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
julia
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"])
julia
using Lux, Random, Optimisers, Zygote
# using LuxCUDA, AMDGPU, Metal, oneAPI # Optional packages for GPU supportWe take randomness very seriously
julia
# Seeding
rng = Random.default_rng()
Random.seed!(rng, 0)Random.TaskLocalRNG()Build the model
julia
# 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.
julia
# 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.
julia
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.
julia
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!
julia
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)2025-03-17 03:05:50.759451: I external/xla/xla/service/service.cc:152] XLA service 0xce707d0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2025-03-17 03:05:50.760041: I external/xla/xla/service/service.cc:160] 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 00:00:1742180750.762183 465569 se_gpu_pjrt_client.cc:951] Using BFC allocator.
I0000 00:00:1742180750.762344 465569 gpu_helpers.cc:136] XLA backend allocating 3825205248 bytes on device 0 for BFCAllocator.
I0000 00:00:1742180750.762683 465569 gpu_helpers.cc:177] XLA backend will use up to 1275068416 bytes on device 0 for CollectiveBFCAllocator.
I0000 00:00:1742180750.778564 465569 cuda_dnn.cc:529] Loaded cuDNN version 90400
E0000 00:00:1742180932.949725 465569 buffer_comparator.cc:156] Difference at 16: 0, expected 18.4532
E0000 00:00:1742180932.950780 465569 buffer_comparator.cc:156] Difference at 17: 0, expected 16.1701
E0000 00:00:1742180932.950794 465569 buffer_comparator.cc:156] Difference at 18: 0, expected 18.5372
E0000 00:00:1742180932.950802 465569 buffer_comparator.cc:156] Difference at 19: 0, expected 17.7684
E0000 00:00:1742180932.950809 465569 buffer_comparator.cc:156] Difference at 20: 0, expected 17.8078
E0000 00:00:1742180932.950816 465569 buffer_comparator.cc:156] Difference at 21: 0, expected 17.412
E0000 00:00:1742180932.950823 465569 buffer_comparator.cc:156] Difference at 22: 0, expected 18.0425
E0000 00:00:1742180932.950830 465569 buffer_comparator.cc:156] Difference at 23: 0, expected 17.7822
E0000 00:00:1742180932.950836 465569 buffer_comparator.cc:156] Difference at 24: 0, expected 16.8692
E0000 00:00:1742180932.950843 465569 buffer_comparator.cc:156] Difference at 25: 0, expected 19.6248
2025-03-17 03:08:52.950857: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.954155 465569 buffer_comparator.cc:156] Difference at 16: 0, expected 18.4532
E0000 00:00:1742180932.954183 465569 buffer_comparator.cc:156] Difference at 17: 0, expected 16.1701
E0000 00:00:1742180932.954192 465569 buffer_comparator.cc:156] Difference at 18: 0, expected 18.5372
E0000 00:00:1742180932.954199 465569 buffer_comparator.cc:156] Difference at 19: 0, expected 17.7684
E0000 00:00:1742180932.954206 465569 buffer_comparator.cc:156] Difference at 20: 0, expected 17.8078
E0000 00:00:1742180932.954212 465569 buffer_comparator.cc:156] Difference at 21: 0, expected 17.412
E0000 00:00:1742180932.954219 465569 buffer_comparator.cc:156] Difference at 22: 0, expected 18.0425
E0000 00:00:1742180932.954226 465569 buffer_comparator.cc:156] Difference at 23: 0, expected 17.7822
E0000 00:00:1742180932.954233 465569 buffer_comparator.cc:156] Difference at 24: 0, expected 16.8692
E0000 00:00:1742180932.954239 465569 buffer_comparator.cc:156] Difference at 25: 0, expected 19.6248
2025-03-17 03:08:52.954250: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.957377 465569 buffer_comparator.cc:156] Difference at 656: 0, expected 15.8892
E0000 00:00:1742180932.957389 465569 buffer_comparator.cc:156] Difference at 657: 0, expected 15.1292
E0000 00:00:1742180932.957393 465569 buffer_comparator.cc:156] Difference at 658: 0, expected 14.0499
E0000 00:00:1742180932.957396 465569 buffer_comparator.cc:156] Difference at 659: 0, expected 13.8377
E0000 00:00:1742180932.957399 465569 buffer_comparator.cc:156] Difference at 660: 0, expected 13.7353
E0000 00:00:1742180932.957402 465569 buffer_comparator.cc:156] Difference at 661: 0, expected 15.7468
E0000 00:00:1742180932.957405 465569 buffer_comparator.cc:156] Difference at 662: 0, expected 14.9101
E0000 00:00:1742180932.957408 465569 buffer_comparator.cc:156] Difference at 663: 0, expected 14.8135
E0000 00:00:1742180932.957411 465569 buffer_comparator.cc:156] Difference at 664: 0, expected 13.6403
E0000 00:00:1742180932.957414 465569 buffer_comparator.cc:156] Difference at 665: 0, expected 15.8348
2025-03-17 03:08:52.957419: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.960361 465569 buffer_comparator.cc:156] Difference at 672: 0, expected 16.0696
E0000 00:00:1742180932.960373 465569 buffer_comparator.cc:156] Difference at 673: 0, expected 14.3019
E0000 00:00:1742180932.960377 465569 buffer_comparator.cc:156] Difference at 674: 0, expected 15.5573
E0000 00:00:1742180932.960380 465569 buffer_comparator.cc:156] Difference at 675: 0, expected 14.6242
E0000 00:00:1742180932.960383 465569 buffer_comparator.cc:156] Difference at 676: 0, expected 14.8486
E0000 00:00:1742180932.960388 465569 buffer_comparator.cc:156] Difference at 677: 0, expected 14.7699
E0000 00:00:1742180932.960391 465569 buffer_comparator.cc:156] Difference at 678: 0, expected 15.1617
E0000 00:00:1742180932.960394 465569 buffer_comparator.cc:156] Difference at 679: 0, expected 14.9394
E0000 00:00:1742180932.960397 465569 buffer_comparator.cc:156] Difference at 680: 0, expected 13.4678
E0000 00:00:1742180932.960400 465569 buffer_comparator.cc:156] Difference at 681: 0, expected 16.1851
2025-03-17 03:08:52.960404: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.963334 465569 buffer_comparator.cc:156] Difference at 688: 0, expected 15.1187
E0000 00:00:1742180932.963346 465569 buffer_comparator.cc:156] Difference at 689: 0, expected 14.6251
E0000 00:00:1742180932.963350 465569 buffer_comparator.cc:156] Difference at 690: 0, expected 14.2005
E0000 00:00:1742180932.963353 465569 buffer_comparator.cc:156] Difference at 691: 0, expected 15.1561
E0000 00:00:1742180932.963357 465569 buffer_comparator.cc:156] Difference at 692: 0, expected 15.4235
E0000 00:00:1742180932.963360 465569 buffer_comparator.cc:156] Difference at 693: 0, expected 14.1331
E0000 00:00:1742180932.963363 465569 buffer_comparator.cc:156] Difference at 694: 0, expected 14.4063
E0000 00:00:1742180932.963366 465569 buffer_comparator.cc:156] Difference at 695: 0, expected 14.0259
E0000 00:00:1742180932.963369 465569 buffer_comparator.cc:156] Difference at 696: 0, expected 15.0279
E0000 00:00:1742180932.963372 465569 buffer_comparator.cc:156] Difference at 729: 0, expected 14.5946
2025-03-17 03:08:52.963376: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.966311 465569 buffer_comparator.cc:156] Difference at 688: 0, expected 15.1187
E0000 00:00:1742180932.966323 465569 buffer_comparator.cc:156] Difference at 689: 0, expected 14.6251
E0000 00:00:1742180932.966326 465569 buffer_comparator.cc:156] Difference at 690: 0, expected 14.2005
E0000 00:00:1742180932.966330 465569 buffer_comparator.cc:156] Difference at 691: 0, expected 15.1561
E0000 00:00:1742180932.966333 465569 buffer_comparator.cc:156] Difference at 692: 0, expected 15.4235
E0000 00:00:1742180932.966336 465569 buffer_comparator.cc:156] Difference at 693: 0, expected 14.1331
E0000 00:00:1742180932.966339 465569 buffer_comparator.cc:156] Difference at 694: 0, expected 14.4063
E0000 00:00:1742180932.966342 465569 buffer_comparator.cc:156] Difference at 695: 0, expected 14.0259
E0000 00:00:1742180932.966345 465569 buffer_comparator.cc:156] Difference at 696: 0, expected 15.0279
E0000 00:00:1742180932.966348 465569 buffer_comparator.cc:156] Difference at 729: 0, expected 14.5946
2025-03-17 03:08:52.966353: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.969281 465569 buffer_comparator.cc:156] Difference at 688: 0, expected 15.1187
E0000 00:00:1742180932.969293 465569 buffer_comparator.cc:156] Difference at 689: 0, expected 14.6251
E0000 00:00:1742180932.969297 465569 buffer_comparator.cc:156] Difference at 690: 0, expected 14.2005
E0000 00:00:1742180932.969300 465569 buffer_comparator.cc:156] Difference at 691: 0, expected 15.1561
E0000 00:00:1742180932.969303 465569 buffer_comparator.cc:156] Difference at 692: 0, expected 15.4235
E0000 00:00:1742180932.969306 465569 buffer_comparator.cc:156] Difference at 693: 0, expected 14.1331
E0000 00:00:1742180932.969309 465569 buffer_comparator.cc:156] Difference at 694: 0, expected 14.4063
E0000 00:00:1742180932.969312 465569 buffer_comparator.cc:156] Difference at 695: 0, expected 14.0259
E0000 00:00:1742180932.969315 465569 buffer_comparator.cc:156] Difference at 696: 0, expected 15.0279
E0000 00:00:1742180932.969320 465569 buffer_comparator.cc:156] Difference at 729: 0, expected 14.5946
2025-03-17 03:08:52.969325: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.981869 465569 buffer_comparator.cc:156] Difference at 16: -nan, expected 29.4863
E0000 00:00:1742180932.981881 465569 buffer_comparator.cc:156] Difference at 17: -nan, expected 25.4275
E0000 00:00:1742180932.981885 465569 buffer_comparator.cc:156] Difference at 18: -nan, expected 29.498
E0000 00:00:1742180932.981888 465569 buffer_comparator.cc:156] Difference at 19: -nan, expected 24.9024
E0000 00:00:1742180932.981891 465569 buffer_comparator.cc:156] Difference at 20: -nan, expected 31.8883
E0000 00:00:1742180932.981893 465569 buffer_comparator.cc:156] Difference at 21: -nan, expected 30.5795
E0000 00:00:1742180932.981896 465569 buffer_comparator.cc:156] Difference at 22: -nan, expected 26.1755
E0000 00:00:1742180932.981899 465569 buffer_comparator.cc:156] Difference at 23: -nan, expected 30.0282
E0000 00:00:1742180932.981902 465569 buffer_comparator.cc:156] Difference at 24: -nan, expected 25.7237
E0000 00:00:1742180932.981905 465569 buffer_comparator.cc:156] Difference at 25: -nan, expected 25.7191
2025-03-17 03:08:52.981910: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.984832 465569 buffer_comparator.cc:156] Difference at 16: -nan, expected 29.4863
E0000 00:00:1742180932.984843 465569 buffer_comparator.cc:156] Difference at 17: -nan, expected 25.4275
E0000 00:00:1742180932.984847 465569 buffer_comparator.cc:156] Difference at 18: -nan, expected 29.498
E0000 00:00:1742180932.984850 465569 buffer_comparator.cc:156] Difference at 19: -nan, expected 24.9024
E0000 00:00:1742180932.984853 465569 buffer_comparator.cc:156] Difference at 20: -nan, expected 31.8883
E0000 00:00:1742180932.984856 465569 buffer_comparator.cc:156] Difference at 21: -nan, expected 30.5795
E0000 00:00:1742180932.984859 465569 buffer_comparator.cc:156] Difference at 22: -nan, expected 26.1755
E0000 00:00:1742180932.984862 465569 buffer_comparator.cc:156] Difference at 23: -nan, expected 30.0282
E0000 00:00:1742180932.984864 465569 buffer_comparator.cc:156] Difference at 24: -nan, expected 25.7237
E0000 00:00:1742180932.984867 465569 buffer_comparator.cc:156] Difference at 25: -nan, expected 25.7191
2025-03-17 03:08:52.984872: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.987795 465569 buffer_comparator.cc:156] Difference at 512: -nan, expected 13.9275
E0000 00:00:1742180932.987807 465569 buffer_comparator.cc:156] Difference at 513: -nan, expected 12.9447
E0000 00:00:1742180932.987810 465569 buffer_comparator.cc:156] Difference at 514: -nan, expected 13.899
E0000 00:00:1742180932.987813 465569 buffer_comparator.cc:156] Difference at 515: -nan, expected 14.1578
E0000 00:00:1742180932.987816 465569 buffer_comparator.cc:156] Difference at 516: -nan, expected 15.4892
E0000 00:00:1742180932.987819 465569 buffer_comparator.cc:156] Difference at 517: -nan, expected 16.545
E0000 00:00:1742180932.987822 465569 buffer_comparator.cc:156] Difference at 518: -nan, expected 17.8581
E0000 00:00:1742180932.987825 465569 buffer_comparator.cc:156] Difference at 519: -nan, expected 13.0536
E0000 00:00:1742180932.987827 465569 buffer_comparator.cc:156] Difference at 520: -nan, expected 16.1329
E0000 00:00:1742180932.987830 465569 buffer_comparator.cc:156] Difference at 521: -nan, expected 14.5245
2025-03-17 03:08:52.987835: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
E0000 00:00:1742180932.990769 465569 buffer_comparator.cc:156] Difference at 528: -nan, expected 17.5032
E0000 00:00:1742180932.990780 465569 buffer_comparator.cc:156] Difference at 529: -nan, expected 15.1785
E0000 00:00:1742180932.990784 465569 buffer_comparator.cc:156] Difference at 530: -nan, expected 15.9473
E0000 00:00:1742180932.990787 465569 buffer_comparator.cc:156] Difference at 531: -nan, expected 14.437
E0000 00:00:1742180932.990790 465569 buffer_comparator.cc:156] Difference at 532: -nan, expected 17.9637
E0000 00:00:1742180932.990793 465569 buffer_comparator.cc:156] Difference at 533: -nan, expected 17.3157
E0000 00:00:1742180932.990795 465569 buffer_comparator.cc:156] Difference at 534: -nan, expected 15.7802
E0000 00:00:1742180932.990798 465569 buffer_comparator.cc:156] Difference at 535: -nan, expected 17.6887
E0000 00:00:1742180932.990801 465569 buffer_comparator.cc:156] Difference at 536: -nan, expected 15.1881
E0000 00:00:1742180932.990804 465569 buffer_comparator.cc:156] Difference at 537: -nan, expected 14.4224
2025-03-17 03:08:52.990809: E external/xla/xla/service/gpu/autotuning/gemm_fusion_autotuner.cc:1138] Results do not match the reference. This is likely a bug/unexpected loss of precision.
Iteration: 0001 Loss: 2.08086824
Iteration: 0101 Loss: 0.135109991
Iteration: 0201 Loss: 0.00448962208
Iteration: 0301 Loss: 0.00111342408
Iteration: 0401 Loss: 0.000457020855
Iteration: 0501 Loss: 0.000461334654
Iteration: 0601 Loss: 0.000210383674
Iteration: 0701 Loss: 0.000209954742
Iteration: 0801 Loss: 0.000158460374
Iteration: 0901 Loss: 9.63655111e-05
Iteration: 1000 Loss: 0.000503897725Training 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.jlfor CUDA support.AMDGPU.jlfor AMDGPU support.Metal.jlfor Apple Metal support.oneAPI.jlfor oneAPI support.