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(3.0f-4))

    ### Lets train the model
    nepochs = 25
    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 \t Time %.5fs \t Training Accuracy: %.2f%% \t Test \
                 Accuracy: %.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 \t Training Accuracy: %.2f%% \t Test Accuracy: \
                 %.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/ 25] 	        MNIST 	 Time 70.34137s 	 Training Accuracy: 24.61% 	 Test Accuracy: 25.00%
[  1/ 25] 	 FashionMNIST 	 Time 0.05165s 	 Training Accuracy: 31.64% 	 Test Accuracy: 28.12%
[  2/ 25] 	        MNIST 	 Time 0.02839s 	 Training Accuracy: 50.98% 	 Test Accuracy: 37.50%
[  2/ 25] 	 FashionMNIST 	 Time 0.02865s 	 Training Accuracy: 54.69% 	 Test Accuracy: 46.88%
[  3/ 25] 	        MNIST 	 Time 0.02797s 	 Training Accuracy: 60.64% 	 Test Accuracy: 59.38%
[  3/ 25] 	 FashionMNIST 	 Time 0.02738s 	 Training Accuracy: 63.18% 	 Test Accuracy: 56.25%
[  4/ 25] 	        MNIST 	 Time 0.02813s 	 Training Accuracy: 69.73% 	 Test Accuracy: 50.00%
[  4/ 25] 	 FashionMNIST 	 Time 0.06439s 	 Training Accuracy: 67.87% 	 Test Accuracy: 56.25%
[  5/ 25] 	        MNIST 	 Time 0.02043s 	 Training Accuracy: 73.24% 	 Test Accuracy: 53.12%
[  5/ 25] 	 FashionMNIST 	 Time 0.02181s 	 Training Accuracy: 69.14% 	 Test Accuracy: 59.38%
[  6/ 25] 	        MNIST 	 Time 0.02097s 	 Training Accuracy: 80.76% 	 Test Accuracy: 62.50%
[  6/ 25] 	 FashionMNIST 	 Time 0.02071s 	 Training Accuracy: 77.54% 	 Test Accuracy: 68.75%
[  7/ 25] 	        MNIST 	 Time 0.02084s 	 Training Accuracy: 82.81% 	 Test Accuracy: 68.75%
[  7/ 25] 	 FashionMNIST 	 Time 0.02176s 	 Training Accuracy: 77.44% 	 Test Accuracy: 68.75%
[  8/ 25] 	        MNIST 	 Time 0.02193s 	 Training Accuracy: 87.40% 	 Test Accuracy: 68.75%
[  8/ 25] 	 FashionMNIST 	 Time 0.03834s 	 Training Accuracy: 81.35% 	 Test Accuracy: 65.62%
[  9/ 25] 	        MNIST 	 Time 0.02064s 	 Training Accuracy: 88.77% 	 Test Accuracy: 65.62%
[  9/ 25] 	 FashionMNIST 	 Time 0.02082s 	 Training Accuracy: 79.39% 	 Test Accuracy: 59.38%
[ 10/ 25] 	        MNIST 	 Time 0.02032s 	 Training Accuracy: 91.80% 	 Test Accuracy: 65.62%
[ 10/ 25] 	 FashionMNIST 	 Time 0.02079s 	 Training Accuracy: 84.18% 	 Test Accuracy: 62.50%
[ 11/ 25] 	        MNIST 	 Time 0.02130s 	 Training Accuracy: 93.95% 	 Test Accuracy: 68.75%
[ 11/ 25] 	 FashionMNIST 	 Time 0.02083s 	 Training Accuracy: 84.28% 	 Test Accuracy: 59.38%
[ 12/ 25] 	        MNIST 	 Time 0.02087s 	 Training Accuracy: 95.80% 	 Test Accuracy: 68.75%
[ 12/ 25] 	 FashionMNIST 	 Time 0.02076s 	 Training Accuracy: 84.18% 	 Test Accuracy: 59.38%
[ 13/ 25] 	        MNIST 	 Time 0.03383s 	 Training Accuracy: 96.09% 	 Test Accuracy: 62.50%
[ 13/ 25] 	 FashionMNIST 	 Time 0.02042s 	 Training Accuracy: 84.28% 	 Test Accuracy: 56.25%
[ 14/ 25] 	        MNIST 	 Time 0.02569s 	 Training Accuracy: 95.51% 	 Test Accuracy: 65.62%
[ 14/ 25] 	 FashionMNIST 	 Time 0.02999s 	 Training Accuracy: 85.94% 	 Test Accuracy: 53.12%
[ 15/ 25] 	        MNIST 	 Time 0.02794s 	 Training Accuracy: 97.85% 	 Test Accuracy: 65.62%
[ 15/ 25] 	 FashionMNIST 	 Time 0.02341s 	 Training Accuracy: 86.72% 	 Test Accuracy: 68.75%
[ 16/ 25] 	        MNIST 	 Time 0.02842s 	 Training Accuracy: 98.34% 	 Test Accuracy: 59.38%
[ 16/ 25] 	 FashionMNIST 	 Time 0.01969s 	 Training Accuracy: 88.77% 	 Test Accuracy: 65.62%
[ 17/ 25] 	        MNIST 	 Time 0.02103s 	 Training Accuracy: 99.02% 	 Test Accuracy: 59.38%
[ 17/ 25] 	 FashionMNIST 	 Time 0.02706s 	 Training Accuracy: 91.21% 	 Test Accuracy: 75.00%
[ 18/ 25] 	        MNIST 	 Time 0.02319s 	 Training Accuracy: 98.93% 	 Test Accuracy: 62.50%
[ 18/ 25] 	 FashionMNIST 	 Time 0.02128s 	 Training Accuracy: 91.31% 	 Test Accuracy: 71.88%
[ 19/ 25] 	        MNIST 	 Time 0.02119s 	 Training Accuracy: 99.22% 	 Test Accuracy: 62.50%
[ 19/ 25] 	 FashionMNIST 	 Time 0.02110s 	 Training Accuracy: 90.14% 	 Test Accuracy: 65.62%
[ 20/ 25] 	        MNIST 	 Time 0.03255s 	 Training Accuracy: 100.00% 	 Test Accuracy: 68.75%
[ 20/ 25] 	 FashionMNIST 	 Time 0.02236s 	 Training Accuracy: 88.09% 	 Test Accuracy: 65.62%
[ 21/ 25] 	        MNIST 	 Time 0.02123s 	 Training Accuracy: 98.93% 	 Test Accuracy: 65.62%
[ 21/ 25] 	 FashionMNIST 	 Time 0.02194s 	 Training Accuracy: 86.23% 	 Test Accuracy: 59.38%
[ 22/ 25] 	        MNIST 	 Time 0.02066s 	 Training Accuracy: 99.22% 	 Test Accuracy: 65.62%
[ 22/ 25] 	 FashionMNIST 	 Time 0.02110s 	 Training Accuracy: 89.84% 	 Test Accuracy: 65.62%
[ 23/ 25] 	        MNIST 	 Time 0.02071s 	 Training Accuracy: 99.80% 	 Test Accuracy: 62.50%
[ 23/ 25] 	 FashionMNIST 	 Time 0.02081s 	 Training Accuracy: 91.60% 	 Test Accuracy: 65.62%
[ 24/ 25] 	        MNIST 	 Time 0.02063s 	 Training Accuracy: 99.32% 	 Test Accuracy: 65.62%
[ 24/ 25] 	 FashionMNIST 	 Time 0.02078s 	 Training Accuracy: 91.11% 	 Test Accuracy: 62.50%
[ 25/ 25] 	        MNIST 	 Time 0.02201s 	 Training Accuracy: 99.61% 	 Test Accuracy: 62.50%
[ 25/ 25] 	 FashionMNIST 	 Time 0.04165s 	 Training Accuracy: 92.09% 	 Test Accuracy: 68.75%

[FINAL] 	        MNIST 	 Training Accuracy: 100.00% 	 Test Accuracy: 62.50%
[FINAL] 	 FashionMNIST 	 Training Accuracy: 92.09% 	 Test Accuracy: 68.75%

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 1.10.5
Commit 6f3fdf7b362 (2024-08-27 14:19 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 12.5, artifact installation
CUDA driver 12.5
NVIDIA driver 555.42.6

CUDA libraries: 
- CUBLAS: 12.5.3
- CURAND: 10.3.6
- CUFFT: 11.2.3
- CUSOLVER: 11.6.3
- CUSPARSE: 12.5.1
- CUPTI: 2024.2.1 (API 23.0.0)
- NVML: 12.0.0+555.42.6

Julia packages: 
- CUDA: 5.4.3
- CUDA_Driver_jll: 0.9.2+0
- CUDA_Runtime_jll: 0.14.1+0

Toolchain:
- Julia: 1.10.5
- 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.