MNIST Classification with SimpleChains

SimpleChains.jl is an excellent framework for training small neural networks. In this tutorial we will demonstrate how to use the same API as Lux.jl to train a model using SimpleChains.jl. We will use the tutorial from SimpleChains.jl as a reference.

Package Imports

using Lux, MLUtils, Optimisers, Zygote, OneHotArrays, Random, Statistics, Printf, Reactant
using MLDatasets: MNIST
using SimpleChains: SimpleChains

Reactant.set_default_backend("cpu")

Precompiling Lux...
   3546.0 ms  ✓ ForwardDiff
    948.8 ms  ✓ ForwardDiff → ForwardDiffStaticArraysExt
    885.5 ms  ✓ NNlib → NNlibForwardDiffExt
   5481.5 ms  ✓ LuxLib
   9395.9 ms  ✓ Lux
  5 dependencies successfully precompiled in 20 seconds. 100 already precompiled.
Precompiling LuxMLUtilsExt...
   1996.1 ms  ✓ Lux → LuxMLUtilsExt
  1 dependency successfully precompiled in 2 seconds. 164 already precompiled.
Precompiling Zygote...
  34106.0 ms  ✓ Zygote
  1 dependency successfully precompiled in 34 seconds. 64 already precompiled.
Precompiling StructArraysExt...
    475.0 ms  ✓ Accessors → StructArraysExt
  1 dependency successfully precompiled in 1 seconds. 20 already precompiled.
Precompiling BangBangStructArraysExt...
    604.5 ms  ✓ BangBang → BangBangStructArraysExt
  1 dependency successfully precompiled in 1 seconds. 24 already precompiled.
Precompiling MLDataDevicesZygoteExt...
   1474.2 ms  ✓ MLDataDevices → MLDataDevicesZygoteExt
  1 dependency successfully precompiled in 2 seconds. 71 already precompiled.
Precompiling LuxZygoteExt...
   2656.2 ms  ✓ Lux → LuxZygoteExt
  1 dependency successfully precompiled in 3 seconds. 143 already precompiled.
Precompiling Reactant...
    693.8 ms  ✓ URIs
  21653.4 ms  ✓ HTTP
  87965.0 ms  ✓ Reactant
  3 dependencies successfully precompiled in 111 seconds. 77 already precompiled.
Precompiling LuxLibEnzymeExt...
   1300.4 ms  ✓ LuxLib → LuxLibEnzymeExt
  1 dependency successfully precompiled in 2 seconds. 133 already precompiled.
Precompiling LuxEnzymeExt...
   7954.5 ms  ✓ Lux → LuxEnzymeExt
  1 dependency successfully precompiled in 8 seconds. 149 already precompiled.
Precompiling OptimisersReactantExt...
  17697.6 ms  ✓ Reactant → ReactantStatisticsExt
  20518.6 ms  ✓ Optimisers → OptimisersReactantExt
  2 dependencies successfully precompiled in 21 seconds. 88 already precompiled.
Precompiling LuxCoreReactantExt...
  17475.2 ms  ✓ LuxCore → LuxCoreReactantExt
  1 dependency successfully precompiled in 18 seconds. 85 already precompiled.
Precompiling MLDataDevicesReactantExt...
  17292.9 ms  ✓ MLDataDevices → MLDataDevicesReactantExt
  1 dependency successfully precompiled in 18 seconds. 82 already precompiled.
Precompiling LuxLibReactantExt...
  18266.7 ms  ✓ LuxLib → LuxLibReactantExt
  18854.9 ms  ✓ Reactant → ReactantSpecialFunctionsExt
  19132.0 ms  ✓ Reactant → ReactantKernelAbstractionsExt
  17435.2 ms  ✓ Reactant → ReactantArrayInterfaceExt
  4 dependencies successfully precompiled in 36 seconds. 158 already precompiled.
Precompiling WeightInitializersReactantExt...
  18044.9 ms  ✓ WeightInitializers → WeightInitializersReactantExt
  1 dependency successfully precompiled in 18 seconds. 96 already precompiled.
Precompiling ReactantAbstractFFTsExt...
  17319.2 ms  ✓ Reactant → ReactantAbstractFFTsExt
  1 dependency successfully precompiled in 18 seconds. 82 already precompiled.
Precompiling ReactantOneHotArraysExt...
  17333.5 ms  ✓ Reactant → ReactantOneHotArraysExt
  1 dependency successfully precompiled in 18 seconds. 104 already precompiled.
Precompiling ReactantNNlibExt...
  19863.2 ms  ✓ Reactant → ReactantNNlibExt
  1 dependency successfully precompiled in 20 seconds. 103 already precompiled.
Precompiling LuxReactantExt...
  12465.3 ms  ✓ Lux → LuxReactantExt
  1 dependency successfully precompiled in 13 seconds. 180 already precompiled.
Precompiling MLDatasets...
   1904.6 ms  ✓ FileIO → HTTPExt
   1962.8 ms  ✓ HDF5_jll
   3334.7 ms  ✓ DataDeps
   7461.3 ms  ✓ HDF5
   2351.6 ms  ✓ MAT
  10278.4 ms  ✓ MLDatasets
  6 dependencies successfully precompiled in 23 seconds. 197 already precompiled.
Precompiling SimpleChains...
    344.7 ms  ✓ UnPack
    494.0 ms  ✓ StaticArrayInterface → StaticArrayInterfaceOffsetArraysExt
    832.8 ms  ✓ HostCPUFeatures
   7695.4 ms  ✓ VectorizationBase
   1055.7 ms  ✓ SLEEFPirates
   1346.1 ms  ✓ VectorizedRNG
    719.5 ms  ✓ VectorizedRNG → VectorizedRNGStaticArraysExt
  28829.5 ms  ✓ LoopVectorization
    993.4 ms  ✓ LoopVectorization → SpecialFunctionsExt
   1249.8 ms  ✓ LoopVectorization → ForwardDiffExt
   6218.2 ms  ✓ SimpleChains
  11 dependencies successfully precompiled in 46 seconds. 58 already precompiled.
Precompiling LuxLibSLEEFPiratesExt...
   2502.3 ms  ✓ LuxLib → LuxLibSLEEFPiratesExt
  1 dependency successfully precompiled in 3 seconds. 93 already precompiled.
Precompiling ReactantOffsetArraysExt...
  17736.7 ms  ✓ Reactant → ReactantOffsetArraysExt
  1 dependency successfully precompiled in 18 seconds. 82 already precompiled.
Precompiling LuxLibLoopVectorizationExt...
   4767.6 ms  ✓ LuxLib → LuxLibLoopVectorizationExt
  1 dependency successfully precompiled in 5 seconds. 101 already precompiled.
Precompiling LuxSimpleChainsExt...
   2087.5 ms  ✓ Lux → LuxSimpleChainsExt
  1 dependency successfully precompiled in 2 seconds. 122 already precompiled.
2025-07-03 16:00:41.042426: I external/xla/xla/service/service.cc:153] XLA service 0x2edf4280 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2025-07-03 16:00:41.042466: I external/xla/xla/service/service.cc:161]   StreamExecutor device (0): NVIDIA A100-PCIE-40GB MIG 1g.5gb, Compute Capability 8.0
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
I0000 00:00:1751558441.043239  129860 se_gpu_pjrt_client.cc:1370] Using BFC allocator.
I0000 00:00:1751558441.043310  129860 gpu_helpers.cc:136] XLA backend allocating 3825205248 bytes on device 0 for BFCAllocator.
I0000 00:00:1751558441.043339  129860 gpu_helpers.cc:177] XLA backend will use up to 1275068416 bytes on device 0 for CollectiveBFCAllocator.
I0000 00:00:1751558441.058531  129860 cuda_dnn.cc:471] Loaded cuDNN version 90800

Loading MNIST

function loadmnist(batchsize, train_split)
    # Load MNIST
    N = parse(Bool, get(ENV, "CI", "false")) ? 1500 : nothing
    dataset = MNIST(; split=:train)
    if N !== nothing
        imgs = dataset.features[:, :, 1:N]
        labels_raw = dataset.targets[1:N]
    else
        imgs = dataset.features
        labels_raw = dataset.targets
    end

    # Process images into (H, W, C, BS) batches
    x_data = Float32.(reshape(imgs, size(imgs, 1), size(imgs, 2), 1, size(imgs, 3)))
    y_data = onehotbatch(labels_raw, 0:9)
    (x_train, y_train), (x_test, y_test) = splitobs((x_data, y_data); at=train_split)

    return (
        # Use DataLoader to automatically minibatch and shuffle the data
        DataLoader(collect.((x_train, y_train)); batchsize, shuffle=true, partial=false),
        # Don't shuffle the test data
        DataLoader(collect.((x_test, y_test)); batchsize, shuffle=false, partial=false),
    )
end

loadmnist (generic function with 1 method)

Define the Model

lux_model = Chain(
    Conv((5, 5), 1 => 6, relu),
    MaxPool((2, 2)),
    Conv((5, 5), 6 => 16, relu),
    MaxPool((2, 2)),
    FlattenLayer(3),
    Chain(Dense(256 => 128, relu), Dense(128 => 84, relu), Dense(84 => 10)),
)

Chain(
    layer_1 = Conv((5, 5), 1 => 6, relu),  # 156 parameters
    layer_2 = MaxPool((2, 2)),
    layer_3 = Conv((5, 5), 6 => 16, relu),  # 2_416 parameters
    layer_4 = MaxPool((2, 2)),
    layer_5 = Lux.FlattenLayer{Static.StaticInt{3}}(static(3)),
    layer_6 = Chain(
        layer_1 = Dense(256 => 128, relu),  # 32_896 parameters
        layer_2 = Dense(128 => 84, relu),  # 10_836 parameters
        layer_3 = Dense(84 => 10),      # 850 parameters
    ),
)         # Total: 47_154 parameters,
          #        plus 0 states.

We now need to convert the lux_model to SimpleChains.jl. We need to do this by defining the ToSimpleChainsAdaptor and providing the input dimensions.

adaptor = ToSimpleChainsAdaptor((28, 28, 1))
simple_chains_model = adaptor(lux_model)

SimpleChainsLayer(
    Chain(
        layer_1 = Conv((5, 5), 1 => 6, relu),  # 156 parameters
        layer_2 = MaxPool((2, 2)),
        layer_3 = Conv((5, 5), 6 => 16, relu),  # 2_416 parameters
        layer_4 = MaxPool((2, 2)),
        layer_5 = Lux.FlattenLayer{Static.StaticInt{3}}(static(3)),
        layer_6 = Chain(
            layer_1 = Dense(256 => 128, relu),  # 32_896 parameters
            layer_2 = Dense(128 => 84, relu),  # 10_836 parameters
            layer_3 = Dense(84 => 10),  # 850 parameters
        ),
    ),
)         # Total: 47_154 parameters,
          #        plus 0 states.

Helper Functions

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

function accuracy(model, ps, st, dataloader)
    total_correct, total = 0, 0
    st = Lux.testmode(st)
    for (x, y) in dataloader
        target_class = onecold(Array(y))
        predicted_class = onecold(Array(first(model(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)

Define the Training Loop

function train(model, dev=cpu_device(); rng=Random.default_rng(), kwargs...)
    train_dataloader, test_dataloader = dev(loadmnist(128, 0.9))
    ps, st = dev(Lux.setup(rng, model))

    vjp = dev isa ReactantDevice ? AutoEnzyme() : AutoZygote()

    train_state = Training.TrainState(model, ps, st, Adam(3.0f-4))

    if dev isa ReactantDevice
        x_ra = first(test_dataloader)[1]
        model_compiled = Reactant.with_config(;
            dot_general_precision=PrecisionConfig.HIGH,
            convolution_precision=PrecisionConfig.HIGH,
        ) do
            @compile model(x_ra, ps, Lux.testmode(st))
        end
    else
        model_compiled = model
    end

    ### Lets train the model
    nepochs = 10
    tr_acc, te_acc = 0.0, 0.0
    for epoch in 1:nepochs
        stime = time()
        for (x, y) in train_dataloader
            _, _, _, train_state = Training.single_train_step!(
                vjp, lossfn, (x, y), train_state
            )
        end
        ttime = time() - stime

        tr_acc =
            accuracy(
                model_compiled, train_state.parameters, train_state.states, train_dataloader
            ) * 100
        te_acc =
            accuracy(
                model_compiled, train_state.parameters, train_state.states, test_dataloader
            ) * 100

        @printf "[%2d/%2d] \t Time %.2fs \t Training Accuracy: %.2f%% \t Test Accuracy: \
                 %.2f%%\n" epoch nepochs ttime tr_acc te_acc
    end

    return tr_acc, te_acc
end

train (generic function with 2 methods)

Finally Training the Model

First we will train the Lux model

tr_acc, te_acc = train(lux_model, reactant_device())

[ 1/10] 	 Time 396.89s 	 Training Accuracy: 22.27% 	 Test Accuracy: 24.22%
[ 2/10] 	 Time 0.11s 	 Training Accuracy: 31.25% 	 Test Accuracy: 26.56%
[ 3/10] 	 Time 0.09s 	 Training Accuracy: 38.91% 	 Test Accuracy: 36.72%
[ 4/10] 	 Time 0.10s 	 Training Accuracy: 50.31% 	 Test Accuracy: 42.19%
[ 5/10] 	 Time 0.09s 	 Training Accuracy: 57.81% 	 Test Accuracy: 47.66%
[ 6/10] 	 Time 0.09s 	 Training Accuracy: 65.31% 	 Test Accuracy: 54.69%
[ 7/10] 	 Time 0.09s 	 Training Accuracy: 71.56% 	 Test Accuracy: 64.84%
[ 8/10] 	 Time 0.10s 	 Training Accuracy: 76.09% 	 Test Accuracy: 71.88%
[ 9/10] 	 Time 0.10s 	 Training Accuracy: 78.83% 	 Test Accuracy: 72.66%
[10/10] 	 Time 0.12s 	 Training Accuracy: 81.17% 	 Test Accuracy: 75.78%

Now we will train the SimpleChains model

tr_acc, te_acc = train(simple_chains_model)

[ 1/10] 	 Time 931.22s 	 Training Accuracy: 19.92% 	 Test Accuracy: 17.97%
[ 2/10] 	 Time 12.17s 	 Training Accuracy: 46.17% 	 Test Accuracy: 45.31%
[ 3/10] 	 Time 12.16s 	 Training Accuracy: 61.33% 	 Test Accuracy: 57.81%
[ 4/10] 	 Time 12.14s 	 Training Accuracy: 74.69% 	 Test Accuracy: 69.53%
[ 5/10] 	 Time 12.17s 	 Training Accuracy: 77.27% 	 Test Accuracy: 71.09%
[ 6/10] 	 Time 12.23s 	 Training Accuracy: 82.97% 	 Test Accuracy: 78.91%
[ 7/10] 	 Time 12.25s 	 Training Accuracy: 83.98% 	 Test Accuracy: 78.12%
[ 8/10] 	 Time 12.22s 	 Training Accuracy: 86.25% 	 Test Accuracy: 82.81%
[ 9/10] 	 Time 12.21s 	 Training Accuracy: 86.48% 	 Test Accuracy: 79.69%
[10/10] 	 Time 12.19s 	 Training Accuracy: 87.19% 	 Test Accuracy: 80.47%

On my local machine we see a 3-4x speedup when using SimpleChains.jl. The conditions of the server this documentation is being built on is not ideal for CPU benchmarking hence, the speedup may not be as significant and even there might be regressions.

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 1.11.5
Commit 760b2e5b739 (2025-04-14 06:53 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
  LLVM: libLLVM-16.0.6 (ORCJIT, znver2)
Threads: 48 default, 0 interactive, 24 GC (on 2 virtual cores)
Environment:
  JULIA_CPU_THREADS = 2
  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
  JULIA_DEPOT_PATH = /root/.cache/julia-buildkite-plugin/depots/01872db4-8c79-43af-ab7d-12abac4f24f6

---

This page was generated using Literate.jl.