaAlbornoz!

Thank you so much for exhaustive explanation. As I stated in my initial post, I already handled this issue (although differently) so you didn't really have to be so specific. I really appreciate it though!

However, as I understand, the issue still persists if I run my original code (with the QASM transpilation and what-nots). Moreover, I guess that's also not the intended behavior, correct? If that's not the case, feel free to close this issue.

Actually, I can look into it myself, especially since you've pinpointed the problem to be QASM conversion-related. Would that be acceptable for you and the team?

Cheers
T

[CatalinaAlbornoz]

Hi @Tomev,

The issue does still persist with the original code.

I'd love to understand the use case better so that we can build the right solution for it.

Can you please help me understand why we need the conversion to QASM and parsing?

Feel free to look into it yourself too if you prefer. The team would surely appreciate that.