Training a Simple LSTM

In this tutorial we will go over using a recurrent neural network to classify clockwise and anticlockwise spirals. By the end of this tutorial you will be able to:

1. Create custom Lux models.

2. Become familiar with the Lux recurrent neural network API.

3. Training using Optimisers.jl and Zygote.jl.

Package Imports

Note: If you wish to use AutoZygote() for automatic differentiation, add Zygote to your project dependencies and include using Zygote.

using ADTypes, Lux, JLD2, MLUtils, Optimisers, Printf, Reactant, Random

Precompiling Reactant...
   2800.8 ms  ✓ Reactant_jll
  88800.1 ms  ✓ Reactant
  2 dependencies successfully precompiled in 92 seconds. 78 already precompiled.
Precompiling OptimisersReactantExt...
  17211.4 ms  ✓ Reactant → ReactantStatisticsExt
  19505.3 ms  ✓ Optimisers → OptimisersReactantExt
  2 dependencies successfully precompiled in 20 seconds. 88 already precompiled.
Precompiling LuxCoreReactantExt...
  17312.9 ms  ✓ LuxCore → LuxCoreReactantExt
  1 dependency successfully precompiled in 18 seconds. 85 already precompiled.
Precompiling MLDataDevicesReactantExt...
  17530.6 ms  ✓ MLDataDevices → MLDataDevicesReactantExt
  1 dependency successfully precompiled in 18 seconds. 82 already precompiled.
Precompiling LuxLibReactantExt...
  17862.2 ms  ✓ Reactant → ReactantSpecialFunctionsExt
  17960.7 ms  ✓ Reactant → ReactantKernelAbstractionsExt
  18346.3 ms  ✓ LuxLib → LuxLibReactantExt
  16886.4 ms  ✓ Reactant → ReactantArrayInterfaceExt
  4 dependencies successfully precompiled in 35 seconds. 158 already precompiled.
Precompiling WeightInitializersReactantExt...
  17177.0 ms  ✓ WeightInitializers → WeightInitializersReactantExt
  1 dependency successfully precompiled in 17 seconds. 96 already precompiled.
Precompiling ReactantNNlibExt...
  20648.6 ms  ✓ Reactant → ReactantNNlibExt
  1 dependency successfully precompiled in 21 seconds. 103 already precompiled.
Precompiling LuxReactantExt...
  12178.0 ms  ✓ Lux → LuxReactantExt
  1 dependency successfully precompiled in 13 seconds. 180 already precompiled.

Dataset

We will use MLUtils to generate 500 (noisy) clockwise and 500 (noisy) anticlockwise spirals. Using this data we will create a MLUtils.DataLoader. Our dataloader will give us sequences of size 2 × seq_len × batch_size and we need to predict a binary value whether the sequence is clockwise or anticlockwise.

function get_dataloaders(; dataset_size=1000, sequence_length=50)
    # Create the spirals
    data = [MLUtils.Datasets.make_spiral(sequence_length) for _ in 1:dataset_size]
    # Get the labels
    labels = vcat(repeat([0.0f0], dataset_size ÷ 2), repeat([1.0f0], dataset_size ÷ 2))
    clockwise_spirals = [
        reshape(d[1][:, 1:sequence_length], :, sequence_length, 1) for
        d in data[1:(dataset_size ÷ 2)]
    ]
    anticlockwise_spirals = [
        reshape(d[1][:, (sequence_length + 1):end], :, sequence_length, 1) for
        d in data[((dataset_size ÷ 2) + 1):end]
    ]
    x_data = Float32.(cat(clockwise_spirals..., anticlockwise_spirals...; dims=3))
    # Split the dataset
    (x_train, y_train), (x_val, y_val) = splitobs((x_data, labels); at=0.8, shuffle=true)
    # Create DataLoaders
    return (
        # Use DataLoader to automatically minibatch and shuffle the data
        DataLoader(
            collect.((x_train, y_train)); batchsize=128, shuffle=true, partial=false
        ),
        # Don't shuffle the validation data
        DataLoader(collect.((x_val, y_val)); batchsize=128, shuffle=false, partial=false),
    )
end

get_dataloaders (generic function with 1 method)

Creating a Classifier

We will be extending the Lux.AbstractLuxContainerLayer type for our custom model since it will contain a lstm block and a classifier head.

We pass the fieldnames lstm_cell and classifier to the type to ensure that the parameters and states are automatically populated and we don't have to define Lux.initialparameters and Lux.initialstates.

To understand more about container layers, please look at Container Layer.

struct SpiralClassifier{L,C} <: AbstractLuxContainerLayer{(:lstm_cell, :classifier)}
    lstm_cell::L
    classifier::C
end

We won't define the model from scratch but rather use the Lux.LSTMCell and Lux.Dense.

function SpiralClassifier(in_dims, hidden_dims, out_dims)
    return SpiralClassifier(
        LSTMCell(in_dims => hidden_dims), Dense(hidden_dims => out_dims, sigmoid)
    )
end

Main.var"##230".SpiralClassifier

We can use default Lux blocks – Recurrence(LSTMCell(in_dims => hidden_dims) – instead of defining the following. But let's still do it for the sake of it.

Now we need to define the behavior of the Classifier when it is invoked.

function (s::SpiralClassifier)(
    x::AbstractArray{T,3}, ps::NamedTuple, st::NamedTuple
) where {T}
    # First we will have to run the sequence through the LSTM Cell
    # The first call to LSTM Cell will create the initial hidden state
    # See that the parameters and states are automatically populated into a field called
    # `lstm_cell` We use `eachslice` to get the elements in the sequence without copying,
    # and `Iterators.peel` to split out the first element for LSTM initialization.
    x_init, x_rest = Iterators.peel(LuxOps.eachslice(x, Val(2)))
    (y, carry), st_lstm = s.lstm_cell(x_init, ps.lstm_cell, st.lstm_cell)
    # Now that we have the hidden state and memory in `carry` we will pass the input and
    # `carry` jointly
    for x in x_rest
        (y, carry), st_lstm = s.lstm_cell((x, carry), ps.lstm_cell, st_lstm)
    end
    # After running through the sequence we will pass the output through the classifier
    y, st_classifier = s.classifier(y, ps.classifier, st.classifier)
    # Finally remember to create the updated state
    st = merge(st, (classifier=st_classifier, lstm_cell=st_lstm))
    return vec(y), st
end

Using the @compact API

We can also define the model using the Lux.@compact API, which is a more concise way of defining models. This macro automatically handles the boilerplate code for you and as such we recommend this way of defining custom layers

function SpiralClassifierCompact(in_dims, hidden_dims, out_dims)
    lstm_cell = LSTMCell(in_dims => hidden_dims)
    classifier = Dense(hidden_dims => out_dims, sigmoid)
    return @compact(; lstm_cell, classifier) do x::AbstractArray{T,3} where {T}
        x_init, x_rest = Iterators.peel(LuxOps.eachslice(x, Val(2)))
        y, carry = lstm_cell(x_init)
        for x in x_rest
            y, carry = lstm_cell((x, carry))
        end
        @return vec(classifier(y))
    end
end

SpiralClassifierCompact (generic function with 1 method)

Defining Accuracy, Loss and Optimiser

Now let's define the binarycrossentropy loss. Typically it is recommended to use logitbinarycrossentropy since it is more numerically stable, but for the sake of simplicity we will use binarycrossentropy.

const lossfn = BinaryCrossEntropyLoss()

function compute_loss(model, ps, st, (x, y))
    ŷ, st_ = model(x, ps, st)
    loss = lossfn(ŷ, y)
    return loss, st_, (; y_pred=ŷ)
end

matches(y_pred, y_true) = sum((y_pred .> 0.5f0) .== y_true)
accuracy(y_pred, y_true) = matches(y_pred, y_true) / length(y_pred)

accuracy (generic function with 1 method)

Training the Model

function main(model_type)
    dev = reactant_device()
    cdev = cpu_device()

    # Get the dataloaders
    train_loader, val_loader = dev(get_dataloaders())

    # Create the model
    model = model_type(2, 8, 1)
    ps, st = dev(Lux.setup(Random.default_rng(), model))

    train_state = Training.TrainState(model, ps, st, Adam(0.01f0))
    model_compiled = if dev isa ReactantDevice
        Reactant.with_config(;
            dot_general_precision=PrecisionConfig.HIGH,
            convolution_precision=PrecisionConfig.HIGH,
        ) do
            @compile model(first(train_loader)[1], ps, Lux.testmode(st))
        end
    else
        model
    end
    ad = dev isa ReactantDevice ? AutoEnzyme() : AutoZygote()

    for epoch in 1:25
        # Train the model
        total_loss = 0.0f0
        total_samples = 0
        for (x, y) in train_loader
            (_, loss, _, train_state) = Training.single_train_step!(
                ad, lossfn, (x, y), train_state
            )
            total_loss += loss * length(y)
            total_samples += length(y)
        end
        @printf "Epoch [%3d]: Loss %4.5f\n" epoch (total_loss / total_samples)

        # Validate the model
        total_acc = 0.0f0
        total_loss = 0.0f0
        total_samples = 0

        st_ = Lux.testmode(train_state.states)
        for (x, y) in val_loader
            ŷ, st_ = model_compiled(x, train_state.parameters, st_)
            ŷ, y = cdev(ŷ), cdev(y)
            total_acc += accuracy(ŷ, y) * length(y)
            total_loss += lossfn(ŷ, y) * length(y)
            total_samples += length(y)
        end

        @printf "Validation:\tLoss %4.5f\tAccuracy %4.5f\n" (total_loss / total_samples) (
            total_acc / total_samples
        )
    end

    return cpu_device()((train_state.parameters, train_state.states))
end

ps_trained, st_trained = main(SpiralClassifier)

┌ Warning: `replicate` doesn't work for `TaskLocalRNG`. Returning the same `TaskLocalRNG`.
└ @ LuxCore /var/lib/buildkite-agent/builds/gpuci-9/julialang/lux-dot-jl/lib/LuxCore/src/LuxCore.jl:18
2025-07-09 04:07:46.917887: I external/xla/xla/service/service.cc:153] XLA service 0x45b36020 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2025-07-09 04:07:46.917968: 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 00:00:1752034066.918782 1104280 se_gpu_pjrt_client.cc:1370] Using BFC allocator.
I0000 00:00:1752034066.918907 1104280 gpu_helpers.cc:136] XLA backend allocating 3825205248 bytes on device 0 for BFCAllocator.
I0000 00:00:1752034066.918993 1104280 gpu_helpers.cc:177] XLA backend will use up to 1275068416 bytes on device 0 for CollectiveBFCAllocator.
I0000 00:00:1752034066.933682 1104280 cuda_dnn.cc:471] Loaded cuDNN version 90800
Epoch [  1]: Loss 0.64280
Validation:	Loss 0.56839	Accuracy 0.46875
Epoch [  2]: Loss 0.54155
Validation:	Loss 0.52683	Accuracy 0.49219
Epoch [  3]: Loss 0.49968
Validation:	Loss 0.48309	Accuracy 1.00000
Epoch [  4]: Loss 0.45279
Validation:	Loss 0.43759	Accuracy 1.00000
Epoch [  5]: Loss 0.40737
Validation:	Loss 0.39060	Accuracy 1.00000
Epoch [  6]: Loss 0.35700
Validation:	Loss 0.33951	Accuracy 1.00000
Epoch [  7]: Loss 0.30293
Validation:	Loss 0.27055	Accuracy 1.00000
Epoch [  8]: Loss 0.22155
Validation:	Loss 0.17741	Accuracy 1.00000
Epoch [  9]: Loss 0.14966
Validation:	Loss 0.12521	Accuracy 1.00000
Epoch [ 10]: Loss 0.10884
Validation:	Loss 0.09290	Accuracy 1.00000
Epoch [ 11]: Loss 0.08258
Validation:	Loss 0.07103	Accuracy 1.00000
Epoch [ 12]: Loss 0.06342
Validation:	Loss 0.05455	Accuracy 1.00000
Epoch [ 13]: Loss 0.04874
Validation:	Loss 0.04165	Accuracy 1.00000
Epoch [ 14]: Loss 0.03807
Validation:	Loss 0.03329	Accuracy 1.00000
Epoch [ 15]: Loss 0.03060
Validation:	Loss 0.02637	Accuracy 1.00000
Epoch [ 16]: Loss 0.02404
Validation:	Loss 0.02021	Accuracy 1.00000
Epoch [ 17]: Loss 0.01894
Validation:	Loss 0.01659	Accuracy 1.00000
Epoch [ 18]: Loss 0.01614
Validation:	Loss 0.01445	Accuracy 1.00000
Epoch [ 19]: Loss 0.01423
Validation:	Loss 0.01292	Accuracy 1.00000
Epoch [ 20]: Loss 0.01278
Validation:	Loss 0.01174	Accuracy 1.00000
Epoch [ 21]: Loss 0.01164
Validation:	Loss 0.01077	Accuracy 1.00000
Epoch [ 22]: Loss 0.01071
Validation:	Loss 0.00993	Accuracy 1.00000
Epoch [ 23]: Loss 0.00990
Validation:	Loss 0.00919	Accuracy 1.00000
Epoch [ 24]: Loss 0.00916
Validation:	Loss 0.00853	Accuracy 1.00000
Epoch [ 25]: Loss 0.00850
Validation:	Loss 0.00793	Accuracy 1.00000

We can also train the compact model with the exact same code!

ps_trained2, st_trained2 = main(SpiralClassifierCompact)

┌ Warning: `replicate` doesn't work for `TaskLocalRNG`. Returning the same `TaskLocalRNG`.
└ @ LuxCore /var/lib/buildkite-agent/builds/gpuci-9/julialang/lux-dot-jl/lib/LuxCore/src/LuxCore.jl:18
Epoch [  1]: Loss 0.49720
Validation:	Loss 0.43628	Accuracy 1.00000
Epoch [  2]: Loss 0.39232
Validation:	Loss 0.31417	Accuracy 1.00000
Epoch [  3]: Loss 0.25683
Validation:	Loss 0.17034	Accuracy 1.00000
Epoch [  4]: Loss 0.13932
Validation:	Loss 0.10147	Accuracy 1.00000
Epoch [  5]: Loss 0.08779
Validation:	Loss 0.06674	Accuracy 1.00000
Epoch [  6]: Loss 0.05975
Validation:	Loss 0.04731	Accuracy 1.00000
Epoch [  7]: Loss 0.04352
Validation:	Loss 0.03584	Accuracy 1.00000
Epoch [  8]: Loss 0.03376
Validation:	Loss 0.02855	Accuracy 1.00000
Epoch [  9]: Loss 0.02730
Validation:	Loss 0.02337	Accuracy 1.00000
Epoch [ 10]: Loss 0.02235
Validation:	Loss 0.01911	Accuracy 1.00000
Epoch [ 11]: Loss 0.01806
Validation:	Loss 0.01526	Accuracy 1.00000
Epoch [ 12]: Loss 0.01438
Validation:	Loss 0.01234	Accuracy 1.00000
Epoch [ 13]: Loss 0.01197
Validation:	Loss 0.01054	Accuracy 1.00000
Epoch [ 14]: Loss 0.01040
Validation:	Loss 0.00929	Accuracy 1.00000
Epoch [ 15]: Loss 0.00929
Validation:	Loss 0.00834	Accuracy 1.00000
Epoch [ 16]: Loss 0.00838
Validation:	Loss 0.00757	Accuracy 1.00000
Epoch [ 17]: Loss 0.00767
Validation:	Loss 0.00693	Accuracy 1.00000
Epoch [ 18]: Loss 0.00704
Validation:	Loss 0.00638	Accuracy 1.00000
Epoch [ 19]: Loss 0.00652
Validation:	Loss 0.00591	Accuracy 1.00000
Epoch [ 20]: Loss 0.00606
Validation:	Loss 0.00549	Accuracy 1.00000
Epoch [ 21]: Loss 0.00565
Validation:	Loss 0.00512	Accuracy 1.00000
Epoch [ 22]: Loss 0.00527
Validation:	Loss 0.00479	Accuracy 1.00000
Epoch [ 23]: Loss 0.00492
Validation:	Loss 0.00449	Accuracy 1.00000
Epoch [ 24]: Loss 0.00463
Validation:	Loss 0.00421	Accuracy 1.00000
Epoch [ 25]: Loss 0.00433
Validation:	Loss 0.00396	Accuracy 1.00000

Saving the Model

We can save the model using JLD2 (and any other serialization library of your choice) Note that we transfer the model to CPU before saving. Additionally, we recommend that you don't save the model struct and only save the parameters and states.

@save "trained_model.jld2" ps_trained st_trained

Let's try loading the model

@load "trained_model.jld2" ps_trained st_trained

2-element Vector{Symbol}:
 :ps_trained
 :st_trained

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.