search

mit-han-lab / torchquantum

A PyTorch-based framework for Quantum Classical Simulation, Quantum Machine Learning, Quantum Neural Networks, Parameterized Quantum Circuits with support for easy deployments on real quantum computers.
https://torchquantum.org
MIT License
1.3k stars 196 forks source link

Train data label and image are different #40

Open manu123416 opened 1 year ago

manu123416 commented 1 year ago

Hi , I tried testing your quantum neural network code on jupyter notebook. I think, there is some bug in the training data.

dataset = MNIST(root='../Data_Manu',
                 train_valid_split_ratio=[0.9, 0.1],
            digits_of_interest=[3, 5],
#             n_train_samples = 75,
            n_test_samples=75)

data_train= dataset['train'][0]

{'image': tensor([[[-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.3733, -0.1696,
           -0.1696,  0.8868,  1.4468,  2.1087,  0.0213, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242,  0.1995,  0.6704,  1.8032,  1.9560,  2.7960,
            2.8088,  2.7960,  2.7960,  2.7960,  2.4142, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
            1.6250,  2.1723,  1.3068,  1.3068,  1.3196,  1.3068,  1.3068,
            1.4978,  2.5542,  2.7069,  2.7960,  2.7960,  2.7960,  2.7960,
            2.8088,  2.7960,  2.7960,  2.3251,  1.8160, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
            2.2996,  2.7960,  2.7960,  2.7960,  2.8088,  2.7960,  2.7960,
            2.7960,  2.7960,  2.8088,  2.7960,  2.7960,  2.7960,  2.7960,
            2.0323,  0.9886,  0.3140, -0.3606, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
            0.7977,  2.8088,  2.8088,  2.8088,  2.8215,  2.8088,  2.8088,
            2.8088,  2.8088,  2.6433,  1.9560,  0.8232,  0.6322, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
            0.5049,  2.7960,  2.7960,  2.7960,  2.8088,  2.4778,  1.2941,
            0.6322,  0.0722, -0.0424, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,  0.4795,
            2.6433,  2.7960,  2.7960,  2.7960,  1.0523, -0.2715, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,  2.2360,
            2.7960,  2.7960,  2.5160,  0.5813, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,  2.8088,
            2.7960,  2.7960,  2.4906,  0.8232,  0.8359,  0.8232,  0.8232,
            0.8232, -0.1315, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,  2.0451,
            2.8088,  2.8088,  2.8088,  2.8088,  2.8215,  2.8088,  2.8088,
            2.8088,  2.8088,  2.0451,  0.1740, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,  0.1740,
            1.8796,  2.5415,  2.5415,  2.6433,  2.5542,  2.5415,  2.5415,
            2.5415,  2.6433,  2.8088,  2.3760, -0.2460, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.0424, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.0424,  2.4142,  2.7960,  2.1087, -0.3478, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.2206,  2.2360,  2.7960,  2.7960, -0.1824, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.0296,
            1.4850,  2.3378,  2.8088,  2.7960,  2.7960,  0.3013, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242,  0.6195,  1.6505,  2.8088,
            2.8088,  2.8088,  2.8215,  2.8088,  1.9051, -0.3224, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.3351,  0.5049,  2.0196,  2.8088,  2.7960,  2.7960,
            2.7960,  2.7960,  2.4524,  1.2050, -0.2715, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.3351,  1.7269,  2.7960,  2.7960,  2.8088,  2.7960,  2.7960,
            2.6306,  1.7905,  0.3395, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.3097,  1.4341,  2.2869,  2.2869,  2.2996,  1.7141,  1.0650,
           -0.1315, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242],
          [-0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242,
           -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242, -0.4242]]]),
 'digit': 1}

The tensor matrix contains 5 but the label shows 1?
Hanrui-Wang commented 1 year ago

Hi manu123416,

As shown here: https://github.com/mit-han-lab/torchquantum/blob/772f0f78a7db50c9f09c866f5bc9e20fed7066c8/torchquantum/datasets/mnist.py#L194 the real digits will be converted to relative digit. This will make it more convenient for cross entropy loss training

Hanrui-Wang commented 1 year ago

For more data, could you try change the n_train_samples = 75 and see whether more data are used in training set?

manu123416 commented 1 year ago

Thank you Hanui-Wang for response. I tested it by putting n_train_samples = 75, but it doesnot increase the data for training and reduces the training data to 1.