Training a HyperNetwork on MNIST and FashionMNIST

Package Imports

using Lux, ADTypes, ComponentArrays, LuxCUDA, MLDatasets, MLUtils, OneHotArrays, Optimisers,
      Printf, Random, Setfield, Statistics, Zygote

CUDA.allowscalar(false)

Loading Datasets

function load_dataset(::Type{dset}, n_train::Int, n_eval::Int, batchsize::Int) where {dset}
    imgs, labels = dset(:train)[1:n_train]
    x_train, y_train = reshape(imgs, 28, 28, 1, n_train), onehotbatch(labels, 0:9)

    imgs, labels = dset(:test)[1:n_eval]
    x_test, y_test = reshape(imgs, 28, 28, 1, n_eval), onehotbatch(labels, 0:9)

    return (
        DataLoader((x_train, y_train); batchsize=min(batchsize, n_train), shuffle=true),
        DataLoader((x_test, y_test); batchsize=min(batchsize, n_eval), shuffle=false)
    )
end

function load_datasets(n_train=1024, n_eval=32, batchsize=256)
    return load_dataset.((MNIST, FashionMNIST), n_train, n_eval, batchsize)
end

load_datasets (generic function with 4 methods)

Implement a HyperNet Layer

function HyperNet(
        weight_generator::Lux.AbstractLuxLayer, core_network::Lux.AbstractLuxLayer)
    ca_axes = Lux.initialparameters(Random.default_rng(), core_network) |>
              ComponentArray |>
              getaxes
    return @compact(; ca_axes, weight_generator, core_network, dispatch=:HyperNet) do (x, y)
        # Generate the weights
        ps_new = ComponentArray(vec(weight_generator(x)), ca_axes)
        @return core_network(y, ps_new)
    end
end

HyperNet (generic function with 1 method)

Defining functions on the CompactLuxLayer requires some understanding of how the layer is structured, as such we don't recommend doing it unless you are familiar with the internals. In this case, we simply write it to ignore the initialization of the core_network parameters.

function Lux.initialparameters(rng::AbstractRNG, hn::CompactLuxLayer{:HyperNet})
    return (; weight_generator=Lux.initialparameters(rng, hn.layers.weight_generator),)
end

Create and Initialize the HyperNet

function create_model()
    # Doesn't need to be a MLP can have any Lux Layer
    core_network = Chain(FlattenLayer(), Dense(784, 256, relu), Dense(256, 10))
    weight_generator = Chain(
        Embedding(2 => 32),
        Dense(32, 64, relu),
        Dense(64, Lux.parameterlength(core_network))
    )

    model = HyperNet(weight_generator, core_network)
    return model
end

create_model (generic function with 1 method)

Define Utility Functions

const loss = CrossEntropyLoss(; logits=Val(true))

function accuracy(model, ps, st, dataloader, data_idx)
    total_correct, total = 0, 0
    st = Lux.testmode(st)
    for (x, y) in dataloader
        target_class = onecold(y)
        predicted_class = onecold(first(model((data_idx, x), ps, st)))
        total_correct += sum(target_class .== predicted_class)
        total += length(target_class)
    end
    return total_correct / total
end

accuracy (generic function with 1 method)

Training

function train()
    model = create_model()
    dataloaders = load_datasets()

    dev = gpu_device()
    rng = Xoshiro(0)
    ps, st = Lux.setup(rng, model) |> dev

    train_state = Training.TrainState(model, ps, st, Adam(0.001f0))

    ### Lets train the model
    nepochs = 50
    for epoch in 1:nepochs, data_idx in 1:2
        train_dataloader, test_dataloader = dataloaders[data_idx] .|> dev

        stime = time()
        for (x, y) in train_dataloader
            (_, _, _, train_state) = Training.single_train_step!(
                AutoZygote(), loss, ((data_idx, x), y), train_state)
        end
        ttime = time() - stime

        train_acc = round(
            accuracy(model, train_state.parameters,
                train_state.states, train_dataloader, data_idx) * 100;
            digits=2)
        test_acc = round(
            accuracy(model, train_state.parameters,
                train_state.states, test_dataloader, data_idx) * 100;
            digits=2)

        data_name = data_idx == 1 ? "MNIST" : "FashionMNIST"

        @printf "[%3d/%3d]\t%12s\tTime %3.5fs\tTraining Accuracy: %3.2f%%\tTest \
                 Accuracy: %3.2f%%\n" epoch nepochs data_name ttime train_acc test_acc
    end

    println()

    test_acc_list = [0.0, 0.0]
    for data_idx in 1:2
        train_dataloader, test_dataloader = dataloaders[data_idx] .|> dev
        train_acc = round(
            accuracy(model, train_state.parameters,
                train_state.states, train_dataloader, data_idx) * 100;
            digits=2)
        test_acc = round(
            accuracy(model, train_state.parameters,
                train_state.states, test_dataloader, data_idx) * 100;
            digits=2)

        data_name = data_idx == 1 ? "MNIST" : "FashionMNIST"

        @printf "[FINAL]\t%12s\tTraining Accuracy: %3.2f%%\tTest Accuracy: \
                 %3.2f%%\n" data_name train_acc test_acc
        test_acc_list[data_idx] = test_acc
    end
    return test_acc_list
end

test_acc_list = train()

[  1/ 50]	       MNIST	Time 71.28585s	Training Accuracy: 56.25%	Test Accuracy: 50.00%
[  1/ 50]	FashionMNIST	Time 0.02745s	Training Accuracy: 42.38%	Test Accuracy: 37.50%
[  2/ 50]	       MNIST	Time 0.02940s	Training Accuracy: 64.75%	Test Accuracy: 65.62%
[  2/ 50]	FashionMNIST	Time 0.02847s	Training Accuracy: 60.35%	Test Accuracy: 50.00%
[  3/ 50]	       MNIST	Time 0.02808s	Training Accuracy: 77.25%	Test Accuracy: 71.88%
[  3/ 50]	FashionMNIST	Time 0.02793s	Training Accuracy: 63.38%	Test Accuracy: 46.88%
[  4/ 50]	       MNIST	Time 0.02838s	Training Accuracy: 75.29%	Test Accuracy: 62.50%
[  4/ 50]	FashionMNIST	Time 0.02797s	Training Accuracy: 68.26%	Test Accuracy: 68.75%
[  5/ 50]	       MNIST	Time 0.02280s	Training Accuracy: 82.52%	Test Accuracy: 68.75%
[  5/ 50]	FashionMNIST	Time 0.03865s	Training Accuracy: 73.44%	Test Accuracy: 68.75%
[  6/ 50]	       MNIST	Time 0.02032s	Training Accuracy: 85.64%	Test Accuracy: 62.50%
[  6/ 50]	FashionMNIST	Time 0.02057s	Training Accuracy: 70.70%	Test Accuracy: 62.50%
[  7/ 50]	       MNIST	Time 0.02057s	Training Accuracy: 89.06%	Test Accuracy: 71.88%
[  7/ 50]	FashionMNIST	Time 0.02106s	Training Accuracy: 77.64%	Test Accuracy: 59.38%
[  8/ 50]	       MNIST	Time 0.02057s	Training Accuracy: 91.60%	Test Accuracy: 78.12%
[  8/ 50]	FashionMNIST	Time 0.02087s	Training Accuracy: 79.79%	Test Accuracy: 65.62%
[  9/ 50]	       MNIST	Time 0.02086s	Training Accuracy: 93.65%	Test Accuracy: 78.12%
[  9/ 50]	FashionMNIST	Time 0.02127s	Training Accuracy: 77.25%	Test Accuracy: 65.62%
[ 10/ 50]	       MNIST	Time 0.03372s	Training Accuracy: 95.21%	Test Accuracy: 78.12%
[ 10/ 50]	FashionMNIST	Time 0.02057s	Training Accuracy: 79.69%	Test Accuracy: 62.50%
[ 11/ 50]	       MNIST	Time 0.02018s	Training Accuracy: 96.68%	Test Accuracy: 81.25%
[ 11/ 50]	FashionMNIST	Time 0.02033s	Training Accuracy: 80.27%	Test Accuracy: 68.75%
[ 12/ 50]	       MNIST	Time 0.02062s	Training Accuracy: 97.46%	Test Accuracy: 84.38%
[ 12/ 50]	FashionMNIST	Time 0.02032s	Training Accuracy: 82.91%	Test Accuracy: 78.12%
[ 13/ 50]	       MNIST	Time 0.02121s	Training Accuracy: 98.05%	Test Accuracy: 84.38%
[ 13/ 50]	FashionMNIST	Time 0.02076s	Training Accuracy: 84.38%	Test Accuracy: 71.88%
[ 14/ 50]	       MNIST	Time 0.02052s	Training Accuracy: 98.54%	Test Accuracy: 78.12%
[ 14/ 50]	FashionMNIST	Time 0.02815s	Training Accuracy: 82.23%	Test Accuracy: 65.62%
[ 15/ 50]	       MNIST	Time 0.02053s	Training Accuracy: 99.22%	Test Accuracy: 78.12%
[ 15/ 50]	FashionMNIST	Time 0.02096s	Training Accuracy: 85.45%	Test Accuracy: 75.00%
[ 16/ 50]	       MNIST	Time 0.02124s	Training Accuracy: 99.32%	Test Accuracy: 78.12%
[ 16/ 50]	FashionMNIST	Time 0.02123s	Training Accuracy: 85.64%	Test Accuracy: 68.75%
[ 17/ 50]	       MNIST	Time 0.02110s	Training Accuracy: 99.71%	Test Accuracy: 81.25%
[ 17/ 50]	FashionMNIST	Time 0.02048s	Training Accuracy: 88.18%	Test Accuracy: 78.12%
[ 18/ 50]	       MNIST	Time 0.02095s	Training Accuracy: 99.71%	Test Accuracy: 81.25%
[ 18/ 50]	FashionMNIST	Time 0.02082s	Training Accuracy: 87.40%	Test Accuracy: 68.75%
[ 19/ 50]	       MNIST	Time 0.02046s	Training Accuracy: 99.71%	Test Accuracy: 81.25%
[ 19/ 50]	FashionMNIST	Time 0.02108s	Training Accuracy: 88.38%	Test Accuracy: 71.88%
[ 20/ 50]	       MNIST	Time 0.02132s	Training Accuracy: 99.71%	Test Accuracy: 78.12%
[ 20/ 50]	FashionMNIST	Time 0.02190s	Training Accuracy: 89.36%	Test Accuracy: 75.00%
[ 21/ 50]	       MNIST	Time 0.02137s	Training Accuracy: 99.71%	Test Accuracy: 81.25%
[ 21/ 50]	FashionMNIST	Time 0.02109s	Training Accuracy: 89.94%	Test Accuracy: 78.12%
[ 22/ 50]	       MNIST	Time 0.02105s	Training Accuracy: 99.80%	Test Accuracy: 78.12%
[ 22/ 50]	FashionMNIST	Time 0.02042s	Training Accuracy: 90.33%	Test Accuracy: 81.25%
[ 23/ 50]	       MNIST	Time 0.02818s	Training Accuracy: 99.90%	Test Accuracy: 84.38%
[ 23/ 50]	FashionMNIST	Time 0.02114s	Training Accuracy: 89.16%	Test Accuracy: 75.00%
[ 24/ 50]	       MNIST	Time 0.02103s	Training Accuracy: 99.90%	Test Accuracy: 78.12%
[ 24/ 50]	FashionMNIST	Time 0.02048s	Training Accuracy: 89.45%	Test Accuracy: 78.12%
[ 25/ 50]	       MNIST	Time 0.02087s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 25/ 50]	FashionMNIST	Time 0.02118s	Training Accuracy: 89.84%	Test Accuracy: 75.00%
[ 26/ 50]	       MNIST	Time 0.02108s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 26/ 50]	FashionMNIST	Time 0.02136s	Training Accuracy: 91.60%	Test Accuracy: 81.25%
[ 27/ 50]	       MNIST	Time 0.02069s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 27/ 50]	FashionMNIST	Time 0.02639s	Training Accuracy: 91.89%	Test Accuracy: 81.25%
[ 28/ 50]	       MNIST	Time 0.02189s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 28/ 50]	FashionMNIST	Time 0.02148s	Training Accuracy: 92.58%	Test Accuracy: 81.25%
[ 29/ 50]	       MNIST	Time 0.02182s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 29/ 50]	FashionMNIST	Time 0.02118s	Training Accuracy: 93.16%	Test Accuracy: 81.25%
[ 30/ 50]	       MNIST	Time 0.02180s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 30/ 50]	FashionMNIST	Time 0.02140s	Training Accuracy: 93.16%	Test Accuracy: 81.25%
[ 31/ 50]	       MNIST	Time 0.02136s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 31/ 50]	FashionMNIST	Time 0.02178s	Training Accuracy: 93.65%	Test Accuracy: 81.25%
[ 32/ 50]	       MNIST	Time 0.02749s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 32/ 50]	FashionMNIST	Time 0.02114s	Training Accuracy: 93.46%	Test Accuracy: 78.12%
[ 33/ 50]	       MNIST	Time 0.02157s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 33/ 50]	FashionMNIST	Time 0.02100s	Training Accuracy: 93.65%	Test Accuracy: 81.25%
[ 34/ 50]	       MNIST	Time 0.02109s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 34/ 50]	FashionMNIST	Time 0.02125s	Training Accuracy: 93.75%	Test Accuracy: 78.12%
[ 35/ 50]	       MNIST	Time 0.02066s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 35/ 50]	FashionMNIST	Time 0.02084s	Training Accuracy: 94.14%	Test Accuracy: 81.25%
[ 36/ 50]	       MNIST	Time 0.02098s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 36/ 50]	FashionMNIST	Time 0.02065s	Training Accuracy: 95.02%	Test Accuracy: 81.25%
[ 37/ 50]	       MNIST	Time 0.02113s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 37/ 50]	FashionMNIST	Time 0.02114s	Training Accuracy: 94.73%	Test Accuracy: 81.25%
[ 38/ 50]	       MNIST	Time 0.02108s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 38/ 50]	FashionMNIST	Time 0.02109s	Training Accuracy: 94.73%	Test Accuracy: 81.25%
[ 39/ 50]	       MNIST	Time 0.02079s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 39/ 50]	FashionMNIST	Time 0.02103s	Training Accuracy: 95.21%	Test Accuracy: 81.25%
[ 40/ 50]	       MNIST	Time 0.02067s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 40/ 50]	FashionMNIST	Time 0.02974s	Training Accuracy: 94.92%	Test Accuracy: 81.25%
[ 41/ 50]	       MNIST	Time 0.02071s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 41/ 50]	FashionMNIST	Time 0.02085s	Training Accuracy: 94.04%	Test Accuracy: 75.00%
[ 42/ 50]	       MNIST	Time 0.02151s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 42/ 50]	FashionMNIST	Time 0.02107s	Training Accuracy: 94.24%	Test Accuracy: 78.12%
[ 43/ 50]	       MNIST	Time 0.02127s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 43/ 50]	FashionMNIST	Time 0.02123s	Training Accuracy: 93.55%	Test Accuracy: 75.00%
[ 44/ 50]	       MNIST	Time 0.02085s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 44/ 50]	FashionMNIST	Time 0.02067s	Training Accuracy: 95.02%	Test Accuracy: 78.12%
[ 45/ 50]	       MNIST	Time 0.03514s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 45/ 50]	FashionMNIST	Time 0.02249s	Training Accuracy: 96.09%	Test Accuracy: 81.25%
[ 46/ 50]	       MNIST	Time 0.02139s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 46/ 50]	FashionMNIST	Time 0.02109s	Training Accuracy: 95.80%	Test Accuracy: 81.25%
[ 47/ 50]	       MNIST	Time 0.02168s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 47/ 50]	FashionMNIST	Time 0.02099s	Training Accuracy: 96.78%	Test Accuracy: 81.25%
[ 48/ 50]	       MNIST	Time 0.02119s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 48/ 50]	FashionMNIST	Time 0.02053s	Training Accuracy: 96.19%	Test Accuracy: 81.25%
[ 49/ 50]	       MNIST	Time 0.02296s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 49/ 50]	FashionMNIST	Time 0.02727s	Training Accuracy: 96.88%	Test Accuracy: 81.25%
[ 50/ 50]	       MNIST	Time 0.02168s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 50/ 50]	FashionMNIST	Time 0.02130s	Training Accuracy: 96.88%	Test Accuracy: 81.25%

[FINAL]	       MNIST	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[FINAL]	FashionMNIST	Training Accuracy: 96.88%	Test Accuracy: 81.25%

Appendix

using InteractiveUtils
InteractiveUtils.versioninfo()

if @isdefined(MLDataDevices)
    if @isdefined(CUDA) && MLDataDevices.functional(CUDADevice)
        println()
        CUDA.versioninfo()
    end

    if @isdefined(AMDGPU) && MLDataDevices.functional(AMDGPUDevice)
        println()
        AMDGPU.versioninfo()
    end
end

Julia Version 1.10.6
Commit 67dffc4a8ae (2024-10-28 12:23 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 48 × AMD EPYC 7402 24-Core Processor
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-15.0.7 (ORCJIT, znver2)
Threads: 48 default, 0 interactive, 24 GC (on 2 virtual cores)
Environment:
  JULIA_CPU_THREADS = 2
  JULIA_DEPOT_PATH = /root/.cache/julia-buildkite-plugin/depots/01872db4-8c79-43af-ab7d-12abac4f24f6
  LD_LIBRARY_PATH = /usr/local/nvidia/lib:/usr/local/nvidia/lib64
  JULIA_PKG_SERVER = 
  JULIA_NUM_THREADS = 48
  JULIA_CUDA_HARD_MEMORY_LIMIT = 100%
  JULIA_PKG_PRECOMPILE_AUTO = 0
  JULIA_DEBUG = Literate

CUDA runtime 12.6, artifact installation
CUDA driver 12.6
NVIDIA driver 560.35.3

CUDA libraries: 
- CUBLAS: 12.6.3
- CURAND: 10.3.7
- CUFFT: 11.3.0
- CUSOLVER: 11.7.1
- CUSPARSE: 12.5.4
- CUPTI: 2024.3.2 (API 24.0.0)
- NVML: 12.0.0+560.35.3

Julia packages: 
- CUDA: 5.5.2
- CUDA_Driver_jll: 0.10.3+0
- CUDA_Runtime_jll: 0.15.4+0

Toolchain:
- Julia: 1.10.6
- LLVM: 15.0.7

Environment:
- JULIA_CUDA_HARD_MEMORY_LIMIT: 100%

1 device:
  0: NVIDIA A100-PCIE-40GB MIG 1g.5gb (sm_80, 2.170 GiB / 4.750 GiB available)

---

This page was generated using Literate.jl.