Getting Started

Installation

Install Julia v1.10 or above. Lux.jl is available through the Julia package manager. You can enter it by pressing ] in the REPL and then typing add Lux. Alternatively, you can also do

import Pkg
Pkg.add("Lux")

Update to v1

If you are using a pre-v1 version of Lux.jl, please see the Updating to v1 section for instructions on how to update.

Quickstart

Pre-Requisites

You need to install Optimisers and Zygote if not done already. Pkg.add(["Optimisers", "Zygote"])

using Lux, Random, Optimisers, Zygote
# using LuxCUDA, AMDGPU, Metal, oneAPI # Optional packages for GPU support

We take randomness very seriously

# Seeding
rng = Random.default_rng()
Random.seed!(rng, 0)

Random.TaskLocalRNG()

Build the model

# Construct the layer
model = Chain(Dense(128, 256, tanh), Chain(Dense(256, 1, tanh), Dense(1, 10)))

Chain(
    layer_1 = Dense(128 => 256, tanh),  # 33_024 parameters
    layer_2 = Chain(
        layer_1 = Dense(256 => 1, tanh),  # 257 parameters
        layer_2 = Dense(1 => 10),       # 20 parameters
    ),
)         # Total: 33_301 parameters,
          #        plus 0 states.

Models don't hold parameters and states so initialize them. From there on, we can just use our standard AD and Optimisers API. However, here we will show how to use Lux's Training API that provides an uniform API over all supported AD systems.

# Get the device determined by Lux
dev = gpu_device()

# Parameter and State Variables
ps, st = Lux.setup(rng, model) |> dev

# Dummy Input
x = rand(rng, Float32, 128, 2) |> dev

# Run the model
y, st = Lux.apply(model, x, ps, st)

# Gradients
## First construct a TrainState
train_state = Lux.Training.TrainState(model, ps, st, Adam(0.0001f0))

## We can compute the gradients using Training.compute_gradients
gs, loss, stats, train_state = Lux.Training.compute_gradients(
    AutoZygote(), MSELoss(),
    (x, dev(rand(rng, Float32, 10, 2))), train_state
)

## Optimization
train_state = Training.apply_gradients!(train_state, gs) # or Training.apply_gradients (no `!` at the end)

# Both these steps can be combined into a single call
gs, loss, stats, train_state = Training.single_train_step!(
    AutoZygote(), MSELoss(),
    (x, dev(rand(rng, Float32, 10, 2))), train_state
)

((layer_1 = (weight = Float32[0.0017983615 0.006062332 … 0.0053392933 0.0056276177; 0.0011292367 0.0041270256 … 0.003585879 0.0038155357; … ; -0.0008762945 -0.0031371699 … -0.0027350332 -0.0029033197; 0.0011154839 0.002197485 … 0.0021741025 0.0021157824], bias = Float32[0.006656272, 0.004425203, 0.0028994146, -0.0116051175, 0.0031301186, 0.0037318026, 0.0136483535, 0.013969757, -0.015173428, -0.005173992  …  -0.0018621369, -0.0015270555, -0.007873881, -0.0076395273, -0.0022123815, 0.0039605754, 0.0034407252, -0.0045406874, -0.003383829, 0.0029306945]), layer_2 = (layer_1 = (weight = Float32[0.04993449 0.03202845 … -0.059382 0.07701616], bias = Float32[0.08797912]), layer_2 = (weight = Float32[-0.094527975; -0.11476975; … ; -0.016841749; -0.0698748;;], bias = Float32[-0.21608135, -0.26255828, -0.23534852, -0.21524015, -0.055711076, -0.20314303, -0.1895644, 0.03666526, -0.03937737, -0.15905891]))), 0.8455785f0, NamedTuple(), Lux.Training.TrainState{Nothing, Nothing, Chain{@NamedTuple{layer_1::Dense{typeof(tanh), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Chain{@NamedTuple{layer_1::Dense{typeof(tanh), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}}, Nothing}, @NamedTuple{layer_1::@NamedTuple{weight::Matrix{Float32}, bias::Vector{Float32}}, layer_2::@NamedTuple{layer_1::@NamedTuple{weight::Matrix{Float32}, bias::Vector{Float32}}, layer_2::@NamedTuple{weight::Matrix{Float32}, bias::Vector{Float32}}}}, @NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{layer_1::@NamedTuple{}, layer_2::@NamedTuple{}}}, Adam{Float32, Tuple{Float64, Float64}, Float64}, @NamedTuple{layer_1::@NamedTuple{weight::Optimisers.Leaf{Adam{Float32, Tuple{Float64, Float64}, Float64}, Tuple{Matrix{Float32}, Matrix{Float32}, Tuple{Float32, Float32}}}, bias::Optimisers.Leaf{Adam{Float32, Tuple{Float64, Float64}, Float64}, Tuple{Vector{Float32}, Vector{Float32}, Tuple{Float32, Float32}}}}, layer_2::@NamedTuple{layer_1::@NamedTuple{weight::Optimisers.Leaf{Adam{Float32, Tuple{Float64, Float64}, Float64}, Tuple{Matrix{Float32}, Matrix{Float32}, Tuple{Float32, Float32}}}, bias::Optimisers.Leaf{Adam{Float32, Tuple{Float64, Float64}, Float64}, Tuple{Vector{Float32}, Vector{Float32}, Tuple{Float32, Float32}}}}, layer_2::@NamedTuple{weight::Optimisers.Leaf{Adam{Float32, Tuple{Float64, Float64}, Float64}, Tuple{Matrix{Float32}, Matrix{Float32}, Tuple{Float32, Float32}}}, bias::Optimisers.Leaf{Adam{Float32, Tuple{Float64, Float64}, Float64}, Tuple{Vector{Float32}, Vector{Float32}, Tuple{Float32, Float32}}}}}}}(nothing, nothing, Chain{@NamedTuple{layer_1::Dense{typeof(tanh), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Chain{@NamedTuple{layer_1::Dense{typeof(tanh), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}}, Nothing}((layer_1 = Dense(128 => 256, tanh), layer_2 = Chain{@NamedTuple{layer_1::Dense{typeof(tanh), Int64, Int64, Nothing, Nothing, Static.True}, layer_2::Dense{typeof(identity), Int64, Int64, Nothing, Nothing, Static.True}}, Nothing}((layer_1 = Dense(256 => 1, tanh), layer_2 = Dense(1 => 10)), nothing)), nothing), (layer_1 = (weight = Float32[-0.22542597 0.22379348 … 0.1997513 -0.018708104; -0.023026714 0.15451026 … -0.065325744 0.18120264; … ; 0.038037397 -0.07125516 … -0.03306083 0.039138064; -0.18810266 -0.09693537 … -0.18102062 0.019230088], bias = Float32[0.030937059, -0.060276944, 0.084569596, 0.00040024254, -0.065509446, -0.08527214, -0.026523968, 0.06347208, 0.042247728, 0.027705256  …  -0.06052852, 0.03504307, -0.028244259, 0.06788022, 0.0027464977, -0.06942153, 0.0064240773, 0.0141069945, -0.029283267, 0.01174226]), layer_2 = (layer_1 = (weight = Float32[0.12008221 0.06026435 … -0.070576 0.1577647], bias = Float32[0.026844418]), layer_2 = (weight = Float32[0.5345728; -0.28288874; … ; -0.32983455; -0.45298168;;], bias = Float32[-0.59751064, -0.7033041, -0.8457602, -0.53789175, -0.31473723, 0.17461234, -0.82945836, 0.67841595, 0.35837248, -0.14941788]))), (layer_1 = NamedTuple(), layer_2 = (layer_1 = NamedTuple(), layer_2 = NamedTuple())), Adam(eta=0.0001, beta=(0.9, 0.999), epsilon=1.0e-8), (layer_1 = (weight = Leaf(Adam(eta=0.0001, beta=(0.9, 0.999), epsilon=1.0e-8), (Float32[0.000926728 0.000860063 … 0.00110328 0.000908301; 0.000480834 0.000574605 … 0.000665883 0.000584197; … ; -0.000391039 -0.000438617 … -0.000520651 -0.000449867; 0.00106235 0.000365587 … 0.000813131 0.000495484], Float32[7.20343f-8 4.46976f-8 … 6.84867f-8 4.63952f-8; 1.79691f-8 2.02649f-8 … 2.45046f-8 1.96227f-8; … ; 1.21215f-8 1.17657f-8 … 1.50136f-8 1.15681f-8; 1.12738f-7 7.45199f-9 … 4.8495f-8 1.44173f-8], (0.729, 0.997003))), bias = Leaf(Adam(eta=0.0001, beta=(0.9, 0.999), epsilon=1.0e-8), (Float32[0.00169459, 0.000977637, 0.00103866, -0.00234933, 0.000659175, 0.000868318, 0.00303222, 0.00271383, -0.00326585, -0.0014993  …  -0.000480712, -0.000501535, -0.00174489, -0.00160158, -0.000470662, 0.00127967, 0.000618911, -0.00103705, -0.000773079, 0.00146704], Float32[1.74884f-7, 5.48983f-8, 7.75433f-8, 3.08981f-7, 2.45763f-8, 4.41623f-8, 5.29156f-7, 4.09021f-7, 6.07287f-7, 1.45678f-7  …  1.4164f-8, 1.73391f-8, 1.7507f-7, 1.44894f-7, 1.25673f-8, 1.1198f-7, 2.11545f-8, 6.25338f-8, 3.4755f-8, 1.78565f-7], (0.729, 0.997003)))), layer_2 = (layer_1 = (weight = Leaf(Adam(eta=0.0001, beta=(0.9, 0.999), epsilon=1.0e-8), (Float32[0.00443555 0.00163654 … -0.0124978 0.0123434], Float32[2.53181f-6 1.32838f-6 … 8.83289f-6 8.58873f-6], (0.729, 0.997003))), bias = Leaf(Adam(eta=0.0001, beta=(0.9, 0.999), epsilon=1.0e-8), (Float32[0.0191175], Float32[2.08743f-5], (0.729, 0.997003)))), layer_2 = (weight = Leaf(Adam(eta=0.0001, beta=(0.9, 0.999), epsilon=1.0e-8), (Float32[-0.0172084; -0.0213176; … ; -0.00376332; -0.0116419;;], Float32[1.63537f-5; 2.51152f-5; … ; 8.16783f-7; 7.55419f-6;;], (0.729, 0.997003))), bias = Leaf(Adam(eta=0.0001, beta=(0.9, 0.999), epsilon=1.0e-8), (Float32[-0.0365001, -0.045083, -0.0507623, -0.0390298, -0.0242259, -0.0404982, -0.0358925, 0.0114351, -0.00803444, -0.0248332], Float32[7.40417f-5, 0.000112652, 0.000146818, 8.41229f-5, 4.60234f-5, 9.15105f-5, 7.13093f-5, 8.78741f-6, 3.62043f-6, 3.51285f-5], (0.729, 0.997003)))))), 2))

Defining Custom Layers

We can train our model using the above code, but let's go ahead and see how to use Reactant. Reactant is a julia frontend that generates MLIR and then compiles it using XLA (after running fancy optimizations). It is the current recommended way to train large models in Lux. For more details on using Reactant, see the manual.

using Lux, Random, Optimisers, Reactant, Enzyme
using Printf # For pretty printing

dev = reactant_device()

(::ReactantDevice{Missing, Missing, Missing}) (generic function with 1 method)

We will define a custom MLP using the @compact macro. The macro takes in a list of parameters, layers and states, and a function defining the forward pass of the neural network.

n_in = 1
n_out = 1
nlayers = 3

model = @compact(
    w1=Dense(n_in => 32),
    w2=[Dense(32 => 32) for i in 1:nlayers],
    w3=Dense(32 => n_out),
    act=relu
) do x
    embed = act(w1(x))
    for w in w2
        embed = act(w(embed))
    end
    out = w3(embed)
    @return out
end

@compact(
    w1 = Dense(1 => 32),                # 64 parameters
    w2 = NamedTuple(
        1 = Dense(32 => 32),            # 1_056 parameters
        2 = Dense(32 => 32),            # 1_056 parameters
        3 = Dense(32 => 32),            # 1_056 parameters
    ),
    w3 = Dense(32 => 1),                # 33 parameters
    act = relu,
) do x 
    embed = act(w1(x))
    for w = w2
        embed = act(w(embed))
    end
    out = w3(embed)
    return out
end       # Total: 3_265 parameters,
          #        plus 1 states.

We can initialize the model and train it with the same code as before!

rng = Random.default_rng()
Random.seed!(rng, 0)

ps, st = Lux.setup(rng, model) |> dev

x = rand(rng, Float32, n_in, 32) |> dev

@jit model(x, ps, st)  # 1×32 Matrix and updated state as output.

x_data = reshape(collect(-2.0f0:0.1f0:2.0f0), 1, :)
y_data = 2 .* x_data .- x_data .^ 3
x_data, y_data = dev(x_data), dev(y_data)

function train_model!(model, ps, st, x_data, y_data)
    train_state = Lux.Training.TrainState(model, ps, st, Adam(0.001f0))

    for iter in 1:1000
        _, loss, _, train_state = Lux.Training.single_train_step!(
            AutoEnzyme(), MSELoss(),
            (x_data, y_data), train_state
        )
        if iter % 100 == 1 || iter == 1000
            @printf "Iteration: %04d \t Loss: %10.9g\n" iter loss
        end
    end

    return model, ps, st
end

train_model!(model, ps, st, x_data, y_data)

2025-07-03 16:27:12.007351: I external/xla/xla/service/service.cc:153] XLA service 0x14f68dc0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2025-07-03 16:27:12.007388: 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 00:00:1751560032.008245  296855 se_gpu_pjrt_client.cc:1370] Using BFC allocator.
I0000 00:00:1751560032.008322  296855 gpu_helpers.cc:136] XLA backend allocating 3825205248 bytes on device 0 for BFCAllocator.
I0000 00:00:1751560032.008373  296855 gpu_helpers.cc:177] XLA backend will use up to 1275068416 bytes on device 0 for CollectiveBFCAllocator.
I0000 00:00:1751560032.022332  296855 cuda_dnn.cc:471] Loaded cuDNN version 90800
2025-07-03 16:27:48.806938: W external/xla/xla/service/gpu/autotuning/dot_search_space.cc:200] All configs were filtered out because none of them sufficiently match the hints. Maybe the hints set does not contain a good representative set of valid configs?Working around this by using the full hints set instead.
2025-07-03 16:27:49.448971: I external/xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_21', 68 bytes spill stores, 68 bytes spill loads

2025-07-03 16:27:49.656395: I external/xla/xla/stream_executor/cuda/subprocess_compilation.cc:346] ptxas warning : Registers are spilled to local memory in function 'gemm_fusion_dot_21', 68 bytes spill stores, 68 bytes spill loads

E0000 00:00:1751560069.734983  296855 buffer_comparator.cc:147] Difference at 16: 0, expected 17.0497
E0000 00:00:1751560069.735737  296855 buffer_comparator.cc:147] Difference at 17: 0, expected 18.0586
E0000 00:00:1751560069.735741  296855 buffer_comparator.cc:147] Difference at 18: 0, expected 16.542
E0000 00:00:1751560069.735745  296855 buffer_comparator.cc:147] Difference at 19: 0, expected 16.819
E0000 00:00:1751560069.735748  296855 buffer_comparator.cc:147] Difference at 20: 0, expected 17.5866
E0000 00:00:1751560069.735751  296855 buffer_comparator.cc:147] Difference at 21: 0, expected 16.2892
E0000 00:00:1751560069.735754  296855 buffer_comparator.cc:147] Difference at 22: 0, expected 17.3707
E0000 00:00:1751560069.735757  296855 buffer_comparator.cc:147] Difference at 23: 0, expected 15.3674
E0000 00:00:1751560069.735760  296855 buffer_comparator.cc:147] Difference at 24: 0, expected 17.0118
E0000 00:00:1751560069.735763  296855 buffer_comparator.cc:147] Difference at 25: 0, expected 16.7842
2025-07-03 16:27:49.735773: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1037] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1751560069.737895  296855 buffer_comparator.cc:147] Difference at 16: 0, expected 17.0497
E0000 00:00:1751560069.737909  296855 buffer_comparator.cc:147] Difference at 17: 0, expected 18.0586
E0000 00:00:1751560069.737913  296855 buffer_comparator.cc:147] Difference at 18: 0, expected 16.542
E0000 00:00:1751560069.737916  296855 buffer_comparator.cc:147] Difference at 19: 0, expected 16.819
E0000 00:00:1751560069.737919  296855 buffer_comparator.cc:147] Difference at 20: 0, expected 17.5866
E0000 00:00:1751560069.737922  296855 buffer_comparator.cc:147] Difference at 21: 0, expected 16.2892
E0000 00:00:1751560069.737925  296855 buffer_comparator.cc:147] Difference at 22: 0, expected 17.3707
E0000 00:00:1751560069.737928  296855 buffer_comparator.cc:147] Difference at 23: 0, expected 15.3674
E0000 00:00:1751560069.737931  296855 buffer_comparator.cc:147] Difference at 24: 0, expected 17.0118
E0000 00:00:1751560069.737934  296855 buffer_comparator.cc:147] Difference at 25: 0, expected 16.7842
2025-07-03 16:27:49.737939: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1037] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1751560069.739929  296855 buffer_comparator.cc:147] Difference at 512: 0, expected 15.077
E0000 00:00:1751560069.739942  296855 buffer_comparator.cc:147] Difference at 513: 0, expected 14.8311
E0000 00:00:1751560069.739945  296855 buffer_comparator.cc:147] Difference at 514: 0, expected 13.6951
E0000 00:00:1751560069.739948  296855 buffer_comparator.cc:147] Difference at 515: 0, expected 16.2599
E0000 00:00:1751560069.739951  296855 buffer_comparator.cc:147] Difference at 516: 0, expected 14.5245
E0000 00:00:1751560069.739954  296855 buffer_comparator.cc:147] Difference at 517: 0, expected 14.1703
E0000 00:00:1751560069.739957  296855 buffer_comparator.cc:147] Difference at 518: 0, expected 14.8673
E0000 00:00:1751560069.739960  296855 buffer_comparator.cc:147] Difference at 519: 0, expected 16.2019
E0000 00:00:1751560069.739963  296855 buffer_comparator.cc:147] Difference at 520: 0, expected 16.0798
E0000 00:00:1751560069.739967  296855 buffer_comparator.cc:147] Difference at 521: 0, expected 14.494
2025-07-03 16:27:49.739972: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1037] Results do not match the reference. This is likely a bug/unexpected loss of precision.
Iteration: 0001 	 Loss: 2.08073235
Iteration: 0101 	 Loss: 0.142574623
Iteration: 0201 	 Loss: 0.0051055951
Iteration: 0301 	 Loss: 0.00118357129
Iteration: 0401 	 Loss: 0.000504208321
Iteration: 0501 	 Loss: 0.000281832268
Iteration: 0601 	 Loss: 0.000203011135
Iteration: 0701 	 Loss: 0.000126347542
Iteration: 0801 	 Loss: 0.00201115524
Iteration: 0901 	 Loss: 9.70276451e-05
Iteration: 1000 	 Loss: 7.81012277e-05

Training with Optimization.jl

If you are coming from the SciML ecosystem and want to use Optimization.jl, please refer to the Optimization.jl Tutorial.

Additional Packages

LuxDL hosts various packages that provide additional functionality for Lux.jl. All packages mentioned in this documentation are available via the Julia General Registry.

You can install all those packages via import Pkg; Pkg.add(<package name>).

XLA (CPU/GPU/TPU) Support

Lux.jl supports XLA compilation for CPU, GPU, and TPU using Reactant.jl.

GPU Support

GPU Support for Lux.jl requires loading additional packages:

• LuxCUDA.jl for CUDA support.

• AMDGPU.jl for AMDGPU support.

• Metal.jl for Apple Metal support.

• oneAPI.jl for oneAPI support.