Quanvolution (Quantum convolution) for MNIST image classification with TorchQuantum.#
Tutorial Author: Zirui Li, Hanrui Wang
Outline#
Introduction to Quanvolutional Neural Network.
Build and train a Quanvolutional Neural Network.
Compare Quanvolutional Neural Network with a classic model.
Evaluate on real quantum computer.
Compare multiple models with or without a trainable quanvolutional filter.
In this tutorial, we use tq.QuantumDevice, tq.GeneralEncoder, tq.RandomLayer, tq.MeasureAll, tq.PauliZ class from TrochQuantum.
You can learn how to build, train and evaluate a quanvolutional filter using TorchQuantum in this tutorial.
Introduction to Quanvolutional Neural Network.#
Convolutional Neural Network#
Convolutional neural network is a classic neural network genre, mostly applied to anylize visual images. They are known for their convolutional layers that perform convolution. Typically the convolution operation is the Frobenius inner product of the convolution filter with the input image followed by an activation function. The convolution filter slides along the input image and generates a feature map. We can use the feature map for classification.
Quantum convolution#
One can extend the same idea also to the context of quantum variational circuits. Replace the classical convolution filters with variational quantum circuits and we get quanvolutional neural networks with quanvolutional filters. The quanvolutional filters perform more complex operations in a higher dimension Hilbert space than Frobenius inner product. Therefore, quanvolutional filters have more potential than traditional convolution filters.
Build and train a Quanvolutional Neural Network.#
Installation#
Install torchquantum and all the libs we need.
[ ]:
!pip install qiskit==0.32.1
Collecting qiskit==0.32.1
Downloading qiskit-0.32.1.tar.gz (13 kB)
Collecting qiskit-terra==0.18.3
Downloading qiskit_terra-0.18.3-cp37-cp37m-manylinux2010_x86_64.whl (6.1 MB)
|████████████████████████████████| 6.1 MB 4.1 MB/s
Collecting qiskit-aer==0.9.1
Downloading qiskit_aer-0.9.1-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (17.9 MB)
|████████████████████████████████| 17.9 MB 633 kB/s
Collecting qiskit-ibmq-provider==0.18.1
Downloading qiskit_ibmq_provider-0.18.1-py3-none-any.whl (237 kB)
|████████████████████████████████| 237 kB 73.3 MB/s
Collecting qiskit-ignis==0.6.0
Downloading qiskit_ignis-0.6.0-py3-none-any.whl (207 kB)
|████████████████████████████████| 207 kB 65.4 MB/s
Collecting qiskit-aqua==0.9.5
Downloading qiskit_aqua-0.9.5-py3-none-any.whl (2.1 MB)
|████████████████████████████████| 2.1 MB 60.5 MB/s
Requirement already satisfied: scipy>=1.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-aer==0.9.1->qiskit==0.32.1) (1.4.1)
Requirement already satisfied: numpy>=1.16.3 in /usr/local/lib/python3.7/dist-packages (from qiskit-aer==0.9.1->qiskit==0.32.1) (1.21.5)
Requirement already satisfied: h5py<3.3.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit==0.32.1) (3.1.0)
Collecting quandl
Downloading Quandl-3.7.0-py2.py3-none-any.whl (26 kB)
Requirement already satisfied: scikit-learn>=0.20.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit==0.32.1) (1.0.2)
Collecting yfinance>=0.1.62
Downloading yfinance-0.1.70-py2.py3-none-any.whl (26 kB)
Requirement already satisfied: psutil>=5 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit==0.32.1) (5.4.8)
Collecting docplex>=2.21.207
Downloading docplex-2.22.213.tar.gz (634 kB)
|████████████████████████████████| 634 kB 68.2 MB/s
Requirement already satisfied: setuptools>=40.1.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit==0.32.1) (57.4.0)
Requirement already satisfied: sympy>=1.3 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit==0.32.1) (1.7.1)
Collecting retworkx>=0.8.0
Downloading retworkx-0.11.0-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (1.6 MB)
|████████████████████████████████| 1.6 MB 21.6 MB/s
Requirement already satisfied: fastdtw<=0.3.4 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit==0.32.1) (0.3.4)
Collecting dlx<=1.0.4
Downloading dlx-1.0.4.tar.gz (5.5 kB)
Requirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit==0.32.1) (1.3.5)
Requirement already satisfied: urllib3>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from qiskit-ibmq-provider==0.18.1->qiskit==0.32.1) (1.24.3)
Requirement already satisfied: python-dateutil>=2.8.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-ibmq-provider==0.18.1->qiskit==0.32.1) (2.8.2)
Collecting requests-ntlm>=1.1.0
Downloading requests_ntlm-1.1.0-py2.py3-none-any.whl (5.7 kB)
Requirement already satisfied: requests>=2.19 in /usr/local/lib/python3.7/dist-packages (from qiskit-ibmq-provider==0.18.1->qiskit==0.32.1) (2.23.0)
Collecting websocket-client>=1.0.1
Downloading websocket_client-1.2.3-py3-none-any.whl (53 kB)
|████████████████████████████████| 53 kB 2.7 MB/s
Collecting python-constraint>=1.4
Downloading python-constraint-1.4.0.tar.bz2 (18 kB)
Collecting fastjsonschema>=2.10
Downloading fastjsonschema-2.15.3-py3-none-any.whl (22 kB)
Collecting symengine>0.7
Downloading symengine-0.8.1-cp37-cp37m-manylinux2010_x86_64.whl (38.2 MB)
|████████████████████████████████| 38.2 MB 116 kB/s
Collecting tweedledum<2.0,>=1.1
Downloading tweedledum-1.1.1-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (943 kB)
|████████████████████████████████| 943 kB 22.1 MB/s
Collecting ply>=3.10
Downloading ply-3.11-py2.py3-none-any.whl (49 kB)
|████████████████████████████████| 49 kB 8.3 MB/s
Requirement already satisfied: dill>=0.3 in /usr/local/lib/python3.7/dist-packages (from qiskit-terra==0.18.3->qiskit==0.32.1) (0.3.4)
Requirement already satisfied: jsonschema>=2.6 in /usr/local/lib/python3.7/dist-packages (from qiskit-terra==0.18.3->qiskit==0.32.1) (4.3.3)
Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from docplex>=2.21.207->qiskit-aqua==0.9.5->qiskit==0.32.1) (1.15.0)
Requirement already satisfied: cached-property in /usr/local/lib/python3.7/dist-packages (from h5py<3.3.0->qiskit-aqua==0.9.5->qiskit==0.32.1) (1.5.2)
Requirement already satisfied: importlib-resources>=1.4.0 in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit==0.32.1) (5.4.0)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit==0.32.1) (3.10.0.2)
Requirement already satisfied: attrs>=17.4.0 in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit==0.32.1) (21.4.0)
Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit==0.32.1) (4.11.0)
Requirement already satisfied: pyrsistent!=0.17.0,!=0.17.1,!=0.17.2,>=0.14.0 in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit==0.32.1) (0.18.1)
Requirement already satisfied: zipp>=3.1.0 in /usr/local/lib/python3.7/dist-packages (from importlib-resources>=1.4.0->jsonschema>=2.6->qiskit-terra==0.18.3->qiskit==0.32.1) (3.7.0)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19->qiskit-ibmq-provider==0.18.1->qiskit==0.32.1) (2021.10.8)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19->qiskit-ibmq-provider==0.18.1->qiskit==0.32.1) (2.10)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19->qiskit-ibmq-provider==0.18.1->qiskit==0.32.1) (3.0.4)
Collecting cryptography>=1.3
Downloading cryptography-36.0.1-cp36-abi3-manylinux_2_24_x86_64.whl (3.6 MB)
|████████████████████████████████| 3.6 MB 31.3 MB/s
Collecting ntlm-auth>=1.0.2
Downloading ntlm_auth-1.5.0-py2.py3-none-any.whl (29 kB)
Requirement already satisfied: cffi>=1.12 in /usr/local/lib/python3.7/dist-packages (from cryptography>=1.3->requests-ntlm>=1.1.0->qiskit-ibmq-provider==0.18.1->qiskit==0.32.1) (1.15.0)
Requirement already satisfied: pycparser in /usr/local/lib/python3.7/dist-packages (from cffi>=1.12->cryptography>=1.3->requests-ntlm>=1.1.0->qiskit-ibmq-provider==0.18.1->qiskit==0.32.1) (2.21)
Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.20.0->qiskit-aqua==0.9.5->qiskit==0.32.1) (1.1.0)
Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.20.0->qiskit-aqua==0.9.5->qiskit==0.32.1) (3.1.0)
Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.7/dist-packages (from sympy>=1.3->qiskit-aqua==0.9.5->qiskit==0.32.1) (1.2.1)
Collecting requests>=2.19
Downloading requests-2.27.1-py2.py3-none-any.whl (63 kB)
|████████████████████████████████| 63 kB 814 kB/s
Requirement already satisfied: multitasking>=0.0.7 in /usr/local/lib/python3.7/dist-packages (from yfinance>=0.1.62->qiskit-aqua==0.9.5->qiskit==0.32.1) (0.0.10)
Collecting lxml>=4.5.1
Downloading lxml-4.7.1-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.manylinux_2_24_x86_64.whl (6.4 MB)
|████████████████████████████████| 6.4 MB 30.3 MB/s
Requirement already satisfied: pytz>=2017.3 in /usr/local/lib/python3.7/dist-packages (from pandas->qiskit-aqua==0.9.5->qiskit==0.32.1) (2018.9)
Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19->qiskit-ibmq-provider==0.18.1->qiskit==0.32.1) (2.0.11)
Collecting inflection>=0.3.1
Downloading inflection-0.5.1-py2.py3-none-any.whl (9.5 kB)
Requirement already satisfied: more-itertools in /usr/local/lib/python3.7/dist-packages (from quandl->qiskit-aqua==0.9.5->qiskit==0.32.1) (8.12.0)
Building wheels for collected packages: qiskit, dlx, docplex, python-constraint
Building wheel for qiskit (setup.py) ... done
Created wheel for qiskit: filename=qiskit-0.32.1-py3-none-any.whl size=11777 sha256=911365fec91e5c648d2569b156af429c0aff7c3d95d453a7d24e6bf8d7d1a315
Stored in directory: /root/.cache/pip/wheels/0f/62/0a/c53eda1ead41c137c47c9730bc2771a8367b1ce00fb64e8cc6
Building wheel for dlx (setup.py) ... done
Created wheel for dlx: filename=dlx-1.0.4-py3-none-any.whl size=5718 sha256=cb913d8c2b19d87e8784f4220a02752c4858e3ebe531b0e80ab22c5625c1bd0b
Stored in directory: /root/.cache/pip/wheels/78/55/c8/dc61e772445a566b7608a476d151e9dcaf4e092b01b0c4bc3c
Building wheel for docplex (setup.py) ... done
Created wheel for docplex: filename=docplex-2.22.213-py3-none-any.whl size=696882 sha256=192bab0a2587503608ce12a090c9f129f2a6b0e88f3a41e568c07ca585b4e3ff
Stored in directory: /root/.cache/pip/wheels/90/69/6b/1375c68a5b7ff94c40263b151c86f58bd72200bf0c465b5ba3
Building wheel for python-constraint (setup.py) ... done
Created wheel for python-constraint: filename=python_constraint-1.4.0-py2.py3-none-any.whl size=24081 sha256=77684dcb7c666715267053a0114cf933c5c52cb7af9a30645de961f9af12c323
Stored in directory: /root/.cache/pip/wheels/07/27/db/1222c80eb1e431f3d2199c12569cb1cac60f562a451fe30479
Successfully built qiskit dlx docplex python-constraint
Installing collected packages: tweedledum, symengine, retworkx, python-constraint, ply, fastjsonschema, requests, qiskit-terra, ntlm-auth, lxml, inflection, cryptography, yfinance, websocket-client, requests-ntlm, quandl, qiskit-ignis, docplex, dlx, qiskit-ibmq-provider, qiskit-aqua, qiskit-aer, qiskit
Attempting uninstall: requests
Found existing installation: requests 2.23.0
Uninstalling requests-2.23.0:
Successfully uninstalled requests-2.23.0
Attempting uninstall: lxml
Found existing installation: lxml 4.2.6
Uninstalling lxml-4.2.6:
Successfully uninstalled lxml-4.2.6
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
google-colab 1.0.0 requires requests~=2.23.0, but you have requests 2.27.1 which is incompatible.
datascience 0.10.6 requires folium==0.2.1, but you have folium 0.8.3 which is incompatible.
Successfully installed cryptography-36.0.1 dlx-1.0.4 docplex-2.22.213 fastjsonschema-2.15.3 inflection-0.5.1 lxml-4.7.1 ntlm-auth-1.5.0 ply-3.11 python-constraint-1.4.0 qiskit-0.32.1 qiskit-aer-0.9.1 qiskit-aqua-0.9.5 qiskit-ibmq-provider-0.18.1 qiskit-ignis-0.6.0 qiskit-terra-0.18.3 quandl-3.7.0 requests-2.27.1 requests-ntlm-1.1.0 retworkx-0.11.0 symengine-0.8.1 tweedledum-1.1.1 websocket-client-1.2.3 yfinance-0.1.70
Download and cd to the repo.
[ ]:
!git clone https://github.com/mit-han-lab/torchquantum.git
Cloning into 'torchquantum'...
remote: Enumerating objects: 10737, done.
remote: Counting objects: 100% (7529/7529), done.
remote: Compressing objects: 100% (3777/3777), done.
remote: Total 10737 (delta 3765), reused 7076 (delta 3348), pack-reused 3208
Receiving objects: 100% (10737/10737), 3.19 MiB | 12.92 MiB/s, done.
Resolving deltas: 100% (5732/5732), done.
Checking out files: 100% (50055/50055), done.
[ ]:
%cd torchquantum
/content/torchquantum
Install torch-quantum.
[ ]:
!pip install --editable .
Obtaining file:///content/torchquantum
Requirement already satisfied: numpy>=1.19.2 in /usr/local/lib/python3.7/dist-packages (from torchquantum==0.1.0) (1.21.5)
Requirement already satisfied: torchvision>=0.9.0.dev20210130 in /usr/local/lib/python3.7/dist-packages (from torchquantum==0.1.0) (0.11.1+cu111)
Requirement already satisfied: tqdm>=4.56.0 in /usr/local/lib/python3.7/dist-packages (from torchquantum==0.1.0) (4.62.3)
Requirement already satisfied: setuptools>=52.0.0 in /usr/local/lib/python3.7/dist-packages (from torchquantum==0.1.0) (57.4.0)
Requirement already satisfied: torch>=1.8.0 in /usr/local/lib/python3.7/dist-packages (from torchquantum==0.1.0) (1.10.0+cu111)
Collecting torchpack>=0.3.0
Downloading torchpack-0.3.1-py3-none-any.whl (34 kB)
Requirement already satisfied: qiskit>=0.32.0 in /usr/local/lib/python3.7/dist-packages (from torchquantum==0.1.0) (0.32.1)
Collecting matplotlib>=3.3.2
Downloading matplotlib-3.5.1-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.whl (11.2 MB)
|████████████████████████████████| 11.2 MB 6.5 MB/s
Collecting pathos>=0.2.7
Downloading pathos-0.2.8-py2.py3-none-any.whl (81 kB)
|████████████████████████████████| 81 kB 12.1 MB/s
Requirement already satisfied: python-dateutil>=2.7 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=3.3.2->torchquantum==0.1.0) (2.8.2)
Requirement already satisfied: pyparsing>=2.2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=3.3.2->torchquantum==0.1.0) (3.0.7)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=3.3.2->torchquantum==0.1.0) (0.11.0)
Collecting fonttools>=4.22.0
Downloading fonttools-4.29.1-py3-none-any.whl (895 kB)
|████████████████████████████████| 895 kB 55.2 MB/s
Requirement already satisfied: pillow>=6.2.0 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=3.3.2->torchquantum==0.1.0) (7.1.2)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=3.3.2->torchquantum==0.1.0) (1.3.2)
Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=3.3.2->torchquantum==0.1.0) (21.3)
Collecting ppft>=1.6.6.4
Downloading ppft-1.6.6.4-py3-none-any.whl (65 kB)
|████████████████████████████████| 65 kB 4.1 MB/s
Collecting pox>=0.3.0
Downloading pox-0.3.0-py2.py3-none-any.whl (30 kB)
Requirement already satisfied: multiprocess>=0.70.12 in /usr/local/lib/python3.7/dist-packages (from pathos>=0.2.7->torchquantum==0.1.0) (0.70.12.2)
Requirement already satisfied: dill>=0.3.4 in /usr/local/lib/python3.7/dist-packages (from pathos>=0.2.7->torchquantum==0.1.0) (0.3.4)
Requirement already satisfied: six>=1.7.3 in /usr/local/lib/python3.7/dist-packages (from ppft>=1.6.6.4->pathos>=0.2.7->torchquantum==0.1.0) (1.15.0)
Requirement already satisfied: qiskit-ibmq-provider==0.18.1 in /usr/local/lib/python3.7/dist-packages (from qiskit>=0.32.0->torchquantum==0.1.0) (0.18.1)
Requirement already satisfied: qiskit-aqua==0.9.5 in /usr/local/lib/python3.7/dist-packages (from qiskit>=0.32.0->torchquantum==0.1.0) (0.9.5)
Requirement already satisfied: qiskit-aer==0.9.1 in /usr/local/lib/python3.7/dist-packages (from qiskit>=0.32.0->torchquantum==0.1.0) (0.9.1)
Requirement already satisfied: qiskit-ignis==0.6.0 in /usr/local/lib/python3.7/dist-packages (from qiskit>=0.32.0->torchquantum==0.1.0) (0.6.0)
Requirement already satisfied: qiskit-terra==0.18.3 in /usr/local/lib/python3.7/dist-packages (from qiskit>=0.32.0->torchquantum==0.1.0) (0.18.3)
Requirement already satisfied: scipy>=1.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-aer==0.9.1->qiskit>=0.32.0->torchquantum==0.1.0) (1.4.1)
Requirement already satisfied: quandl in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (3.7.0)
Requirement already satisfied: scikit-learn>=0.20.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (1.0.2)
Requirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (1.3.5)
Requirement already satisfied: h5py<3.3.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (3.1.0)
Requirement already satisfied: fastdtw<=0.3.4 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (0.3.4)
Requirement already satisfied: dlx<=1.0.4 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (1.0.4)
Requirement already satisfied: retworkx>=0.8.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (0.11.0)
Requirement already satisfied: sympy>=1.3 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (1.7.1)
Requirement already satisfied: docplex>=2.21.207 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (2.22.213)
Requirement already satisfied: yfinance>=0.1.62 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (0.1.70)
Requirement already satisfied: psutil>=5 in /usr/local/lib/python3.7/dist-packages (from qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (5.4.8)
Requirement already satisfied: urllib3>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (1.24.3)
Requirement already satisfied: requests-ntlm>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (1.1.0)
Requirement already satisfied: websocket-client>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (1.2.3)
Requirement already satisfied: requests>=2.19 in /usr/local/lib/python3.7/dist-packages (from qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (2.27.1)
Requirement already satisfied: symengine>0.7 in /usr/local/lib/python3.7/dist-packages (from qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (0.8.1)
Requirement already satisfied: fastjsonschema>=2.10 in /usr/local/lib/python3.7/dist-packages (from qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (2.15.3)
Requirement already satisfied: ply>=3.10 in /usr/local/lib/python3.7/dist-packages (from qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (3.11)
Requirement already satisfied: python-constraint>=1.4 in /usr/local/lib/python3.7/dist-packages (from qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (1.4.0)
Requirement already satisfied: jsonschema>=2.6 in /usr/local/lib/python3.7/dist-packages (from qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (4.3.3)
Requirement already satisfied: tweedledum<2.0,>=1.1 in /usr/local/lib/python3.7/dist-packages (from qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (1.1.1)
Requirement already satisfied: cached-property in /usr/local/lib/python3.7/dist-packages (from h5py<3.3.0->qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (1.5.2)
Requirement already satisfied: attrs>=17.4.0 in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (21.4.0)
Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (4.11.0)
Requirement already satisfied: importlib-resources>=1.4.0 in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (5.4.0)
Requirement already satisfied: pyrsistent!=0.17.0,!=0.17.1,!=0.17.2,>=0.14.0 in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (0.18.1)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from jsonschema>=2.6->qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (3.10.0.2)
Requirement already satisfied: zipp>=3.1.0 in /usr/local/lib/python3.7/dist-packages (from importlib-resources>=1.4.0->jsonschema>=2.6->qiskit-terra==0.18.3->qiskit>=0.32.0->torchquantum==0.1.0) (3.7.0)
Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19->qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (2.10)
Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19->qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (2.0.11)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19->qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (2021.10.8)
Requirement already satisfied: cryptography>=1.3 in /usr/local/lib/python3.7/dist-packages (from requests-ntlm>=1.1.0->qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (36.0.1)
Requirement already satisfied: ntlm-auth>=1.0.2 in /usr/local/lib/python3.7/dist-packages (from requests-ntlm>=1.1.0->qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (1.5.0)
Requirement already satisfied: cffi>=1.12 in /usr/local/lib/python3.7/dist-packages (from cryptography>=1.3->requests-ntlm>=1.1.0->qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (1.15.0)
Requirement already satisfied: pycparser in /usr/local/lib/python3.7/dist-packages (from cffi>=1.12->cryptography>=1.3->requests-ntlm>=1.1.0->qiskit-ibmq-provider==0.18.1->qiskit>=0.32.0->torchquantum==0.1.0) (2.21)
Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.20.0->qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (1.1.0)
Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.20.0->qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (3.1.0)
Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.7/dist-packages (from sympy>=1.3->qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (1.2.1)
Requirement already satisfied: tensorboard in /usr/local/lib/python3.7/dist-packages (from torchpack>=0.3.0->torchquantum==0.1.0) (2.8.0)
Collecting toml
Downloading toml-0.10.2-py2.py3-none-any.whl (16 kB)
Collecting tensorpack
Downloading tensorpack-0.11-py2.py3-none-any.whl (296 kB)
|████████████████████████████████| 296 kB 57.1 MB/s
Collecting multimethod
Downloading multimethod-1.7-py3-none-any.whl (9.5 kB)
Collecting loguru
Downloading loguru-0.6.0-py3-none-any.whl (58 kB)
|████████████████████████████████| 58 kB 2.6 MB/s
Requirement already satisfied: pyyaml in /usr/local/lib/python3.7/dist-packages (from torchpack>=0.3.0->torchquantum==0.1.0) (3.13)
Requirement already satisfied: multitasking>=0.0.7 in /usr/local/lib/python3.7/dist-packages (from yfinance>=0.1.62->qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (0.0.10)
Requirement already satisfied: lxml>=4.5.1 in /usr/local/lib/python3.7/dist-packages (from yfinance>=0.1.62->qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (4.7.1)
Requirement already satisfied: pytz>=2017.3 in /usr/local/lib/python3.7/dist-packages (from pandas->qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (2018.9)
Requirement already satisfied: more-itertools in /usr/local/lib/python3.7/dist-packages (from quandl->qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (8.12.0)
Requirement already satisfied: inflection>=0.3.1 in /usr/local/lib/python3.7/dist-packages (from quandl->qiskit-aqua==0.9.5->qiskit>=0.32.0->torchquantum==0.1.0) (0.5.1)
Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (3.3.6)
Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (1.8.1)
Requirement already satisfied: absl-py>=0.4 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (1.0.0)
Requirement already satisfied: grpcio>=1.24.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (1.43.0)
Requirement already satisfied: wheel>=0.26 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (0.37.1)
Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (0.4.6)
Requirement already satisfied: google-auth<3,>=1.6.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (1.35.0)
Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (0.6.1)
Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (1.0.1)
Requirement already satisfied: protobuf>=3.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (3.17.3)
Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (0.2.8)
Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (4.2.4)
Requirement already satisfied: rsa<5,>=3.1.4 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (4.8)
Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.7/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (1.3.1)
Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.7/dist-packages (from pyasn1-modules>=0.2.1->google-auth<3,>=1.6.3->tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (0.4.8)
Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard->torchpack>=0.3.0->torchquantum==0.1.0) (3.2.0)
Requirement already satisfied: msgpack>=0.5.2 in /usr/local/lib/python3.7/dist-packages (from tensorpack->torchpack>=0.3.0->torchquantum==0.1.0) (1.0.3)
Requirement already satisfied: pyzmq>=16 in /usr/local/lib/python3.7/dist-packages (from tensorpack->torchpack>=0.3.0->torchquantum==0.1.0) (22.3.0)
Collecting msgpack-numpy>=0.4.4.2
Downloading msgpack_numpy-0.4.7.1-py2.py3-none-any.whl (6.7 kB)
Requirement already satisfied: tabulate>=0.7.7 in /usr/local/lib/python3.7/dist-packages (from tensorpack->torchpack>=0.3.0->torchquantum==0.1.0) (0.8.9)
Requirement already satisfied: termcolor>=1.1 in /usr/local/lib/python3.7/dist-packages (from tensorpack->torchpack>=0.3.0->torchquantum==0.1.0) (1.1.0)
Installing collected packages: msgpack-numpy, toml, tensorpack, ppft, pox, multimethod, loguru, fonttools, torchpack, pathos, matplotlib, torchquantum
Attempting uninstall: matplotlib
Found existing installation: matplotlib 3.2.2
Uninstalling matplotlib-3.2.2:
Successfully uninstalled matplotlib-3.2.2
Running setup.py develop for torchquantum
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
albumentations 0.1.12 requires imgaug<0.2.7,>=0.2.5, but you have imgaug 0.2.9 which is incompatible.
Successfully installed fonttools-4.29.1 loguru-0.6.0 matplotlib-3.5.1 msgpack-numpy-0.4.7.1 multimethod-1.7 pathos-0.2.8 pox-0.3.0 ppft-1.6.6.4 tensorpack-0.11 toml-0.10.2 torchpack-0.3.1 torchquantum-0.1.0
Data type cannot be displayed: application/vnd.colab-display-data+json
Change PYTHONPATH and install other packages.
[ ]:
%env PYTHONPATH=.
env: PYTHONPATH=.
Run the following code to store a qiskit token. You can replace it with your own token from your IBMQ account if you like.
[ ]:
from qiskit import IBMQ
# IBMQ.save_account('', overwrite=True)
[ ]:
!pip install matplotlib==3.1.3
Collecting matplotlib==3.1.3
Downloading matplotlib-3.1.3-cp37-cp37m-manylinux1_x86_64.whl (13.1 MB)
|████████████████████████████████| 13.1 MB 4.3 MB/s
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.1.3) (3.0.7)
Requirement already satisfied: numpy>=1.11 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.1.3) (1.21.5)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.1.3) (0.11.0)
Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.1.3) (2.8.2)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.1.3) (1.3.2)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib==3.1.3) (1.15.0)
Installing collected packages: matplotlib
Attempting uninstall: matplotlib
Found existing installation: matplotlib 3.5.1
Uninstalling matplotlib-3.5.1:
Successfully uninstalled matplotlib-3.5.1
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
torchquantum 0.1.0 requires matplotlib>=3.3.2, but you have matplotlib 3.1.3 which is incompatible.
albumentations 0.1.12 requires imgaug<0.2.7,>=0.2.5, but you have imgaug 0.2.9 which is incompatible.
Successfully installed matplotlib-3.1.3
Data type cannot be displayed: application/vnd.colab-display-data+json
Step#
Our code requires torchquantum lib, mnist dataset, pytorch and numpy. We need torch and the logsoftmax function from torch.nn.functional, optimizers(optim), torchquantum module, MNIST dataset(MNIST), cosine annealing learning rate(CosineAnnealingLR).
[ ]:
import torch
import torch.nn.functional as F
import torch.optim as optim
import numpy as np
import torchquantum as tq
import random
from torchquantum.datasets import MNIST
from torch.optim.lr_scheduler import CosineAnnealingLR
Build a quanvolutional filter#
Our quanvolution model is a hybrid model. It consists of two parts, the quanvolutional filter part and the classical layer part. To build the model, firstly we define our quanvolutional filter.
Our quanvolutional filter’s structure is the same as the figure described above. It has four qubits. The tq.QuantumDevice module stores the state vector. Usually a Quantum Neural Network module consists of three parts: encoder, ansatz and measurement. We can create an encoder by passing a list of gates to tq.GeneralEncoder. Each entry in the list contains input_idx, func, and wires. Here, each qubit has a rotation-Y gate. 4 RY gates in total. They can encode the 2x2 input
data to the quantum state. Then we decide our ansatz to be a random layer. We call tq.RandomLayer to create an ansatz composed by 8 basic gates with no more than 8 trainable parameters. And finally we perform Pauli-Z measurements on each qubit by creating a tq.MeasureAll module and passing tq.PauliZ to it. The measure function will return four expectation values from four qubits. The four results go to four channels.
Next look at how quanvolutional filter works. We get the batch size. Our image is 28x28. So we reshape our input data to (bsz, 28, 28).
We initialize the data_list. The list stores the outputs in each stride.
The double loop is to iterate all the possible positions that the quanvolutional filter window may stride in. Here the stride is 2.
Then we catenate the data in the 2x2 window. Here we catenate four lists to one big list, so we need to reshape the list to (4, bsz) and transpose it to (bsz, 4).
Next if you want to use qiskit’s remote noise model or real quantum machine, you can set use_qiskit=True and pass these 5 parameters: q_device, encoder, q_layer, measure, and data. The qiskit_processor will receive these parameters, put the data in the encoder, run the while circuits and return the measurement result. Remember only when the model is doing an inference can you use qiskit remote.
If you are training or not using qiskit remote, you can run the three parts one by one on google colab’s GPU.
After each stride, we append the measurement result to data_list.
Finally, we catenate the data_list along dimension 1 and return the result.
[ ]:
class QuanvolutionFilter(tq.QuantumModule):
def __init__(self):
super().__init__()
self.n_wires = 4
self.q_device = tq.QuantumDevice(n_wires=self.n_wires)
self.encoder = tq.GeneralEncoder(
[ {'input_idx': [0], 'func': 'ry', 'wires': [0]},
{'input_idx': [1], 'func': 'ry', 'wires': [1]},
{'input_idx': [2], 'func': 'ry', 'wires': [2]},
{'input_idx': [3], 'func': 'ry', 'wires': [3]},])
self.q_layer = tq.RandomLayer(n_ops=8, wires=list(range(self.n_wires)))
self.measure = tq.MeasureAll(tq.PauliZ)
def forward(self, x, use_qiskit=False):
bsz = x.shape[0]
size = 28
x = x.view(bsz, size, size)
data_list = []
for c in range(0, size, 2):
for r in range(0, size, 2):
data = torch.transpose(torch.cat((x[:, c, r], x[:, c, r+1], x[:, c+1, r], x[:, c+1, r+1])).view(4, bsz), 0, 1)
if use_qiskit:
data = self.qiskit_processor.process_parameterized(
self.q_device, self.encoder, self.q_layer, self.measure, data)
else:
self.encoder(self.q_device, data)
self.q_layer(self.q_device)
data = self.measure(self.q_device)
data_list.append(data.view(bsz, 4))
result = torch.cat(data_list, dim=1).float()
return result
Build the whole hybrid model.#
Then we look at the whole model. The whole model consists of a QuanvolutionFilter and full connect layer(torch.nn.Linear). The size of input is 4*14*14 because a 28x28 image after quanvolutional filter turns into a 4 channel 14x14 feature. As the task is MNIST 10 digits classification, the size of output is 10. At last the model perform F.logsoftmax to the result for classification.
Here, we also has a model without quanvolutional filters used for comparison. Its full connect layer’s input size is simple 28x28.
[ ]:
class HybridModel(torch.nn.Module):
def __init__(self):
super().__init__()
self.qf = QuanvolutionFilter()
self.linear = torch.nn.Linear(4*14*14, 10)
def forward(self, x, use_qiskit=False):
with torch.no_grad():
x = self.qf(x, use_qiskit)
x = self.linear(x)
return F.log_softmax(x, -1)
class HybridModel_without_qf(torch.nn.Module):
def __init__(self):
super().__init__()
self.linear = torch.nn.Linear(28*28, 10)
def forward(self, x, use_qiskit=False):
x = x.view(-1, 28*28)
x = self.linear(x)
return F.log_softmax(x, -1)
Load the dataset MNIST#
We use MNIST classification dataset(10 digits and 1000 training samples).
The root is the folder that stores the dataset. If there’s no MNIST dataset in root, it will automatically download MNIST. Next, we set the train_valid_split_ratio, n_test_samples, and n_train_samples.
The dataset now contains three splits, ‘train’, ‘valid’ and ‘test’. For each split, we create a dataloader with a random sampler, batch_size is 10, num_workers is 8 and pin_memory is true.
[ ]:
random.seed(42)
np.random.seed(42)
torch.manual_seed(42)
dataset = MNIST(
root='./mnist_data',
train_valid_split_ratio=[0.9, 0.1],
n_test_samples=300,
n_train_samples=500,
)
dataflow = dict()
for split in dataset:
sampler = torch.utils.data.RandomSampler(dataset[split])
dataflow[split] = torch.utils.data.DataLoader(
dataset[split],
batch_size=10,
sampler=sampler,
num_workers=8,
pin_memory=True)
Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to ./mnist_data/MNIST/raw/train-images-idx3-ubyte.gz
Extracting ./mnist_data/MNIST/raw/train-images-idx3-ubyte.gz to ./mnist_data/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to ./mnist_data/MNIST/raw/train-labels-idx1-ubyte.gz
Extracting ./mnist_data/MNIST/raw/train-labels-idx1-ubyte.gz to ./mnist_data/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to ./mnist_data/MNIST/raw/t10k-images-idx3-ubyte.gz
Extracting ./mnist_data/MNIST/raw/t10k-images-idx3-ubyte.gz to ./mnist_data/MNIST/raw
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to ./mnist_data/MNIST/raw/t10k-labels-idx1-ubyte.gz
[2022-02-16 04:00:40.771] Only use the front 500 images as TRAIN set.
Extracting ./mnist_data/MNIST/raw/t10k-labels-idx1-ubyte.gz to ./mnist_data/MNIST/raw
[2022-02-16 04:00:40.868] Only use the front 300 images as TEST set.
/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py:481: UserWarning: This DataLoader will create 8 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
cpuset_checked))
Then we set use_cuda, it depends on whether cuda is available.
Create a device.
Initialize the model, n_epochs to 15, Adam optimizer and cosine annealing learning rate scheduler.
[ ]:
use_cuda = torch.cuda.is_available()
device = torch.device("cuda" if use_cuda else "cpu")
model = HybridModel().to(device)
model_without_qf = HybridModel_without_qf().to(device)
n_epochs = 15
optimizer = optim.Adam(model.parameters(), lr=5e-3, weight_decay=1e-4)
scheduler = CosineAnnealingLR(optimizer, T_max=n_epochs)
Train the model.#
When training the model, we iterate the dataloader. Get the inputs and targets data. Feed inputs to the model and get outputs. Calculate the negative loss likelihood loss(F.nll_loss). Reset all the gradients of parameters in the model to zero. Call loss.backward() to perform backpropagation. Call optimizer.step() to update all the parameters.
After each epoch, we will valid the model. In validation, we can use qiskit remote because we don’t need to calculate gradients.
[ ]:
accu_list1 = []
loss_list1 = []
accu_list2 = []
loss_list2 = []
def train(dataflow, model, device, optimizer):
for feed_dict in dataflow['train']:
inputs = feed_dict['image'].to(device)
targets = feed_dict['digit'].to(device)
outputs = model(inputs)
loss = F.nll_loss(outputs, targets)
optimizer.zero_grad()
loss.backward()
optimizer.step()
print(f"loss: {loss.item()}", end='\r')
def valid_test(dataflow, split, model, device, qiskit=False):
target_all = []
output_all = []
with torch.no_grad():
for feed_dict in dataflow[split]:
inputs = feed_dict['image'].to(device)
targets = feed_dict['digit'].to(device)
outputs = model(inputs, use_qiskit=qiskit)
target_all.append(targets)
output_all.append(outputs)
target_all = torch.cat(target_all, dim=0)
output_all = torch.cat(output_all, dim=0)
_, indices = output_all.topk(1, dim=1)
masks = indices.eq(target_all.view(-1, 1).expand_as(indices))
size = target_all.shape[0]
corrects = masks.sum().item()
accuracy = corrects / size
loss = F.nll_loss(output_all, target_all).item()
print(f"{split} set accuracy: {accuracy}")
print(f"{split} set loss: {loss}")
return accuracy, loss
for epoch in range(1, n_epochs + 1):
# train
print(f"Epoch {epoch}:")
train(dataflow, model, device, optimizer)
print(optimizer.param_groups[0]['lr'])
# valid
accu, loss = valid_test(dataflow, 'test', model, device, )
accu_list1.append(accu)
loss_list1.append(loss)
scheduler.step()
Epoch 1:
/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py:481: UserWarning: This DataLoader will create 8 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
cpuset_checked))
0.005
test set accuracy: 0.8066666666666666
test set loss: 0.6323180794715881
Epoch 2:
0.004945369001834514
test set accuracy: 0.7766666666666666
test set loss: 0.5900668501853943
Epoch 3:
0.004783863644106502
test set accuracy: 0.8466666666666667
test set loss: 0.48249581456184387
Epoch 4:
0.0045225424859373685
test set accuracy: 0.8133333333333334
test set loss: 0.5225163698196411
Epoch 5:
0.0041728265158971455
test set accuracy: 0.8033333333333333
test set loss: 0.6009621620178223
Epoch 6:
0.00375
test set accuracy: 0.8333333333333334
test set loss: 0.44394049048423767
Epoch 7:
0.0032725424859373687
test set accuracy: 0.84
test set loss: 0.4330306053161621
Epoch 8:
0.002761321158169134
test set accuracy: 0.8366666666666667
test set loss: 0.45171523094177246
Epoch 9:
0.002238678841830867
test set accuracy: 0.8633333333333333
test set loss: 0.4244077205657959
Epoch 10:
0.001727457514062632
test set accuracy: 0.8633333333333333
test set loss: 0.40085339546203613
Epoch 11:
0.0012500000000000007
test set accuracy: 0.8533333333333334
test set loss: 0.40397733449935913
Epoch 12:
0.0008271734841028553
test set accuracy: 0.87
test set loss: 0.3975270986557007
Epoch 13:
0.00047745751406263163
test set accuracy: 0.8566666666666667
test set loss: 0.4006715416908264
Epoch 14:
0.00021613635589349755
test set accuracy: 0.8666666666666667
test set loss: 0.39790403842926025
Epoch 15:
5.463099816548578e-05
test set accuracy: 0.8666666666666667
test set loss: 0.3979119062423706
Train the model without quanvolutional filters.
[ ]:
optimizer = optim.Adam(model_without_qf.parameters(), lr=5e-3, weight_decay=1e-4)
scheduler = CosineAnnealingLR(optimizer, T_max=n_epochs)
for epoch in range(1, n_epochs + 1):
# train
print(f"Epoch {epoch}:")
train(dataflow, model_without_qf, device, optimizer)
print(optimizer.param_groups[0]['lr'])
# valid
accu, loss = valid_test(dataflow, 'test', model_without_qf, device)
accu_list2.append(accu)
loss_list2.append(loss)
scheduler.step()
Epoch 1:
/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py:481: UserWarning: This DataLoader will create 8 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
cpuset_checked))
0.005
test set accuracy: 0.7733333333333333
test set loss: 0.6043258905410767
Epoch 2:
0.004945369001834514
test set accuracy: 0.8166666666666667
test set loss: 0.5571645498275757
Epoch 3:
0.004783863644106502
test set accuracy: 0.8466666666666667
test set loss: 0.46128183603286743
Epoch 4:
0.0045225424859373685
test set accuracy: 0.8366666666666667
test set loss: 0.5158915519714355
Epoch 5:
0.0041728265158971455
test set accuracy: 0.8666666666666667
test set loss: 0.45338067412376404
Epoch 6:
0.00375
test set accuracy: 0.8466666666666667
test set loss: 0.4563254714012146
Epoch 7:
0.0032725424859373687
test set accuracy: 0.8566666666666667
test set loss: 0.4633018374443054
Epoch 8:
0.002761321158169134
test set accuracy: 0.86
test set loss: 0.46147480607032776
Epoch 9:
0.002238678841830867
test set accuracy: 0.85
test set loss: 0.45319321751594543
Epoch 10:
0.001727457514062632
test set accuracy: 0.84
test set loss: 0.46221110224723816
Epoch 11:
0.0012500000000000007
test set accuracy: 0.8533333333333334
test set loss: 0.4611275792121887
Epoch 12:
0.0008271734841028553
test set accuracy: 0.8533333333333334
test set loss: 0.4614029824733734
Epoch 13:
0.00047745751406263163
test set accuracy: 0.8533333333333334
test set loss: 0.4610340893268585
Epoch 14:
0.00021613635589349755
test set accuracy: 0.8533333333333334
test set loss: 0.46056315302848816
Epoch 15:
5.463099816548578e-05
test set accuracy: 0.8533333333333334
test set loss: 0.4606676697731018
Compare Quanvolutional Neural Network with classical model.#
After training, we can plot the accuracy and loss curve. We can see that model with quanvolutional filter can achieve slightly higher accuracy than model without quanvolution.
[ ]:
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
import matplotlib
fig, (ax1, ax2) = plt.subplots(2, 1, sharex=True)
ax1.plot(accu_list1, label="with quanvolution filter")
ax1.plot(accu_list2, label="without quanvolution filter")
ax1.set_ylabel("Accuracy")
ax1.set_ylim([0.6, 1])
ax1.set_xlabel("Epoch")
ax1.legend()
ax2.plot(loss_list1, label="with quanvolution filter")
ax2.plot(loss_list2, label="without quanvolution filter")
ax2.set_ylabel("Loss")
ax2.set_ylim([0, 2])
ax2.set_xlabel("Epoch")
ax2.legend()
plt.tight_layout()
plt.show()
Here we can also see the image before quanvolutional filter and after quanvolutional filter.
[ ]:
import matplotlib.pyplot as plt
import matplotlib
n_samples = 10
n_channels = 4
for feed_dict in dataflow['test']:
inputs = feed_dict['image'].to(device)
break
sample = inputs[:n_samples]
after_quanv = model.qf(sample).view(n_samples, 14*14, 4).cpu().detach().numpy()
fig, axes = plt.subplots(1 + n_channels, n_samples, figsize=(10, 10))
for k in range(n_samples):
axes[0, 0].set_ylabel("image")
if k != 0:
axes[0, k].yaxis.set_visible(False)
norm = matplotlib.colors.Normalize(vmin=0, vmax=1)
axes[0, k].imshow(sample[k, 0, :, :].cpu(), norm=norm, cmap="gray")
for c in range(n_channels):
axes[c + 1, 0].set_ylabel("channel {}".format(c))
if k != 0:
axes[c, k].yaxis.set_visible(False)
axes[c + 1, k].imshow(after_quanv[k, :, c].reshape(14, 14), norm=norm, cmap="gray")
plt.tight_layout()
plt.show()
/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py:481: UserWarning: This DataLoader will create 8 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
cpuset_checked))
Evaluate on real quantum computer.#
At last, we can run our quanvolutional filter on IBMQ’s real quantum machine. The process is really slow so I will not show it here. If you have higher priority access to IBMQ qiskit, you can check the code cell in installation and replace our token with your advance token. That will make the process faster.
[ ]:
# test
valid_test(dataflow, 'test', model, device, qiskit=False)
# run on Qiskit simulator and real Quantum Computers
try:
from qiskit import IBMQ
from torchquantum.plugin import QiskitProcessor
# firstly perform simulate
print(f"\nTest with Qiskit Simulator")
processor_simulation = QiskitProcessor(use_real_qc=False)
model.qf.set_qiskit_processor(processor_simulation)
valid_test(dataflow, 'test', model, device, qiskit=True)
# then try to run on REAL QC
backend_name = 'ibmq_quito'
print(f"\nTest on Real Quantum Computer {backend_name}")
processor_real_qc = QiskitProcessor(use_real_qc=True,backend_name=backend_name)
model.qf.set_qiskit_processor(processor_real_qc)
valid_test(dataflow, 'test', model, device, qiskit=True)
except ImportError:
print("Please install qiskit, create an IBM Q Experience Account and "
"save the account token according to the instruction at "
"'https://github.com/Qiskit/qiskit-ibmq-provider', "
"then try again.")
Trainable Quanvolutional Filter#
In this section, we consider the case that quanvolutional filters are trainable, and we compare various models with nearly the same number of trainale parameters. The four model compared here are described by the following figure.
The Model1 contains a trainable quanvolutional filter and a fully connected layer.
The Model2 contains a trainable quanvolutional filter and a quantum fully connected layer. We use U3CU3Layer0 from torchquantum.layers to implement the QFC layer.
When building the ansatz part of the QFC, we need to pass a dict describing the architecture of the ansatz. Here the dict is {'n_wires': self.n_wires, 'n_blocks': 4, 'n_layers_per_block': 2}, which means the ansatz contains n_wires qubits, there are 4 blocks and in each block are 2 layers. Passing the arch to U3CU3Layer0 we will get a trainable ansatz with 4 blocks and in each block contains 4 U3 gates followed by 4 CU3 gates.
The Model3 is simply a QFC layer.
The Model4 is two fully connected layers.
[ ]:
from torchquantum.encoding import encoder_op_list_name_dict
from torchquantum.layers import U3CU3Layer0
class TrainableQuanvFilter(tq.QuantumModule):
def __init__(self):
super().__init__()
self.n_wires = 4
self.q_device = tq.QuantumDevice(n_wires=self.n_wires)
self.encoder = tq.GeneralEncoder(
[ {'input_idx': [0], 'func': 'ry', 'wires': [0]},
{'input_idx': [1], 'func': 'ry', 'wires': [1]},
{'input_idx': [2], 'func': 'ry', 'wires': [2]},
{'input_idx': [3], 'func': 'ry', 'wires': [3]},])
self.arch = {'n_wires': self.n_wires, 'n_blocks': 5, 'n_layers_per_block': 2}
self.q_layer = U3CU3Layer0(self.arch)
self.measure = tq.MeasureAll(tq.PauliZ)
def forward(self, x, use_qiskit=False):
bsz = x.shape[0]
x = F.avg_pool2d(x, 6).view(bsz, 4, 4)
size = 4
stride = 2
x = x.view(bsz, size, size)
data_list = []
for c in range(0, size, stride):
for r in range(0, size, stride):
data = torch.transpose(torch.cat((x[:, c, r], x[:, c, r+1], x[:, c+1, r], x[:, c+1, r+1])).view(4, bsz), 0, 1)
if use_qiskit:
data = self.qiskit_processor.process_parameterized(
self.q_device, self.encoder, self.q_layer, self.measure, data)
else:
self.encoder(self.q_device, data)
self.q_layer(self.q_device)
data = self.measure(self.q_device)
data_list.append(data.view(bsz, 4))
# transpose to (bsz, channel, 2x2)
result = torch.transpose(torch.cat(data_list, dim=1).view(bsz, 4, 4), 1, 2).float()
return result
class QuantumClassifier(torch.nn.Module):
def __init__(self):
super().__init__()
self.n_wires = 4
self.q_device = tq.QuantumDevice(n_wires=4)
self.encoder = tq.GeneralEncoder(encoder_op_list_name_dict['4x4_ryzxy'])
self.arch = {'n_wires': self.n_wires, 'n_blocks': 8, 'n_layers_per_block': 2}
self.ansatz = U3CU3Layer0(self.arch)
self.measure = tq.MeasureAll(tq.PauliZ)
def forward(self, x, use_qiskit=False):
bsz = x.shape[0]
x = F.avg_pool2d(x, 6).view(bsz, 16)
if use_qiskit:
x = self.qiskit_processor.process_parameterized(
self.q_device, self.encoder, self.q_layer, self.measure, x)
else:
self.encoder(self.q_device, x)
self.ansatz(self.q_device)
x = self.measure(self.q_device)
return x
class QFC(tq.QuantumModule):
def __init__(self):
super().__init__()
self.n_wires = 4
self.q_device = tq.QuantumDevice(n_wires=self.n_wires)
self.encoder = tq.GeneralEncoder(encoder_op_list_name_dict['4x4_ryzxy'])
self.arch = {'n_wires': self.n_wires, 'n_blocks': 4, 'n_layers_per_block': 2}
self.q_layer = U3CU3Layer0(self.arch)
self.measure = tq.MeasureAll(tq.PauliZ)
def forward(self, x, use_qiskit=False):
bsz = x.shape[0]
data = x
if use_qiskit:
data = self.qiskit_processor.process_parameterized(
self.q_device, self.encoder, self.q_layer, self.measure, data)
else:
self.encoder(self.q_device, data)
self.q_layer(self.q_device)
data = self.measure(self.q_device)
return data
class Model1(torch.nn.Module):
def __init__(self):
super().__init__()
self.qf = TrainableQuanvFilter()
self.linear = torch.nn.Linear(16, 4)
def forward(self, x, use_qiskit=False):
x = x.view(-1, 28, 28)
x = self.qf(x)
x = x.reshape(-1, 16)
x = self.linear(x)
return F.log_softmax(x, -1)
class Model2(torch.nn.Module):
def __init__(self):
super().__init__()
self.qf = TrainableQuanvFilter()
self.qfc = QFC()
def forward(self, x, use_qiskit=False):
x = x.view(-1, 28, 28)
x = self.qf(x)
x = x.reshape(-1, 16)
x = self.qfc(x)
return F.log_softmax(x, -1)
class Model3(torch.nn.Module):
def __init__(self):
super().__init__()
self.qfc = QuantumClassifier()
def forward(self, x, use_qiskit=False):
x = self.qfc(x)
return F.log_softmax(x, -1)
class Model4(torch.nn.Module):
def __init__(self):
super().__init__()
self.linear1 = torch.nn.Linear(16, 9)
self.linear2 = torch.nn.Linear(9, 4)
def forward(self, x, use_qiskit=False):
x = x.view(-1, 28, 28)
bsz = x.shape[0]
x = F.avg_pool2d(x, 6).view(bsz, 16)
x = self.linear1(x)
x = self.linear2(x)
return F.log_softmax(x, -1)
Here we do the MNIST 4 classification tasks.
[ ]:
dataset = MNIST(
root='./mnist_data',
train_valid_split_ratio=[0.9, 0.1],
digits_of_interest=[0, 1, 2, 3],
n_test_samples=300,
n_train_samples=500,
)
dataflow = dict()
for split in dataset:
sampler = torch.utils.data.RandomSampler(dataset[split])
dataflow[split] = torch.utils.data.DataLoader(
dataset[split],
batch_size=10,
sampler=sampler,
num_workers=8,
pin_memory=True)
[ ]:
use_cuda = torch.cuda.is_available()
device = torch.device("cuda" if use_cuda else "cpu")
accus = []
model_list = [Model1().to(device), Model2().to(device), Model3().to(device), Model4().to(device)]
for model in model_list:
n_epochs = 15
optimizer = optim.Adam(model.parameters(), lr=5e-3, weight_decay=1e-4)
scheduler = CosineAnnealingLR(optimizer, T_max=n_epochs)
for epoch in range(1, n_epochs + 1):
# train
print(f"Epoch {epoch}:")
train(dataflow, model, device, optimizer)
print(optimizer.param_groups[0]['lr'])
# valid
accu, loss = valid_test(dataflow, 'test', model, device)
scheduler.step()
accus.append(accu)
[ ]:
for i, accu in enumerate(accus):
print('accuracy of model{0}: {1}'.format(i+1, accu))