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 69.48855s	Training Accuracy: 53.03%	Test Accuracy: 43.75%
[  1/ 50]	FashionMNIST	Time 0.02735s	Training Accuracy: 46.19%	Test Accuracy: 40.62%
[  2/ 50]	       MNIST	Time 0.02876s	Training Accuracy: 64.16%	Test Accuracy: 62.50%
[  2/ 50]	FashionMNIST	Time 0.02771s	Training Accuracy: 55.37%	Test Accuracy: 50.00%
[  3/ 50]	       MNIST	Time 0.02763s	Training Accuracy: 70.70%	Test Accuracy: 68.75%
[  3/ 50]	FashionMNIST	Time 0.02792s	Training Accuracy: 65.04%	Test Accuracy: 59.38%
[  4/ 50]	       MNIST	Time 0.02800s	Training Accuracy: 76.56%	Test Accuracy: 65.62%
[  4/ 50]	FashionMNIST	Time 0.02205s	Training Accuracy: 59.77%	Test Accuracy: 53.12%
[  5/ 50]	       MNIST	Time 0.05418s	Training Accuracy: 79.39%	Test Accuracy: 59.38%
[  5/ 50]	FashionMNIST	Time 0.02052s	Training Accuracy: 66.02%	Test Accuracy: 62.50%
[  6/ 50]	       MNIST	Time 0.02051s	Training Accuracy: 83.98%	Test Accuracy: 65.62%
[  6/ 50]	FashionMNIST	Time 0.02000s	Training Accuracy: 73.34%	Test Accuracy: 62.50%
[  7/ 50]	       MNIST	Time 0.02019s	Training Accuracy: 87.50%	Test Accuracy: 78.12%
[  7/ 50]	FashionMNIST	Time 0.02047s	Training Accuracy: 75.98%	Test Accuracy: 68.75%
[  8/ 50]	       MNIST	Time 0.02123s	Training Accuracy: 89.94%	Test Accuracy: 75.00%
[  8/ 50]	FashionMNIST	Time 0.02050s	Training Accuracy: 78.22%	Test Accuracy: 71.88%
[  9/ 50]	       MNIST	Time 0.02225s	Training Accuracy: 91.21%	Test Accuracy: 75.00%
[  9/ 50]	FashionMNIST	Time 0.03643s	Training Accuracy: 78.32%	Test Accuracy: 68.75%
[ 10/ 50]	       MNIST	Time 0.02161s	Training Accuracy: 92.97%	Test Accuracy: 78.12%
[ 10/ 50]	FashionMNIST	Time 0.02192s	Training Accuracy: 79.30%	Test Accuracy: 75.00%
[ 11/ 50]	       MNIST	Time 0.02128s	Training Accuracy: 95.51%	Test Accuracy: 78.12%
[ 11/ 50]	FashionMNIST	Time 0.02116s	Training Accuracy: 79.79%	Test Accuracy: 68.75%
[ 12/ 50]	       MNIST	Time 0.02093s	Training Accuracy: 96.68%	Test Accuracy: 78.12%
[ 12/ 50]	FashionMNIST	Time 0.02049s	Training Accuracy: 81.54%	Test Accuracy: 71.88%
[ 13/ 50]	       MNIST	Time 0.02077s	Training Accuracy: 97.56%	Test Accuracy: 78.12%
[ 13/ 50]	FashionMNIST	Time 0.02088s	Training Accuracy: 84.08%	Test Accuracy: 75.00%
[ 14/ 50]	       MNIST	Time 0.01966s	Training Accuracy: 98.34%	Test Accuracy: 84.38%
[ 14/ 50]	FashionMNIST	Time 0.02043s	Training Accuracy: 83.69%	Test Accuracy: 71.88%
[ 15/ 50]	       MNIST	Time 0.02064s	Training Accuracy: 98.93%	Test Accuracy: 84.38%
[ 15/ 50]	FashionMNIST	Time 0.02044s	Training Accuracy: 84.28%	Test Accuracy: 75.00%
[ 16/ 50]	       MNIST	Time 0.02043s	Training Accuracy: 99.51%	Test Accuracy: 84.38%
[ 16/ 50]	FashionMNIST	Time 0.02020s	Training Accuracy: 83.79%	Test Accuracy: 71.88%
[ 17/ 50]	       MNIST	Time 0.02046s	Training Accuracy: 99.80%	Test Accuracy: 84.38%
[ 17/ 50]	FashionMNIST	Time 0.02075s	Training Accuracy: 85.64%	Test Accuracy: 78.12%
[ 18/ 50]	       MNIST	Time 0.02761s	Training Accuracy: 99.80%	Test Accuracy: 84.38%
[ 18/ 50]	FashionMNIST	Time 0.02039s	Training Accuracy: 87.79%	Test Accuracy: 71.88%
[ 19/ 50]	       MNIST	Time 0.02000s	Training Accuracy: 99.80%	Test Accuracy: 84.38%
[ 19/ 50]	FashionMNIST	Time 0.02043s	Training Accuracy: 88.87%	Test Accuracy: 71.88%
[ 20/ 50]	       MNIST	Time 0.02062s	Training Accuracy: 99.90%	Test Accuracy: 84.38%
[ 20/ 50]	FashionMNIST	Time 0.02044s	Training Accuracy: 88.57%	Test Accuracy: 78.12%
[ 21/ 50]	       MNIST	Time 0.02068s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 21/ 50]	FashionMNIST	Time 0.02089s	Training Accuracy: 89.45%	Test Accuracy: 75.00%
[ 22/ 50]	       MNIST	Time 0.02111s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 22/ 50]	FashionMNIST	Time 0.02686s	Training Accuracy: 89.94%	Test Accuracy: 75.00%
[ 23/ 50]	       MNIST	Time 0.02091s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 23/ 50]	FashionMNIST	Time 0.02061s	Training Accuracy: 90.23%	Test Accuracy: 75.00%
[ 24/ 50]	       MNIST	Time 0.02035s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 24/ 50]	FashionMNIST	Time 0.02028s	Training Accuracy: 90.62%	Test Accuracy: 75.00%
[ 25/ 50]	       MNIST	Time 0.02072s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 25/ 50]	FashionMNIST	Time 0.02099s	Training Accuracy: 91.89%	Test Accuracy: 75.00%
[ 26/ 50]	       MNIST	Time 0.02096s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 26/ 50]	FashionMNIST	Time 0.02050s	Training Accuracy: 92.29%	Test Accuracy: 75.00%
[ 27/ 50]	       MNIST	Time 0.02667s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 27/ 50]	FashionMNIST	Time 0.02048s	Training Accuracy: 92.48%	Test Accuracy: 75.00%
[ 28/ 50]	       MNIST	Time 0.02106s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 28/ 50]	FashionMNIST	Time 0.02163s	Training Accuracy: 92.97%	Test Accuracy: 75.00%
[ 29/ 50]	       MNIST	Time 0.02089s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 29/ 50]	FashionMNIST	Time 0.02078s	Training Accuracy: 92.87%	Test Accuracy: 75.00%
[ 30/ 50]	       MNIST	Time 0.02094s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 30/ 50]	FashionMNIST	Time 0.02115s	Training Accuracy: 93.46%	Test Accuracy: 75.00%
[ 31/ 50]	       MNIST	Time 0.02098s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 31/ 50]	FashionMNIST	Time 0.01995s	Training Accuracy: 93.46%	Test Accuracy: 75.00%
[ 32/ 50]	       MNIST	Time 0.02116s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 32/ 50]	FashionMNIST	Time 0.02099s	Training Accuracy: 93.55%	Test Accuracy: 75.00%
[ 33/ 50]	       MNIST	Time 0.02383s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 33/ 50]	FashionMNIST	Time 0.02110s	Training Accuracy: 93.46%	Test Accuracy: 71.88%
[ 34/ 50]	       MNIST	Time 0.02138s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 34/ 50]	FashionMNIST	Time 0.02148s	Training Accuracy: 93.16%	Test Accuracy: 78.12%
[ 35/ 50]	       MNIST	Time 0.02164s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 35/ 50]	FashionMNIST	Time 0.03411s	Training Accuracy: 91.21%	Test Accuracy: 75.00%
[ 36/ 50]	       MNIST	Time 0.02090s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 36/ 50]	FashionMNIST	Time 0.02145s	Training Accuracy: 93.65%	Test Accuracy: 75.00%
[ 37/ 50]	       MNIST	Time 0.02162s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 37/ 50]	FashionMNIST	Time 0.02253s	Training Accuracy: 94.53%	Test Accuracy: 75.00%
[ 38/ 50]	       MNIST	Time 0.02072s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 38/ 50]	FashionMNIST	Time 0.02149s	Training Accuracy: 94.73%	Test Accuracy: 75.00%
[ 39/ 50]	       MNIST	Time 0.02165s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 39/ 50]	FashionMNIST	Time 0.02153s	Training Accuracy: 95.21%	Test Accuracy: 75.00%
[ 40/ 50]	       MNIST	Time 0.02693s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 40/ 50]	FashionMNIST	Time 0.02076s	Training Accuracy: 95.31%	Test Accuracy: 75.00%
[ 41/ 50]	       MNIST	Time 0.02114s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 41/ 50]	FashionMNIST	Time 0.02092s	Training Accuracy: 95.90%	Test Accuracy: 75.00%
[ 42/ 50]	       MNIST	Time 0.02075s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 42/ 50]	FashionMNIST	Time 0.02058s	Training Accuracy: 95.51%	Test Accuracy: 75.00%
[ 43/ 50]	       MNIST	Time 0.02102s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 43/ 50]	FashionMNIST	Time 0.02080s	Training Accuracy: 95.70%	Test Accuracy: 75.00%
[ 44/ 50]	       MNIST	Time 0.02096s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 44/ 50]	FashionMNIST	Time 0.02750s	Training Accuracy: 96.00%	Test Accuracy: 75.00%
[ 45/ 50]	       MNIST	Time 0.02079s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 45/ 50]	FashionMNIST	Time 0.02078s	Training Accuracy: 95.12%	Test Accuracy: 78.12%
[ 46/ 50]	       MNIST	Time 0.02047s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 46/ 50]	FashionMNIST	Time 0.02077s	Training Accuracy: 94.14%	Test Accuracy: 75.00%
[ 47/ 50]	       MNIST	Time 0.02090s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 47/ 50]	FashionMNIST	Time 0.02249s	Training Accuracy: 95.41%	Test Accuracy: 75.00%
[ 48/ 50]	       MNIST	Time 0.02217s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 48/ 50]	FashionMNIST	Time 0.02085s	Training Accuracy: 95.61%	Test Accuracy: 75.00%
[ 49/ 50]	       MNIST	Time 0.02025s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 49/ 50]	FashionMNIST	Time 0.02117s	Training Accuracy: 96.39%	Test Accuracy: 75.00%
[ 50/ 50]	       MNIST	Time 0.02096s	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[ 50/ 50]	FashionMNIST	Time 0.02397s	Training Accuracy: 96.39%	Test Accuracy: 75.00%

[FINAL]	       MNIST	Training Accuracy: 100.00%	Test Accuracy: 84.38%
[FINAL]	FashionMNIST	Training Accuracy: 96.39%	Test Accuracy: 75.00%

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.