Manual RNN on GPU

kailukowiak · April 16, 2022, 7:13pm

I’m trying to redo this tutorial in Flux. The neural net runs fine on a CPU but when I try transferring it to the GPU I get an error which is caused by indexing a CUDA array.

Here’s an MRE:

##
using Flux
using Flux: onehotbatch
using CUDA
using Parameters: @with_kw

#3##
alphabet = collect('a':'z')
##
name_ = "somename"
all_langs = ["option1", "option2"]
##
x = onehotbatch(name_, alphabet)
y = onehotbatch(["option1"], all_langs)

@with_kw mutable struct Model
    device::Function = CUDA.functional() ? gpu : cpu
    input_size::Int = length(alphabet)
    output_size::Int = length(all_langs)
    hidden_size::Int = length(alphabet)
    lr::Float64 = 3e-4
    epochs::Int = 20
    batch_size::Int = 128
    in2out = device(Chain(Dense(input_size + hidden_size, output_size), softmax))
    in2hidden = device(Dense(input_size + hidden_size, hidden_size))
    ps = params(in2out, in2hidden)
end
model = Model(device = cpu)

function rnn(input, hidden, model)
    combined = [input; hidden] |> model.device
    out = model.in2out(combined)
    hidden = model.in2hidden(combined)
    return out, hidden
end


function predict(X, model::Model)
    hidden = zeros(model.hidden_size) |> model.device
    local ŷ
    for i ∈ eachcol(X)
        ŷ, hidden = rnn(i, hidden, model) #  |> model.device
    end
    return ŷ
end
random_pred = predict(x, model);
println(random_pred)


model_gpu = Model(device = gpu)
x_gpu = x |> model_gpu.device
y_gpu = y |> model_gpu.device
random_pred_gpu = predict(x, model_gpu); # Works but with worning and fails if: `CUDA.allowscalar(false)`
random_pred_gpu

loss_fn(out, lab) = -log(out[lab.indices][1])
gradient(() -> loss_fn(predict(x_gpu, model_gpu), y_gpu)) # Fails

which gives the error:
ERROR: GPU compilation of kernel broadcast_kernel(CUDA.CuKernelContext, CuDeviceMatrix{Float32, 1}, Base.Broadcast.Broadcasted{CUDA.CuArrayStyle{2}, Tuple{Base.OneTo{Int64}, Base.OneTo{Int64}}, typeof(*), Tuple{Base.Broadcast.Extruded{Vector{Float32}, Tuple{Bool}, Tuple{Int64}}, Base.Broadcast.Extruded{LinearAlgebra.Adjoint{Float32, CuDeviceVector{Float32, 1}}, Tuple{Bool, Bool}, Tuple{Int64, Int64}}}}, Int64) failed KernelError: passing and using non-bitstype argument

I believe I could just use the built-in in RNN in Flux.jl however, that somewhat defeats the purpose of a from scratch tutorial.

What would be the best way to get around this error?

ToucheSir · April 16, 2022, 9:32pm

No good MWE is complete without a stacktrace (most of the time it’s more useful than the error message itself). If you add one, it would be much easier for someone to start troubleshooting.

kailukowiak · April 19, 2022, 1:29pm

Of course. Here it is:

julia> gradient(() -> loss_fn(predict(x_gpu, model_gpu), y_gpu))# Fails
ERROR: CUDA error: unspecified launch failure (code 719, ERROR_LAUNCH_FAILED)
Stacktrace:
  [1] throw_api_error(res::CUDA.cudaError_enum)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/error.jl:91
  [2] isdone
    @ ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/stream.jl:109 [inlined]
  [3] nonblocking_synchronize
    @ ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/stream.jl:139 [inlined]
  [4] (::CUDA.var"#217#218"{Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Int64, Vector{Float32}, Int64, Int64})()
    @ CUDA ~/.julia/packages/CUDA/5jdFl/src/array.jl:388
  [5] #context!#63
    @ ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/state.jl:164 [inlined]
  [6] context!
    @ ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/state.jl:161 [inlined]
  [7] unsafe_copyto!(dest::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, doffs::Int64, src::Vector{Float32}, soffs::Int64, n::Int64)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/src/array.jl:384
  [8] copyto!
    @ ~/.julia/packages/CUDA/5jdFl/src/array.jl:342 [inlined]
  [9] copyto!
    @ ~/.julia/packages/CUDA/5jdFl/src/array.jl:346 [inlined]
 [10] CuArray
    @ ~/.julia/packages/CUDA/5jdFl/src/array.jl:288 [inlined]
 [11] adjoint
    @ ~/.julia/packages/Zygote/H6vD3/src/lib/broadcast.jl:275 [inlined]
 [12] _pullback
    @ ~/.julia/packages/ZygoteRules/AIbCs/src/adjoint.jl:65 [inlined]
 [13] _pullback
    @ ~/.julia/packages/CUDA/5jdFl/src/array.jl:535 [inlined]
 [14] _pullback
    @ ~/.julia/packages/Adapt/wASZA/src/Adapt.jl:42 [inlined]
 [15] _pullback
    @ ~/.julia/packages/Adapt/wASZA/src/Adapt.jl:40 [inlined]
 [16] _pullback (repeats 2 times)
    @ ~/.julia/packages/CUDA/5jdFl/src/array.jl:545 [inlined]
 [17] _pullback
    @ ~/.julia/packages/Flux/qAdFM/src/functor.jl:97 [inlined]
 [18] _pullback
    @ ~/.julia/packages/Adapt/wASZA/src/Adapt.jl:42 [inlined]
 [19] _pullback
    @ ~/.julia/packages/Adapt/wASZA/src/Adapt.jl:40 [inlined]
 [20] _pullback
    @ ~/.julia/packages/Flux/qAdFM/src/functor.jl:177 [inlined]
 [21] _pullback
    @ ~/.julia/packages/Functors/qBIlC/src/functor.jl:50 [inlined]
 [22] _pullback(::Zygote.Context, ::Functors.var"##fmap#17", ::typeof(Flux._isbitsarray), ::typeof(Functors._default_walk), ::IdDict{Any, Any}, ::Functors.NoKeyword, ::typeof(fmap), ::Flux.var"#138#139", ::Vector{Float64})
    @ Zygote ~/.julia/packages/Zygote/H6vD3/src/compiler/interface2.jl:0
 [23] _pullback
    @ ~/.julia/packages/Functors/qBIlC/src/functor.jl:49 [inlined]
 [24] _pullback(::Zygote.Context, ::Functors.var"#fmap##kw", ::WARNING: Error while freeing DeviceBuffer(104 bytes at 0x0000000302006c00):
CUDA.CuError(code=CUDA.cudaError_enum(0x000002cf), meta=nothing)

Stacktrace:
  [1] throw_api_error(res::CUDA.cudaError_enum)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/error.jl:91
  [2] macro expansion
    @ ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/error.jl:101 [inlined]
  [3] cuMemFreeAsync(dptr::CUDA.Mem.DeviceBuffer, hStream::CuStream)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/lib/utils/call.jl:26
  [4] #free#2
    @ ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/memory.jl:97 [inlined]
  [5] macro expansion
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:58 [inlined]
  [6] macro expansion
    @ ./timing.jl:299 [inlined]
  [7] #actual_free#189
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:57 [inlined]
  [8] #_free#207
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:375 [inlined]
  [9] macro expansion
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:340 [inlined]
 [10] macro expansion
    @ ./timing.jl:299 [inlined]
 [11] #free#206
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:339 [inlined]
 [12] #212
    @ ~/.julia/packages/CUDA/5jdFl/src/array.jl:79 [inlined]
 [13] context!(f::CUDA.var"#212#213"{CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuStream}, ctx::CuContext; skip_destroyed::Bool)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/state.jl:164
 [14] unsafe_free!(xs::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, stream::CuStream)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/src/array.jl:78
 [15] unsafe_finalize!(xs::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer})
    @ CUDA ~/.julia/packages/CUDA/5jdFl/src/array.jl:99
 [16] show_typeparams(io::IOContext{IOBuffer}, env::Core.SimpleVector, orig::Core.SimpleVector, wheres::Vector{TypeVar})
    @ Base ./show.jl:640
 [17] show_datatype(io::IOContext{IOBuffer}, x::DataType, wheres::Vector{TypeVar})
    @ Base ./show.jl:1011
 [18] show_datatype
    @ ./show.jl:989 [inlined]
 [19] _show_type(io::IOContext{IOBuffer}, x::Type)
    @ Base ./show.jl:889
 [20] show(io::IOContext{IOBuffer}, x::Type)
    @ Base ./show.jl:881
 [21] sprint(f::Function, args::Type; context::IOContext{Base.TTY}, sizehint::Int64)
    @ Base ./strings/io.jl:112
 [22] print_type_stacktrace(io::IOContext{Base.TTY}, type::Type; color::Symbol)
    @ Base ./show.jl:2399
 [23] print_type_stacktrace(io::IOContext{Base.TTY}, type::Type)
    @ Base ./show.jl:2399
 [24] show_tuple_as_call(io::IOContext{Base.TTY}, name::Symbol, sig::Type; demangle::Bool, kwargs::Nothing, argnames::Vector{Symbol}, qualified::Bool, hasfirst::Bool)
    @ Base ./show.jl:2380
 [25] show_spec_linfo(io::IOContext{Base.TTY}, frame::Base.StackTraces.StackFrame)
    @ Base.StackTraces ./stacktraces.jl:244
 [26] print_stackframe(io::IOContext{Base.TTY}, i::Int64, frame::Base.StackTraces.StackFrame, n::Int64, digit_align_width::Int64, modulecolor::Symbol)
    @ Base ./errorshow.jl:709
 [27] print_stackframe(io::IOContext{Base.TTY}, i::Int64, frame::Base.StackTraces.StackFrame, n::Int64, digit_align_width::Int64, modulecolordict::IdDict{Module, Symbol}, modulecolorcycler::Base.Iterators.Stateful{Base.Iterators.Cycle{Vector{Symbol}}, Union{Nothing, Tuple{Symbol, Int64}}})
    @ Base ./errorshow.jl:685
 [28] show_full_backtrace(io::IOContext{Base.TTY}, trace::Vector{Any}; print_linebreaks::Bool)
    @ Base ./errorshow.jl:574
 [29] show_backtrace(io::IOContext{Base.TTY}, t::Vector{Base.StackTraces.StackFrame})
    @ Base ./errorshow.jl:769
 [30] showerror(io::IOContext{Base.TTY}, ex::CuError, bt::Vector{Base.StackTraces.StackFrame}; backtrace::Bool)
    @ Base ./errorshow.jl:90
 [31] showerror(io::IOContext{Base.TTY}, ex::CuError, bt::Vector{Base.StackTraces.StackFrame})
    @ Base ./errorshow.jl:87
 [32] display_repl_error(io::Base.TTY, stack::VSCodeServer.EvalErrorStack)
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/repl.jl:173
 [33] #invokelatest#2
    @ ./essentials.jl:716 [inlined]
 [34] invokelatest
    @ ./essentials.jl:714 [inlined]
 [35] display_error(io::Base.TTY, err::VSCodeServer.EvalErrorStack)
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:319
 [36] (::VSCodeServer.var"#65#69"{Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:170
 [37] withpath(f::VSCodeServer.var"#65#69"{Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams}, path::String)
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/repl.jl:184
 [38] (::VSCodeServer.var"#64#68"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:153
 [39] hideprompt(f::VSCodeServer.var"#64#68"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/repl.jl:36
 [40] (::VSCodeServer.var"#63#67"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:124
 [41] with_logstate(f::Function, logstate::Any)
    @ Base.CoreLogging ./logging.jl:511
 [42] with_logger
    @ ./logging.jl:623 [inlined]
 [43] (::VSCodeServer.var"#62#66"{VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:201
 [44] #invokelatest#2
    @ ./essentials.jl:716 [inlined]
 [45] invokelatest(::Any)
    @ Base ./essentials.jl:714
 [46] macro expansion
    @ ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:34 [inlined]
 [47] (::VSCodeServer.var"#60#61")()
    @ VSCodeServer ./task.jl:423
WARNING: Error while freeing DeviceBuffer(104 bytes at 0x0000000302006000):
CUDA.CuError(code=CUDA.cudaError_enum(0x000002cf), meta=nothing)

Stacktrace:
  [1] throw_api_error(res::CUDA.cudaError_enum)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/error.jl:91
  [2] macro expansion
    @ ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/error.jl:101 [inlined]
  [3] cuMemFreeAsync(dptr::CUDA.Mem.DeviceBuffer, hStream::CuStream)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/lib/utils/call.jl:26
  [4] #free#2
    @ ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/memory.jl:97 [inlined]
  [5] macro expansion
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:58 [inlined]
  [6] macro expansion
    @ ./timing.jl:299 [inlined]
  [7] #actual_free#189
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:57 [inlined]
  [8] #_free#207
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:375 [inlined]
  [9] macro expansion
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:340 [inlined]
 [10] macro expansion
    @ ./timing.jl:299 [inlined]
 [11] #free#206
    @ ~/.julia/packages/CUDA/5jdFl/src/pool.jl:339 [inlined]
 [12] #212
    @ ~/.julia/packages/CUDA/5jdFl/src/array.jl:79 [inlined]
 [13] context!(f::CUDA.var"#212#213"{CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuStream}, ctx::CuContext; skip_destroyed::Bool)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/lib/cudadrv/state.jl:164
 [14] unsafe_free!(xs::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, stream::CuStream)
    @ CUDA ~/.julia/packages/CUDA/5jdFl/src/array.jl:78
 [15] unsafe_finalize!(xs::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer})
    @ CUDA ~/.julia/packages/CUDA/5jdFl/src/array.jl:99
 [16] show_typeparams(io::IOContext{IOBuffer}, env::Core.SimpleVector, orig::Core.SimpleVector, wheres::Vector{TypeVar})
    @ Base ./show.jl:640
 [17] show_datatype(io::IOContext{IOBuffer}, x::DataType, wheres::Vector{TypeVar})
    @ Base ./show.jl:1011
 [18] show_datatype
    @ ./show.jl:989 [inlined]
 [19] _show_type(io::IOContext{IOBuffer}, x::Type)
    @ Base ./show.jl:889
 [20] show(io::IOContext{IOBuffer}, x::Type)
    @ Base ./show.jl:881
 [21] sprint(f::Function, args::Type; context::IOContext{Base.TTY}, sizehint::Int64)
    @ Base ./strings/io.jl:112
 [22] print_type_stacktrace(io::IOContext{Base.TTY}, type::Type; color::Symbol)
    @ Base ./show.jl:2399
 [23] print_type_stacktrace(io::IOContext{Base.TTY}, type::Type)
    @ Base ./show.jl:2399
 [24] show_tuple_as_call(io::IOContext{Base.TTY}, name::Symbol, sig::Type; demangle::Bool, kwargs::Nothing, argnames::Vector{Symbol}, qualified::Bool, hasfirst::Bool)
    @ Base ./show.jl:2380
 [25] show_spec_linfo(io::IOContext{Base.TTY}, frame::Base.StackTraces.StackFrame)
    @ Base.StackTraces ./stacktraces.jl:244
 [26] print_stackframe(io::IOContext{Base.TTY}, i::Int64, frame::Base.StackTraces.StackFrame, n::Int64, digit_align_width::Int64, modulecolor::Symbol)
    @ Base ./errorshow.jl:709
 [27] print_stackframe(io::IOContext{Base.TTY}, i::Int64, frame::Base.StackTraces.StackFrame, n::Int64, digit_align_width::Int64, modulecolordict::IdDict{Module, Symbol}, modulecolorcycler::Base.Iterators.Stateful{Base.Iterators.Cycle{Vector{Symbol}}, Union{Nothing, Tuple{Symbol, Int64}}})
    @ Base ./errorshow.jl:685
 [28] show_full_backtrace(io::IOContext{Base.TTY}, trace::Vector{Any}; print_linebreaks::Bool)
    @ Base ./errorshow.jl:574
 [29] show_backtrace(io::IOContext{Base.TTY}, t::Vector{Base.StackTraces.StackFrame})
    @ Base ./errorshow.jl:769
 [30] showerror(io::IOContext{Base.TTY}, ex::CuError, bt::Vector{Base.StackTraces.StackFrame}; backtrace::Bool)
    @ Base ./errorshow.jl:90
 [31] showerror(io::IOContext{Base.TTY}, ex::CuError, bt::Vector{Base.StackTraces.StackFrame})
    @ Base ./errorshow.jl:87
 [32] display_repl_error(io::Base.TTY, stack::VSCodeServer.EvalErrorStack)
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/repl.jl:173
 [33] #invokelatest#2
    @ ./essentials.jl:716 [inlined]
 [34] invokelatest
    @ ./essentials.jl:714 [inlined]
 [35] display_error(io::Base.TTY, err::VSCodeServer.EvalErrorStack)
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:319
 [36] (::VSCodeServer.var"#65#69"{Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:170
 [37] withpath(f::VSCodeServer.var"#65#69"{Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams}, path::String)
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/repl.jl:184
 [38] (::VSCodeServer.var"#64#68"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:153
 [39] hideprompt(f::VSCodeServer.var"#64#68"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/repl.jl:36
 [40] (::VSCodeServer.var"#63#67"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:124
 [41] with_logstate(f::Function, logstate::Any)
    @ Base.CoreLogging ./logging.jl:511
 [42] with_logger
    @ ./logging.jl:623 [inlined]
 [43] (::VSCodeServer.var"#62#66"{VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:201
 [44] #invokelatest#2
    @ ./essentials.jl:716 [inlined]
 [45] invokelatest(::Any)
    @ Base ./essentials.jl:714
 [46] macro expansion
    @ ~/.vscode/extensions/julialang.language-julia-1.6.15/scripts/packages/VSCodeServer/src/eval.jl:34 [inlined]
 [47] (::VSCodeServer.var"#60#61")()
    @ VSCodeServer ./task.jl:423
NamedTuple{(:exclude,), Tuple{typeof(Flux._isbitsarray)}}, ::typeof(fmap), ::Flux.var"#138#139", ::Vector{Float64})
    @ Zygote ~/.julia/packages/Zygote/H6vD3/src/compiler/interface2.jl:0
 [25] _pullback
    @ ~/.julia/packages/Flux/qAdFM/src/functor.jl:177 [inlined]
 [26] _pullback
    @ ./operators.jl:966 [inlined]
 [27] _pullback(::Zygote.Context, ::typeof(|>), ::Vector{Float64}, ::typeof(gpu))
    @ Zygote ~/.julia/packages/Zygote/H6vD3/src/compiler/interface2.jl:0
 [28] _pullback
    @ ~/Documents/julia_tutorial/minimal_repro_ex.jl:39 [inlined]
 [29] _pullback(::Zygote.Context, ::typeof(predict), ::Flux.OneHotArray{UInt32, 26, 1, 2, CuArray{UInt32, 1, CUDA.Mem.DeviceBuffer}}, ::Model)
    @ Zygote ~/.julia/packages/Zygote/H6vD3/src/compiler/interface2.jl:0
 [30] _pullback
    @ ~/Documents/julia_tutorial/minimal_repro_ex.jl:57 [inlined]
 [31] _pullback(::Zygote.Context, ::var"#12#13")
    @ Zygote ~/.julia/packages/Zygote/H6vD3/src/compiler/interface2.jl:0
 [32] _pullback(::Function)
    @ Zygote ~/.julia/packages/Zygote/H6vD3/src/compiler/interface.jl:34
 [33] pullback(::Function)
    @ Zygote ~/.julia/packages/Zygote/H6vD3/src/compiler/interface.jl:40
 [34] gradient(::Function)
    @ Zygote ~/.julia/packages/Zygote/H6vD3/src/compiler/interface.jl:75
 [35] top-level scope
    @ ~/Documents/julia_tutorial/minimal_repro_ex.jl:57

ToucheSir · April 20, 2022, 12:29am

If I’m not mistaken, that’s a completely different error from the one in the OP. What happens if you run the code outside of VSCode?

Topic		Replies	Views
Flux RNN on a GPU - unnecessary copying Machine Learning	0	451	June 12, 2019
Uploading vector of vectors to GPU in flux.jl Machine Learning question	4	613	October 1, 2021
Flux on GPU too slow Machine Learning gpu , cuda , flux	5	1068	September 22, 2022
Parallel feedforward computation with RNN Machine Learning flux	0	422	July 15, 2019
Is it possible to train sparse neural networks in GPU? Machine Learning question , gpu , cuda , flux	3	500	June 17, 2022

Manual RNN on GPU

Related topics