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 72.33321s	Training Accuracy: 57.91%	Test Accuracy: 53.12%
[  1/ 50]	FashionMNIST	Time 0.02845s	Training Accuracy: 48.63%	Test Accuracy: 43.75%
[  2/ 50]	       MNIST	Time 0.02912s	Training Accuracy: 69.43%	Test Accuracy: 65.62%
[  2/ 50]	FashionMNIST	Time 0.02984s	Training Accuracy: 63.57%	Test Accuracy: 56.25%
[  3/ 50]	       MNIST	Time 0.02899s	Training Accuracy: 77.73%	Test Accuracy: 68.75%
[  3/ 50]	FashionMNIST	Time 0.02873s	Training Accuracy: 64.94%	Test Accuracy: 62.50%
[  4/ 50]	       MNIST	Time 0.02721s	Training Accuracy: 82.71%	Test Accuracy: 65.62%
[  4/ 50]	FashionMNIST	Time 0.02093s	Training Accuracy: 65.82%	Test Accuracy: 59.38%
[  5/ 50]	       MNIST	Time 0.02106s	Training Accuracy: 85.84%	Test Accuracy: 68.75%
[  5/ 50]	FashionMNIST	Time 0.02239s	Training Accuracy: 71.68%	Test Accuracy: 53.12%
[  6/ 50]	       MNIST	Time 0.02133s	Training Accuracy: 86.52%	Test Accuracy: 71.88%
[  6/ 50]	FashionMNIST	Time 0.02146s	Training Accuracy: 76.46%	Test Accuracy: 53.12%
[  7/ 50]	       MNIST	Time 0.02169s	Training Accuracy: 89.36%	Test Accuracy: 78.12%
[  7/ 50]	FashionMNIST	Time 0.02320s	Training Accuracy: 77.34%	Test Accuracy: 62.50%
[  8/ 50]	       MNIST	Time 0.02666s	Training Accuracy: 92.58%	Test Accuracy: 81.25%
[  8/ 50]	FashionMNIST	Time 0.05175s	Training Accuracy: 78.61%	Test Accuracy: 65.62%
[  9/ 50]	       MNIST	Time 0.02292s	Training Accuracy: 94.14%	Test Accuracy: 78.12%
[  9/ 50]	FashionMNIST	Time 0.02085s	Training Accuracy: 77.44%	Test Accuracy: 62.50%
[ 10/ 50]	       MNIST	Time 0.02117s	Training Accuracy: 95.70%	Test Accuracy: 78.12%
[ 10/ 50]	FashionMNIST	Time 0.02135s	Training Accuracy: 81.84%	Test Accuracy: 68.75%
[ 11/ 50]	       MNIST	Time 0.02182s	Training Accuracy: 96.39%	Test Accuracy: 81.25%
[ 11/ 50]	FashionMNIST	Time 0.02111s	Training Accuracy: 81.84%	Test Accuracy: 65.62%
[ 12/ 50]	       MNIST	Time 0.02164s	Training Accuracy: 97.75%	Test Accuracy: 81.25%
[ 12/ 50]	FashionMNIST	Time 0.02159s	Training Accuracy: 82.71%	Test Accuracy: 68.75%
[ 13/ 50]	       MNIST	Time 0.02879s	Training Accuracy: 97.85%	Test Accuracy: 78.12%
[ 13/ 50]	FashionMNIST	Time 0.02271s	Training Accuracy: 84.28%	Test Accuracy: 65.62%
[ 14/ 50]	       MNIST	Time 0.02079s	Training Accuracy: 98.73%	Test Accuracy: 78.12%
[ 14/ 50]	FashionMNIST	Time 0.02091s	Training Accuracy: 83.11%	Test Accuracy: 62.50%
[ 15/ 50]	       MNIST	Time 0.02113s	Training Accuracy: 99.22%	Test Accuracy: 78.12%
[ 15/ 50]	FashionMNIST	Time 0.02264s	Training Accuracy: 83.79%	Test Accuracy: 65.62%
[ 16/ 50]	       MNIST	Time 0.02123s	Training Accuracy: 99.32%	Test Accuracy: 78.12%
[ 16/ 50]	FashionMNIST	Time 0.02098s	Training Accuracy: 84.96%	Test Accuracy: 65.62%
[ 17/ 50]	       MNIST	Time 0.02143s	Training Accuracy: 99.51%	Test Accuracy: 78.12%
[ 17/ 50]	FashionMNIST	Time 0.02652s	Training Accuracy: 85.64%	Test Accuracy: 68.75%
[ 18/ 50]	       MNIST	Time 0.02070s	Training Accuracy: 99.71%	Test Accuracy: 78.12%
[ 18/ 50]	FashionMNIST	Time 0.02071s	Training Accuracy: 85.55%	Test Accuracy: 65.62%
[ 19/ 50]	       MNIST	Time 0.02077s	Training Accuracy: 99.90%	Test Accuracy: 78.12%
[ 19/ 50]	FashionMNIST	Time 0.02108s	Training Accuracy: 86.82%	Test Accuracy: 71.88%
[ 20/ 50]	       MNIST	Time 0.02127s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 20/ 50]	FashionMNIST	Time 0.02089s	Training Accuracy: 87.21%	Test Accuracy: 62.50%
[ 21/ 50]	       MNIST	Time 0.02149s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 21/ 50]	FashionMNIST	Time 0.02119s	Training Accuracy: 87.79%	Test Accuracy: 68.75%
[ 22/ 50]	       MNIST	Time 0.02002s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 22/ 50]	FashionMNIST	Time 0.02091s	Training Accuracy: 89.06%	Test Accuracy: 71.88%
[ 23/ 50]	       MNIST	Time 0.02035s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 23/ 50]	FashionMNIST	Time 0.02051s	Training Accuracy: 90.43%	Test Accuracy: 68.75%
[ 24/ 50]	       MNIST	Time 0.02092s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 24/ 50]	FashionMNIST	Time 0.02109s	Training Accuracy: 91.02%	Test Accuracy: 68.75%
[ 25/ 50]	       MNIST	Time 0.02049s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 25/ 50]	FashionMNIST	Time 0.02134s	Training Accuracy: 92.48%	Test Accuracy: 71.88%
[ 26/ 50]	       MNIST	Time 0.02814s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 26/ 50]	FashionMNIST	Time 0.02124s	Training Accuracy: 91.41%	Test Accuracy: 71.88%
[ 27/ 50]	       MNIST	Time 0.02085s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 27/ 50]	FashionMNIST	Time 0.02181s	Training Accuracy: 92.68%	Test Accuracy: 78.12%
[ 28/ 50]	       MNIST	Time 0.02123s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 28/ 50]	FashionMNIST	Time 0.02125s	Training Accuracy: 92.68%	Test Accuracy: 78.12%
[ 29/ 50]	       MNIST	Time 0.02102s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 29/ 50]	FashionMNIST	Time 0.02081s	Training Accuracy: 93.65%	Test Accuracy: 75.00%
[ 30/ 50]	       MNIST	Time 0.02106s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 30/ 50]	FashionMNIST	Time 0.02752s	Training Accuracy: 94.14%	Test Accuracy: 78.12%
[ 31/ 50]	       MNIST	Time 0.02086s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 31/ 50]	FashionMNIST	Time 0.02075s	Training Accuracy: 94.53%	Test Accuracy: 78.12%
[ 32/ 50]	       MNIST	Time 0.02064s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 32/ 50]	FashionMNIST	Time 0.02118s	Training Accuracy: 94.63%	Test Accuracy: 78.12%
[ 33/ 50]	       MNIST	Time 0.02125s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 33/ 50]	FashionMNIST	Time 0.02206s	Training Accuracy: 94.34%	Test Accuracy: 78.12%
[ 34/ 50]	       MNIST	Time 0.02159s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 34/ 50]	FashionMNIST	Time 0.02064s	Training Accuracy: 94.73%	Test Accuracy: 78.12%
[ 35/ 50]	       MNIST	Time 0.02600s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 35/ 50]	FashionMNIST	Time 0.02224s	Training Accuracy: 95.12%	Test Accuracy: 78.12%
[ 36/ 50]	       MNIST	Time 0.02068s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 36/ 50]	FashionMNIST	Time 0.02117s	Training Accuracy: 95.31%	Test Accuracy: 78.12%
[ 37/ 50]	       MNIST	Time 0.02063s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 37/ 50]	FashionMNIST	Time 0.02126s	Training Accuracy: 95.31%	Test Accuracy: 78.12%
[ 38/ 50]	       MNIST	Time 0.02142s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 38/ 50]	FashionMNIST	Time 0.02136s	Training Accuracy: 95.90%	Test Accuracy: 78.12%
[ 39/ 50]	       MNIST	Time 0.02099s	Training Accuracy: 100.00%	Test Accuracy: 81.25%
[ 39/ 50]	FashionMNIST	Time 0.02024s	Training Accuracy: 95.80%	Test Accuracy: 78.12%
[ 40/ 50]	       MNIST	Time 0.02162s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 40/ 50]	FashionMNIST	Time 0.02015s	Training Accuracy: 96.19%	Test Accuracy: 78.12%
[ 41/ 50]	       MNIST	Time 0.02085s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 41/ 50]	FashionMNIST	Time 0.02181s	Training Accuracy: 96.09%	Test Accuracy: 78.12%
[ 42/ 50]	       MNIST	Time 0.02032s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 42/ 50]	FashionMNIST	Time 0.02672s	Training Accuracy: 96.39%	Test Accuracy: 78.12%
[ 43/ 50]	       MNIST	Time 0.02700s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 43/ 50]	FashionMNIST	Time 0.02651s	Training Accuracy: 96.39%	Test Accuracy: 78.12%
[ 44/ 50]	       MNIST	Time 0.02030s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 44/ 50]	FashionMNIST	Time 0.02073s	Training Accuracy: 96.78%	Test Accuracy: 78.12%
[ 45/ 50]	       MNIST	Time 0.02073s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 45/ 50]	FashionMNIST	Time 0.02101s	Training Accuracy: 96.78%	Test Accuracy: 78.12%
[ 46/ 50]	       MNIST	Time 0.02036s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 46/ 50]	FashionMNIST	Time 0.02085s	Training Accuracy: 96.88%	Test Accuracy: 78.12%
[ 47/ 50]	       MNIST	Time 0.02098s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 47/ 50]	FashionMNIST	Time 0.02038s	Training Accuracy: 96.88%	Test Accuracy: 78.12%
[ 48/ 50]	       MNIST	Time 0.02740s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 48/ 50]	FashionMNIST	Time 0.02044s	Training Accuracy: 97.17%	Test Accuracy: 78.12%
[ 49/ 50]	       MNIST	Time 0.02146s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 49/ 50]	FashionMNIST	Time 0.02053s	Training Accuracy: 96.97%	Test Accuracy: 78.12%
[ 50/ 50]	       MNIST	Time 0.02046s	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[ 50/ 50]	FashionMNIST	Time 0.02094s	Training Accuracy: 97.27%	Test Accuracy: 78.12%

[FINAL]	       MNIST	Training Accuracy: 100.00%	Test Accuracy: 78.12%
[FINAL]	FashionMNIST	Training Accuracy: 97.27%	Test Accuracy: 78.12%

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: 0.15.3+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.