search

qiskit-community / qiskit-aqua

Quantum Algorithms & Applications (**DEPRECATED** since April 2021 - see readme for more info)
https://qiskit.org/aqua
Apache License 2.0
572 stars 377 forks source link

QSVM algorithm in IBMQ experience, real backend doesn't work #550

Closed hykavitha closed 5 years ago

hykavitha commented 5 years ago

Informations

What is the current behavior?

Steps to reproduce the problem

What is the expected behavior?

Suggested solutions

hykavitha commented 5 years ago

Qiskit Aqua version: pip list | grep "qiskit" qiskit 0.10.1
qiskit-aer 0.2.0
qiskit-aqua 0.5.0
qiskit-chemistry 0.5.0
qiskit-ibmq-provider 0.2.2
qiskit-ignis 0.1.1
qiskit-terra 0.8.0
Python version: 3.73 Operating system: OSX

What is the current behavior? I'm trying to run QSVM algorithm in IBMQ experience, want to run in one of those real quantum computers.

from qiskit import IBMQ IBMQ.load_accounts() print("Available backends:") IBMQ.backends()

[<IBMQSimulator('ibmq_qasm_simulator') from IBMQ(ibm-q, open, main)>, <IBMQBackend('ibmqx4') from IBMQ(ibm-q, open, main)>, <IBMQBackend('ibmqx2') from IBMQ(ibm-q, open, main)>, <IBMQBackend('ibmq_16_melbourne') from IBMQ(ibm-q, open, main)>, <IBMQSimulator('ibmq_qasm_simulator') from IBMQ()>]

backend = BasicAer.get_backend('ibmq_qasm_simulator') With the BasicAer I realize I'm using IBMQ's simulator and not the actual IBMQ device. if I chose 'ibmq_qasm_simulator', it works!!!

device = IBMQ.get_backend('ibmq_16_melbourne') # ibmx4

Choose a real - device

quantum_instance = QuantumInstance(device, shots=100, seed=seed, seed_transpiler=seed) result = qsvm.run(quantum_instance) The above code which is supposed to use real Quantum computer "ibmq_16_melbourne" throws me a

KeyError Traceback (most recent call last)

in 1 ----> 2 result = qsvm.run(quantum_instance) 3 print("testing success ratio: ", result['testing_accuracy']) 4 end = timer() 5 print("timedelta: " , timedelta(seconds=end-start)) /opt/conda/lib/python3.7/site- packages/qiskit/aqua/algorithms/quantum_algorithm.py in run(self, quantum_instance, **kwargs) 62 quantum_instance.set_config(**kwargs) 63 self._quantum_instance = quantum_instance ---> 64 return self._run() 65 66 @abstractmethod /opt/conda/lib/python3.7/site- packages/qiskit/aqua/algorithms/many_sample/qsvm/qsvm.py in _run(self) 306 307 def _run(self): --> 308 return self.instance.run() 309 310 @property /opt/conda/lib/python3.7/site- packages/qiskit/aqua/algorithms/many_sample/qsvm/_qsvm_binary.py in run(self) 127 def run(self): 128 """Put the train, test, predict together.""" --> 129 self.train(self._qalgo.training_dataset[0], self._qalgo.training_dataset[1]) 130 if self._qalgo.test_dataset is not None: 131 self.test(self._qalgo.test_dataset[0], self._qalgo.test_dataset[1]) /opt/conda/lib/python3.7/site- packages/qiskit/aqua/algorithms/many_sample/qsvm/_qsvm_binary.py in train(self, data, labels) 71 """ 72 scaling = 1.0 if self._qalgo.quantum_instance.is_statevector else None ---> 73 kernel_matrix = self._qalgo.construct_kernel_matrix(data) 74 labels = labels * 2 - 1 # map label from 0 --> -1 and 1 --> 1 75 labels = labels.astype(np.float) /opt/conda/lib/python3.7/site- packages/qiskit/aqua/algorithms/many_sample/qsvm/qsvm.py in construct_kernel_matrix(self, x1_vec, x2_vec, quantum_instance) 243 num_processes=aqua_globals.num_processes) 244 --> 245 results = self.quantum_instance.execute(circuits) 246 247 if logger.isEnabledFor(logging.DEBUG): /opt/conda/lib/python3.7/site-packages/qiskit/aqua/quantum_instance.py in execute(self, circuits, **kwargs) 192 qobjs = compile_circuits(circuits, self._backend, self._backend_config, self._compile_config, self._run_config, 193 show_circuit_summary=self._circuit_summary, circuit_cache=self._circuit_cache, --> 194 **kwargs) 195 196 if self._measurement_error_mitigation_cls is not None: /opt/conda/lib/python3.7/site-packages/qiskit/aqua/utils/run_circuits.py in compile_circuits(circuits, backend, backend_config, compile_config, run_config, show_circuit_summary, circuit_cache, **kwargs) 195 transpiled_sub_circuits = compiler.transpile(sub_circuits, backend, **backend_config, 196 **compile_config) --> 197 qobj = circuit_cache.load_qobj_from_cache(transpiled_sub_circuits, i, run_config=run_config) 198 else: 199 qobj = circuit_cache.load_qobj_from_cache(sub_circuits, i, run_config=run_config) /opt/conda/lib/python3.7/site-packages/qiskit/aqua/utils/circuit_cache.py in load_qobj_from_cache(self, circuits, chunk, run_config) 178 if not hasattr(compiled_gate, 'params') or len(compiled_gate.params) < 1: continue 179 if compiled_gate.name == 'snapshot': continue --> 180 cache_index = self.mappings[chunk][circ_num][gate_num] 181 (uncompiled_gate, regs, _) = raw_gates[cache_index] 182 KeyError: 0 Steps to reproduce the problem Above code What is the expected behavior? I expect to connect to the IBM-Q backend, not a simulator.
Matt-Stypulkoski commented 5 years ago

Just to add a little more information, the user stated that they were receiving this error when they were running this tutorial: https://github.com/Qiskit/qiskit-tutorials/tree/master/qiskit/artificial_intelligence, I am assuming qsvm_classification.ipynb. (User stated this in the original stack exchange question https://quantumcomputing.stackexchange.com/questions/6295/how-to-run-quantum-svm-algorithm-from-qiskit-in-real-ibm-quantum-computer-using)

I tried running through this tutorial on ibmq_16_melbourne, and received the same error. This is what the logger showed before throwing the error:

2019-06-04 11:51:16,594:qiskit.aqua.algorithms.many_sample.qsvm.qsvm:DEBUG: Building circuits:
|██████████████████████████████████████████████████| 780/780 [00:00:00:00]
2019-06-04 11:51:18,869:qiskit.aqua.utils.run_circuits:INFO: Circuit cache is empty, compiling from scratch.
chunfuchen commented 5 years ago

Do you mind elaborating more details on the size of dataset or do you mind sharing the full scripts?

I just tried the notebook with ibmq_16_melbourne backend, and it does not crash. To test that, I reduce the size of the dataset to avoid submitting too many circuits. (4 training samples, 2 testing samples, and 2 unknown label samples)

https://github.com/Qiskit/qiskit-tutorials/blob/master/qiskit/artificial_intelligence/qsvm_classification.ipynb

Qiskit version: 

qiskit                0.10.3
qiskit-aer            0.2.1
qiskit-aqua           0.5.1
qiskit-chemistry      0.5.0
qiskit-ibmq-provider  0.2.2
qiskit-ignis          0.1.1
qiskit-terra          0.8.1

Here is the log I had, it is still running but at least one of jobs is completed

019-06-04 13:30:01,138:qiskit.aqua.components.feature_maps.pauli_expansion:INFO: Pauli terms include: ['IZ', 'ZI', 'ZZ']
2019-06-04 13:30:01,142:qiskit.aqua.quantum_instance:INFO: backend_options can not used with the backends in IBMQ provider.
2019-06-04 13:30:01,143:qiskit.aqua.quantum_instance:INFO: 
Qiskit Terra version: 0.8.1
Backend: 'ibmq_16_melbourne (<qiskit.providers.ibmq.ibmqprovider.IBMQProvider object at 0x127563400>)', with following setting:
{'basis_gates': ['u1', 'u2', 'u3', 'cx', 'id'], 'coupling_map': [[1, 0], [1, 2], [2, 3], [4, 3], [4, 10], [5, 4], [5, 6], [5, 9], [6, 8], [7, 8], [9, 8], [9, 10], [11, 3], [11, 10], [11, 12], [12, 2], [13, 1], [13, 12]]}
{'pass_manager': None, 'initial_layout': None, 'seed_transpiler': 10598}
RunConfig(max_credits=10, seed=10598, shots=1024)
{'timeout': None, 'wait': 5}
{}
{}
Measurement mitigation: None
2019-06-04 13:30:01,145:qiskit.aqua.algorithms.many_sample.qsvm.qsvm:DEBUG: Building circuits:
|██████████████████████████████████████████████████| 6/6 [00:00:00:00]
2019-06-04 13:30:01,289:qiskit.aqua.utils.run_circuits:INFO: Circuit cache is empty, compiling from scratch.
2019-06-04 13:30:09,195:qiskit.aqua.utils.run_circuits:INFO: Backend status: BackendStatus(backend_name='ibmq_16_melbourne', backend_version='1.0.0', operational=True, pending_jobs=119, status_msg='active')
2019-06-04 13:30:09,196:qiskit.aqua.utils.run_circuits:INFO: There are 1 jobs are submitted.
2019-06-04 13:30:09,196:qiskit.aqua.utils.run_circuits:INFO: All job ids:
['5cf6aa9d464bd40076d8f4e0']
2019-06-04 13:30:09,197:qiskit.aqua.utils.run_circuits:INFO: Running 0-th qobj, job id: 5cf6aa9d464bd40076d8f4e0
Got a 502 code response to /api/Jobs/5cf6aa9d464bd40076d8f4e0/status: 502 Bad Gateway: Registered endpoint failed to handle the request.

2019-06-04 16:53:44,053:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 0-th qobj, job id: 5cf6aa9d464bd40076d8f4e0
2019-06-04 16:53:44,067:qiskit.aqua.algorithms.many_sample.qsvm.qsvm:DEBUG: Calculating overlap:
|██████████████████████████████████████████████████| 6/6 [00:00:00:00]
2019-06-04 16:53:44,251:qiskit.aqua.utils.qp_solver:DEBUG: Solving QP problem is completed.
2019-06-04 16:53:44,252:qiskit.aqua.algorithms.many_sample.qsvm.qsvm:DEBUG: Building circuits:
|██████████████████████████████████████████████████| 8/8 [00:00:00:00]
2019-06-04 16:53:46,479:qiskit.aqua.utils.run_circuits:INFO: Circuit cache miss, recompiling. Cache miss reason: AquaError('Gate mismatch at gate 11 (u3, 3 params) of circuit against gate 11 (u2, 2 params) of cached qobj')
2019-06-04 16:53:52,610:qiskit.aqua.utils.run_circuits:INFO: Backend status: BackendStatus(backend_name='ibmq_16_melbourne', backend_version='1.0.0', operational=True, pending_jobs=80, status_msg='active')
2019-06-04 16:53:52,612:qiskit.aqua.utils.run_circuits:INFO: There are 1 jobs are submitted.
2019-06-04 16:53:52,613:qiskit.aqua.utils.run_circuits:INFO: All job ids:
['5cf6da5ea2f98a00755614a8']
Matt-Stypulkoski commented 5 years ago

I am unsure about the original user, however when I received the error I was using the defaults for everything. Whatever was already set in the notebook was what I was using.

However, when I made the changes you had made (changing the number of training samples and testing samples) the job was sent.

nonhermitian commented 5 years ago

So it seems to work if run on a 20q machine. Fails on the 14 and 5 qubit devices. I had to dig into the internals to check, but it seems that the circuits use only two qubits, so it should work on all the devices. The fact that it gives an key error means that the chunks are likely not being computed correctly for the public devices. The public devices have different circuit limits than the larger devices, so perhaps that is not being taken into account correctly. So this is indeed a bug.

hykavitha commented 5 years ago

how do we fix this? And how do we know how many qubits are we using? Is this an error from our code or from qiskit (then this needsa bug filing)

nonhermitian commented 5 years ago

It needs to be fixed in aqua. I dug into the internals to find the number of qubits. But the aqua developers can probably point out an easier way.

chunfuchen commented 5 years ago

@nonhermitian and @hykavitha

Regarding the number of circuits in a job, Aqua is based on the backend property max_experiments to split the number of circuits into jobs. E.g. if you submit 100 circuits, but the max_experiments is 50, Aqua will automatically split the circuits into 2 jobs each with 50 circuits and then get back a single result for you.

Regarding the required number of qubits, you can check out the qiskit-tutorial to learn how the QSVM encodes the data into a quantum state. E.g., the default QSVM requires the same number of qubits to the dimension of data. The data we used in the example is two-dimension; thus we need two qubits. https://github.com/Qiskit/qiskit-tutorials/blob/master/qiskit/artificial_intelligence/qsvm_classification.ipynb

Furthermore, if you have a circuit whose qubits is larger than the targeted device, I believe that qiskit-terra will raise an error. 

TranspilerError: 'Number of qubits greater than device.'

Regarding the key error for circuit caching in the first question, I can reproduce the error if I do not reduce the problem size. Will check that later in the stable branch. On the other hand, the error you encounter should already be fixed in the master branch.

In the meanwhile, you can disable the circuit cache and do qobj validation in the quantum instance to avoid the error, you should be able to run on the real device then.

quantum_instance = QuantumInstance(backend, shots=1024, seed=seed, 
seed_transpiler=seed, circuit_caching=False, skip_qobj_validation=False)

I just ran the codes with the above modification (with the original amount of data) on ibmq_16_melbourne. here is my log, It seems that there is some issues on the api to the backend. will update the post if I can successfully run on public devices. (I am able to run on the 20q device without an issue,)

updated: the jobs are completed on ibmq_16_melbourne

2019-06-11 01:53:18,973:qiskit.aqua.algorithms.many_sample.qsvm.qsvm:DEBUG: Building circuits:
|██████████████████████████████████████████████████| 780/780 [00:00:00:00]
Got a 400 code response to /api/Jobs: {"error":{"status":400,"message":"Generic error","code":"GENERIC_ERROR"}}
Got a 400 code response to /api/Jobs: {"error":{"status":400,"message":"Generic error","code":"GENERIC_ERROR"}}
Got a 400 code response to /api/Jobs: {"error":{"status":400,"message":"Generic error","code":"GENERIC_ERROR"}}
2019-06-11 04:49:35,089:qiskit.aqua.utils.run_circuits:INFO: Backend status: BackendStatus(backend_name='ibmq_16_melbourne', backend_version='1.0.0', operational=True, pending_jobs=21, status_msg='active')
2019-06-11 04:49:35,090:qiskit.aqua.utils.run_circuits:INFO: There are 11 jobs are submitted.
2019-06-11 04:49:35,091:qiskit.aqua.utils.run_circuits:INFO: All job ids:
['5cff424e43e51900734e221c', '5cff42b9711ca80077371510', '5cff43239bb71c007888317b', '5cff4391bd9211006b03f1e9', '5cff4394752445007027ceb9', '5cff43f75ee341007ad7aec4', '5cff43fe68bfba0075bad960', '5cff445b752445007027cebb', '5cff44be06c6bc0073aab339', '5cff4ddd68bfba0075bad974', '5cff6b1d42519f006f7a6f1d']
2019-06-11 04:49:35,092:qiskit.aqua.utils.run_circuits:INFO: Running 0-th qobj, job id: 5cff424e43e51900734e221c
2019-06-11 04:49:35,987:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 0-th qobj, job id: 5cff424e43e51900734e221c
2019-06-11 04:49:35,988:qiskit.aqua.utils.run_circuits:INFO: Running 1-th qobj, job id: 5cff42b9711ca80077371510
2019-06-11 05:10:22,821:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 1-th qobj, job id: 5cff42b9711ca80077371510
2019-06-11 05:10:22,822:qiskit.aqua.utils.run_circuits:INFO: Running 2-th qobj, job id: 5cff43239bb71c007888317b
2019-06-11 05:10:26,684:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 2-th qobj, job id: 5cff43239bb71c007888317b
2019-06-11 05:10:26,685:qiskit.aqua.utils.run_circuits:INFO: Running 3-th qobj, job id: 5cff4391bd9211006b03f1e9
2019-06-11 05:10:27,681:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 3-th qobj, job id: 5cff4391bd9211006b03f1e9
2019-06-11 05:10:27,681:qiskit.aqua.utils.run_circuits:INFO: Running 4-th qobj, job id: 5cff4394752445007027ceb9
2019-06-11 05:10:28,573:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 4-th qobj, job id: 5cff4394752445007027ceb9
2019-06-11 05:10:28,573:qiskit.aqua.utils.run_circuits:INFO: Running 5-th qobj, job id: 5cff43f75ee341007ad7aec4
2019-06-11 05:10:29,523:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 5-th qobj, job id: 5cff43f75ee341007ad7aec4
2019-06-11 05:10:29,523:qiskit.aqua.utils.run_circuits:INFO: Running 6-th qobj, job id: 5cff43fe68bfba0075bad960
2019-06-11 05:10:30,805:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 6-th qobj, job id: 5cff43fe68bfba0075bad960
2019-06-11 05:10:30,807:qiskit.aqua.utils.run_circuits:INFO: Running 7-th qobj, job id: 5cff445b752445007027cebb
2019-06-11 05:10:32,531:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 7-th qobj, job id: 5cff445b752445007027cebb
2019-06-11 05:10:32,535:qiskit.aqua.utils.run_circuits:INFO: Running 8-th qobj, job id: 5cff44be06c6bc0073aab339
2019-06-11 05:10:33,755:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 8-th qobj, job id: 5cff44be06c6bc0073aab339
2019-06-11 05:10:33,758:qiskit.aqua.utils.run_circuits:INFO: Running 9-th qobj, job id: 5cff4ddd68bfba0075bad974
2019-06-11 05:31:17,780:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 9-th qobj, job id: 5cff4ddd68bfba0075bad974
2019-06-11 05:31:17,781:qiskit.aqua.utils.run_circuits:INFO: Running 10-th qobj, job id: 5cff6b1d42519f006f7a6f1d
2019-06-11 05:31:20,498:qiskit.aqua.utils.run_circuits:INFO: COMPLETED the 10-th qobj, job id: 5cff6b1d42519f006f7a6f1d
nonhermitian commented 5 years ago

Ok great. So it is fixed in master.

If I wanted to programmatically find out the number of qubits used, perhaps by accessing the underlying circuits or the like, how would I go about doing that?

chunfuchen commented 5 years ago

Yes, for any algorithm and components, we usually have a construct_circuit method e.g. in QSVM https://github.com/Qiskit/qiskit-aqua/blob/master/qiskit/aqua/algorithms/many_sample/qsvm/qsvm.py#L174 But each algorithm/component might need different parameters to get the circuits.

You can get the circuit easily after you construct the algorithm object.

E.g., QSVM uses the circuits to calculate the overlap between two data; thus, its construct_circuit method requires two data and an optional flag to add the measurement gate or not.

like in the tutorials


feature_map = SecondOrderExpansion(feature_dimension=feature_dim, depth=2, entanglement='linear')
qsvm = QSVM(feature_map, training_input, test_input, datapoints[0])
qc = qsvm.construct_circuit(x1=datapoints[0][0], x2=datapoints[0][1]) # these are two data points.
qc.draw(output='latex')

and this is the circuit.

image

hykavitha commented 5 years ago

@chunfuchen : thats great to hear its fixed in the master branch. does it mean, this works at IBM-Q experience now? Also does we can access ibm-Q computer from a general public accessing public devices? Also, i have 2 questions,

  1. wanted to know how are you printing while qsvm.run(quantum_instance) is running? I'm running the algorithm for training size 4 & test 2, its been 10 mins still that statement is executing with no sign of output. how did you get those informative print-outs while your qsvm.run() is happening? image

  2. in your circuit output, its using just 2 qubits q0, q1 & using U & C gates. When its using 2 qubits, shoudl it behave like a classical computer with two states (0 or 1) ?

hykavitha commented 5 years ago

@nonhermitian & @chunfuchen : I ran the test as per chunfuchen suggestion to get the qsvm work on tranining size 4 & testing size 2 I still couldnt use real Q. Here is the code & output -error

print("Available backends:") IBMQ.backends() Available backends: [<IBMQSimulator('ibmq_qasm_simulator') from IBMQ(ibm-q, open, main)>, <IBMQBackend('ibmqx4') from IBMQ(ibm-q, open, main)>, <IBMQBackend('ibmqx2') from IBMQ(ibm-q, open, main)>, <IBMQBackend('ibmq_16_melbourne') from IBMQ(ibm-q, open, main)>, <IBMQBackend('ibmqx4') from IBMQ()>, <IBMQBackend('ibmqx2') from IBMQ()>, <IBMQBackend('ibmq_16_melbourne') from IBMQ()>, <IBMQSimulator('ibmq_qasm_simulator') from IBMQ()>]

backend = IBMQ.get_backend('ibmq_16_melbourne', hub=None) print(backend)

backend = IBMQ.get_backend('ibmq_16_melbourne', hub=None) print(backend) ​ ibmq_16_melbourne

quantum_instance = QuantumInstance(backend, shots=1024, seed=seed, seed_transpiler=seed) print(quantum_instance)

Qiskit Terra version: 0.8.1 Backend: 'ibmq_16_melbourne (<qiskit.providers.ibmq.ibmqprovider.IBMQProvider object at 0x7f400abfd3c8>)', with following setting: {'basis_gates': ['u1', 'u2', 'u3', 'cx', 'id'], 'coupling_map': [[1, 0], [1, 2], [2, 3], [4, 3], [4, 10], [5, 4], [5, 6], [5, 9], [6, 8], [7, 8], [9, 8], [9, 10], [11, 3], [11, 10], [11, 12], [12, 2], [13, 1], [13, 12]]} {'pass_manager': None, 'initial_layout': None, 'seed_transpiler': 10598} RunConfig(max_credits=10, seed=10598, shots=1024) {'timeout': None, 'wait': 5} {} {} Measurement mitigation: None

result = qsvm.run(quantum_instance)

result = qsvm.run(quantum_instance) ​

KeyError Traceback (most recent call last)

in ----> 1 result = qsvm.run(quantum_instance) /opt/conda/lib/python3.7/site-packages/qiskit/aqua/algorithms/quantum_algorithm.py in run(self, quantum_instance, **kwargs) 62 quantum_instance.set_config(**kwargs) 63 self._quantum_instance = quantum_instance ---> 64 return self._run() 65 66 @abstractmethod /opt/conda/lib/python3.7/site-packages/qiskit/aqua/algorithms/many_sample/qsvm/qsvm.py in _run(self) 306 307 def _run(self): --> 308 return self.instance.run() 309 310 @property /opt/conda/lib/python3.7/site-packages/qiskit/aqua/algorithms/many_sample/qsvm/_qsvm_binary.py in run(self) 131 self.test(self._qalgo.test_dataset[0], self._qalgo.test_dataset[1]) 132 if self._qalgo.datapoints is not None: --> 133 predicted_labels = self.predict(self._qalgo.datapoints) 134 predicted_classes = map_label_to_class_name(predicted_labels, 135 self._qalgo.label_to_class) /opt/conda/lib/python3.7/site-packages/qiskit/aqua/algorithms/many_sample/qsvm/_qsvm_binary.py in predict(self, data) 120 numpy.ndarray: predicted labels, Nx1 array 121 """ --> 122 predicted_confidence = self.get_predicted_confidence(data) 123 binarized_predictions = (np.sign(predicted_confidence) + 1) / 2 # remap -1 --> 0, 1 --> 1 124 predicted_labels = binarized_predictions.astype(int) /opt/conda/lib/python3.7/site-packages/qiskit/aqua/algorithms/many_sample/qsvm/_qsvm_binary.py in get_predicted_confidence(self, data, return_kernel_matrix) 52 svms = self._ret['svm']['support_vectors'] 53 yin = self._ret['svm']['yin'] ---> 54 kernel_matrix = self._qalgo.construct_kernel_matrix(data, svms) 55 56 confidence = np.sum(yin * alphas * kernel_matrix, axis=1) + bias /opt/conda/lib/python3.7/site-packages/qiskit/aqua/algorithms/many_sample/qsvm/qsvm.py in construct_kernel_matrix(self, x1_vec, x2_vec, quantum_instance) 243 num_processes=aqua_globals.num_processes) 244 --> 245 results = self.quantum_instance.execute(circuits) 246 247 if logger.isEnabledFor(logging.DEBUG): /opt/conda/lib/python3.7/site-packages/qiskit/aqua/quantum_instance.py in execute(self, circuits, **kwargs) 192 qobjs = compile_circuits(circuits, self._backend, self._backend_config, self._compile_config, self._run_config, 193 show_circuit_summary=self._circuit_summary, circuit_cache=self._circuit_cache, --> 194 **kwargs) 195 196 if self._measurement_error_mitigation_cls is not None: /opt/conda/lib/python3.7/site-packages/qiskit/aqua/utils/run_circuits.py in compile_circuits(circuits, backend, backend_config, compile_config, run_config, show_circuit_summary, circuit_cache, **kwargs) 195 transpiled_sub_circuits = compiler.transpile(sub_circuits, backend, **backend_config, 196 **compile_config) --> 197 qobj = circuit_cache.load_qobj_from_cache(transpiled_sub_circuits, i, run_config=run_config) 198 else: 199 qobj = circuit_cache.load_qobj_from_cache(sub_circuits, i, run_config=run_config) /opt/conda/lib/python3.7/site-packages/qiskit/aqua/utils/circuit_cache.py in load_qobj_from_cache(self, circuits, chunk, run_config) 178 if not hasattr(compiled_gate, 'params') or len(compiled_gate.params) < 1: continue 179 if compiled_gate.name == 'snapshot': continue --> 180 cache_index = self.mappings[chunk][circ_num][gate_num] 181 (uncompiled_gate, regs, _) = raw_gates[cache_index] 182 KeyError: 0
chunfuchen commented 5 years ago

@hykavitha, to see the log, you can turn Aqua log as we mentioned in the first cell of the notebook

# setup aqua logging
import logging
from qiskit.aqua import set_qiskit_aqua_logging
set_qiskit_aqua_logging(logging.DEBUG)  # choose INFO, DEBUG to see the log

On the other hand, as I posted above, I am able to run the experiment with the original data size on the public device ibmq_16_melbourne at the stable branch with the following modification:

quantum_instance = QuantumInstance(backend, shots=1024, seed=seed, 
seed_transpiler=seed, circuit_caching=False, skip_qobj_validation=False)
chunfuchen commented 5 years ago

@hykavitha do you still encounter any other issue? if no, I think we can close this issue.

hykavitha commented 5 years ago

@chunfuchen : I still face this issue. Once its fixed and confirmed after release is when you would close? or its fixed & tested at qiskit aqua developing stage?

chunfuchen commented 5 years ago

@hykavitha what is the issue you encounter now? could changing setting of the quantum instance be a workaround?

I think this is fixed in the master branch. I can run the notebook as is if the master branch is used.

hykavitha commented 5 years ago

@chunfuchen : I still cant use the IBMQ device from my public account. Qiskit errors out with error key 1. When you say its fixed in the master branch. What does it mean for us having these versions of qiskit.

I have these versions: Qiskit Aqua version: pip list | grep "qiskit" qiskit 0.10.1 qiskit-aer 0.2.0 qiskit-aqua 0.5.0 qiskit-chemistry 0.5.0 qiskit-ibmq-provider 0.2.2 qiskit-ignis 0.1.1 qiskit-terra 0.8.0 Python version: 3.73 Operating system: OSX

chunfuchen commented 5 years ago

I mean that you need to install everything from the source code, here is my qiskit version

qiskit-aer                  0.3.0
qiskit-aqua                 0.5.2      
qiskit-chemistry            0.5.1    
qiskit-ibmq-provider        0.2.3rc2  
qiskit-ignis                0.2.0    
qiskit-terra                0.9.0

however, do you mind posting the error log when you turn off the circuit cache? Thanks

woodsp-ibm commented 5 years ago

This should no longer occur in Aqua 0.6. Closing, if the problem that you experienced still continues for you either reopen or raise a new issue.