CUDA: add support for onehot #2203

yolhan83 · 2024-12-16T20:46:18Z

Is Enzyme supposed to differentiate through KA kernels ? I wanted to try it on this simple exemple and I get an error GPU compilation of MethodInstance for (::GPUArrays.var"#34#36")(::CUDA.CuKernelContext, ::CuDeviceVector{…}, ::Base.Broadcast.Broadcasted{…}, ::Int64) failed KernelError: passing and using non-bitstype argument

using KernelAbstractions,CUDA
using DifferentiationInterface
import Enzyme
bck = AutoEnzyme(mode=Enzyme.Reverse)
uin() = 0.0f0
uout() = 0.0f0
@kernel inbounds=true function diffkernel!(du, u)
    i = @index(Global, Linear)
    if i == 1
        ug = uin()
        ud = u[i+1]
    elseif i == length(u)
        ug = u[i-1]
        ud = uout()
    else
        ug = u[i-1]
        ud = u[i+1] 
    end
    du[i] = ug + ud -2*u[i]
end
function Diffusion(u)
    bck_kernel = get_backend(u)
    du = zero(u)
    ker = diffkernel!(bck_kernel)
    ker(du, u, ndrange=length(du))
    return du
end
u = CUDA.rand(1000)
jacobian(Diffusion, bck,u)

The text was updated successfully, but these errors were encountered:

yolhan83 · 2024-12-16T20:47:10Z

Also, in forward, I get the error Scalar indexing is disallowed. Invocation of setindex! resulted in scalar indexing of a GPU array. This is typically caused by calling an iterating implementation of a method. Such implementations *do not* execute on the GPU, but very slowly on the CPU, and therefore should be avoided.

wsmoses · 2024-12-16T20:48:05Z

It does explicitly support KA.

Can you post a MWE without DifferentiationInterface?

yolhan83 · 2024-12-16T21:16:44Z

Yes, here it is

using KernelAbstractions,CUDA
import Enzyme
uin() = 0.0f0
uout() = 0.0f0
@kernel inbounds=true function diffkernel!(du, u)
    i = @index(Global, Linear)
    if i == 1
        ug = uin()
        ud = u[i+1]
    elseif i == length(u)
        ug = u[i-1]
        ud = uout()
    else
        ug = u[i-1]
        ud = u[i+1] 
    end
    du[i] = ug + ud -2*u[i]
end
function Diffusion(u)
    bck_kernel = get_backend(u)
    du = zero(u)
    ker = diffkernel!(bck_kernel)
    ker(du, u, ndrange=length(du))
    return du
end
u = CUDA.rand(1000)
Enzyme.jacobian(Enzyme.Reverse, Diffusion, u)
Enzyme.jacobian(Enzyme.Forward, Diffusion, u)

Same errors (takes a lot more times to throw for some reason). I don't remember how to make a jacobian from autodiff which is why I stayed on high level jacobian.

yolhan83 · 2024-12-16T21:20:44Z

full error :

forward

ERROR: Scalar indexing is disallowed.
Invocation of setindex! resulted in scalar indexing of a GPU array.
This is typically caused by calling an iterating implementation of a method.
Such implementations *do not* execute on the GPU, but very slowly on the CPU,
and therefore should be avoided.

If you want to allow scalar iteration, use `allowscalar` or `@allowscalar`
to enable scalar iteration globally or for the operations in question.
Stacktrace:
  [1] error(s::String)
    @ Base .\error.jl:35
  [2] errorscalar(op::String)
    @ GPUArraysCore C:\Users\yolha\.julia\packages\GPUArraysCore\GMsgk\src\GPUArraysCore.jl:155
  [3] _assertscalar(op::String, behavior::GPUArraysCore.ScalarIndexing)
    @ GPUArraysCore C:\Users\yolha\.julia\packages\GPUArraysCore\GMsgk\src\GPUArraysCore.jl:128
  [4] assertscalar(op::String)
    @ GPUArraysCore C:\Users\yolha\.julia\packages\GPUArraysCore\GMsgk\src\GPUArraysCore.jl:116
  [5] setindex!
    @ C:\Users\yolha\.julia\packages\GPUArrays\qt4ax\src\host\indexing.jl:58 [inlined]
  [6] #110
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:137 [inlined]
  [7] macro expansion
    @ .\ntuple.jl:72 [inlined]
  [8] ntuple(f::Enzyme.var"#110#111"{CuArray{Float32, 1, CUDA.DeviceMemory}, Int64}, ::Val{1000})
    @ Base .\ntuple.jl:69
  [9] onehot
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:133 [inlined]
 [10] macro expansion
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:440 [inlined]
 [11] create_shadows
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:418 [inlined]
 [12] gradient(::EnzymeCore.ForwardMode{…}, ::typeof(Diffusion), ::CuArray{…})
    @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:577
 [13] #jacobian#128
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:784 [inlined]
 [14] jacobian(::EnzymeCore.ForwardMode{…}, ::Function, ::CuArray{…})
    @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:783

reverse

ERROR: Enzyme compilation failed due to an internal error.
 Please open an issue with the code to reproduce and full error log on github.com/EnzymeAD/Enzyme.jl
 To toggle more information for debugging (needed for bug reports), set Enzyme.Compiler.VERBOSE_ERRORS[] = true (default false)

Stacktrace:
 [1] setindex!
   @ .\array.jl:987
 [2] _sort!
   @ .\sort.jl:831
 [3] multiple call sites
   @ unknown:0

Stacktrace:
  [1] (::Enzyme.Compiler.var"#getparent#69"{…})(v::LLVM.Value, offset::LLVM.Value, hasload::Bool)
    @ Enzyme.Compiler C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\llvm\transforms.jl:888
  [2] (::Enzyme.Compiler.var"#getparent#69"{…})(v::LLVM.Value, offset::LLVM.Value, hasload::Bool)
    @ Enzyme.Compiler C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\llvm\transforms.jl:777
  [3] nodecayed_phis!(mod::LLVM.Module)
    @ Enzyme.Compiler C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\llvm\transforms.jl:891
  [4] optimize!(mod::LLVM.Module, tm::LLVM.TargetMachine)
    @ Enzyme.Compiler C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\compiler\optimize.jl:582
  [5] codegen(output::Symbol, job::GPUCompiler.CompilerJob{…}; libraries::Bool, deferred_codegen::Bool, optimize::Bool, toplevel::Bool, strip::Bool, validate::Bool, only_entry::Bool, parent_job::Nothing) 
    @ Enzyme.Compiler C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\compiler.jl:4107
  [6] codegen
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\compiler.jl:3239 [inlined]
  [7] _thunk(job::GPUCompiler.CompilerJob{Enzyme.Compiler.EnzymeTarget, Enzyme.Compiler.EnzymeCompilerParams}, postopt::Bool)
    @ Enzyme.Compiler C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\compiler.jl:5288
  [8] _thunk
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\compiler.jl:5288 [inlined]
  [9] cached_compilation
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\compiler.jl:5340 [inlined]
 [10] thunkbase(mi::Core.MethodInstance, World::UInt64, FA::Type{…}, A::Type{…}, TT::Type, Mode::Enzyme.API.CDerivativeMode, width::Int64, ModifiedBetween::NTuple{…} where N, ReturnPrimal::Bool, ShadowInit::Bool, ABI::Type, ErrIfFuncWritten::Bool, RuntimeActivity::Bool, edges::Vector{…})
    @ Enzyme.Compiler C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\compiler.jl:5451
 [11] thunk_generator(world::UInt64, source::LineNumberNode, FA::Type, A::Type, TT::Type, Mode::Enzyme.API.CDerivativeMode, Width::Int64, ModifiedBetween::NTuple{…} where N, ReturnPrimal::Bool, ShadowInit::Bool, ABI::Type, ErrIfFuncWritten::Bool, RuntimeActivity::Bool, self::Any, fakeworld::Any, fa::Type, a::Type, tt::Type, mode::Type, width::Type, modifiedbetween::Type, returnprimal::Type, shadowinit::Type, abi::Type, erriffuncwritten::Type, runtimeactivity::Type)
    @ Enzyme.Compiler C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\compiler.jl:5636
 [12] autodiff_thunk
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\Enzyme.jl:975 [inlined]
 [13] jacobian(mode::EnzymeCore.ReverseMode{…}, f::typeof(Diffusion), x::CuArray{…}; n_outs::Val{…}, chunk::Nothing)
    @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:908
 [14] jacobian
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:841 [inlined]
 [15] #jacobian#129
    @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:856 [inlined]
 [16] jacobian(mode::EnzymeCore.ReverseMode{…}, f::typeof(Diffusion), x::CuArray{…})
    @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\sugar.jl:841
 [17] top-level scope
    @ c:\Users\yolha\Desktop\juju_tests\AutoDiff\main.jl:29
Some type information was truncated. Use `show(err)` to see complete types.

yolhan83 · 2024-12-16T21:35:34Z

works like a charm on cpu btw

using KernelAbstractions,CUDA
import Enzyme
uin() = 0.0f0
uout() = 0.0f0
@kernel inbounds=true function diffkernel!(du, u)
    i = @index(Global, Linear)
    if i == 1
        ug = uin()
        ud = u[i+1]
    elseif i == length(u)
        ug = u[i-1]
        ud = uout()
    else
        ug = u[i-1]
        ud = u[i+1] 
    end
    du[i] = ug + ud -2*u[i]
end
function Diffusion(u)
    bck_kernel = get_backend(u)
    du = zero(u)
    ker = diffkernel!(bck_kernel)
    ker(du, u, ndrange=length(du))
    return du
end
u = rand(Float32,1000)
Enzyme.jacobian(Enzyme.Forward, Diffusion, u)
Enzyme.jacobian(Enzyme.Reverse, Diffusion, u)

wsmoses · 2024-12-16T22:18:34Z

the reverse error I tjhink is a 1.11 issue, can you try 1.10?

For forward mode, it looks like we just ought add a onehot implementation for a CuArray

yolhan83 · 2024-12-16T22:39:19Z

not working in 1.10 but different error (for reverse)

LoadError: MethodError: no method matching lookup_world(::Type{Tuple{typeof(gpu_diffkernel!), KernelAbstractions.CompilerMetadata{KernelAbstractions.NDIteration.DynamicSize, KernelAbstractions.NDIteration.DynamicCheck, Nothing, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}, KernelAbstractions.NDIteration.NDRange{1, KernelAbstractions.NDIteration.DynamicSize, KernelAbstractions.NDIteration.DynamicSize, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}}}, CuDeviceVector{Float32, 1}, CuDeviceVector{Float32, 1}}}, ::UInt64, ::Core.Compiler.CachedMethodTable{Core.Compiler.OverlayMethodTable}, ::Base.RefValue{UInt64}, ::Base.RefValue{UInt64})

Closest candidates are:
  lookup_world(::Type, ::UInt64, ::Core.Compiler.OverlayMethodTable, ::Ref{UInt64}, ::Ref{UInt64})
   @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\utils.jl:215
  lookup_world(::Type, ::UInt64, ::Core.Compiler.InternalMethodTable, ::Ref{UInt64}, ::Ref{UInt64})
   @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\utils.jl:204
  lookup_world(::Type, ::UInt64, ::Union{Nothing, Core.MethodTable}, ::Ref{UInt64}, ::Ref{UInt64})
   @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\utils.jl:184

Stacktrace:
 [1] my_methodinstance
   @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\utils.jl:241 [inlined]
 [2] my_methodinstance (repeats 2 times)
   @ C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\utils.jl:256 [inlined]
 [3] tape_type(::GPUCompiler.CompilerJob{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}, ::EnzymeCore.ReverseModeSplit{true, true, false, 0, (false, false, true, false), EnzymeCore.FFIABI, false, false, false}, ::Type{EnzymeCore.Const{typeof(gpu_diffkernel!)}}, ::Type{EnzymeCore.Const{Nothing}}, ::Type{EnzymeCore.Const{KernelAbstractions.CompilerMetadata{KernelAbstractions.NDIteration.DynamicSize, KernelAbstractions.NDIteration.DynamicCheck, Nothing, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}, KernelAbstractions.NDIteration.NDRange{1, KernelAbstractions.NDIteration.DynamicSize, KernelAbstractions.NDIteration.DynamicSize, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}}}}}, ::Type{EnzymeCore.Duplicated{CuDeviceVector{Float32, 1}}}, ::Type{EnzymeCore.Duplicated{CuDeviceVector{Float32, 1}}})
   @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\QVjE5\src\Enzyme.jl:1249
 [4] _create_tape_kernel(::KernelAbstractions.Kernel{CUDABackend, KernelAbstractions.NDIteration.DynamicSize, KernelAbstractions.NDIteration.DynamicSize, typeof(gpu_diffkernel!)}, ::EnzymeCore.ReverseModeSplit{true, true, false, 0, (false, false, true, false), EnzymeCore.FFIABI, false, false, false}, ::Type, ::Type, ::Tuple{Int64}, ::KernelAbstractions.NDIteration.NDRange{1, KernelAbstractions.NDIteration.DynamicSize, KernelAbstractions.NDIteration.DynamicSize, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}, CartesianIndices{1, Tuple{Base.OneTo{Int64}}}}, ::EnzymeCore.Duplicated{CuArray{Float32, 1, CUDA.DeviceMemory}}, ::Vararg{EnzymeCore.Duplicated{CuArray{Float32, 1, CUDA.DeviceMemory}}})
   @ EnzymeExt C:\Users\yolha\.julia\packages\KernelAbstractions\0r40T\ext\EnzymeCore08Ext.jl:97
 [5] #augmented_primal#7
   @ C:\Users\yolha\.julia\packages\KernelAbstractions\0r40T\ext\EnzymeCore08Ext.jl:255
in expression starting at c:\Users\yolha\Desktop\juju_tests\AutoDiff\main.jl:31

wsmoses · 2024-12-16T22:46:51Z

Try with this? #2204

yolhan83 · 2024-12-16T22:55:39Z

new one hehe

ERROR: type CachedMethodTable has no field world
Stacktrace:
 [1] getproperty
   @ .\Base.jl:37 [inlined]
 [2] lookup_world
   @ C:\Users\yolha\.julia\packages\Enzyme\rwEsE\src\utils.jl:222 [inlined]
 [3] my_methodinstance
   @ C:\Users\yolha\.julia\packages\Enzyme\rwEsE\src\utils.jl:252 [inlined]
 [4] my_methodinstance (repeats 2 times)
   @ C:\Users\yolha\.julia\packages\Enzyme\rwEsE\src\utils.jl:267 [inlined]
 [5] tape_type(::GPUCompiler.CompilerJob{…}, ::ReverseModeSplit{…}, ::Type{…}, ::Type{…}, ::Type{…}, ::Type{…}, ::Type{…})
   @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\rwEsE\src\Enzyme.jl:1249
 [6] _create_tape_kernel(::KernelAbstractions.Kernel{…}, ::ReverseModeSplit{…}, ::Type, ::Type, ::Tuple{…}, ::KernelAbstractions.NDIteration.NDRange{…}, ::Duplicated{…}, ::Vararg{…})
   @ EnzymeExt C:\Users\yolha\.julia\packages\KernelAbstractions\0r40T\ext\EnzymeCore08Ext.jl:97
 [7] #augmented_primal#7
   @ C:\Users\yolha\.julia\packages\KernelAbstractions\0r40T\ext\EnzymeCore08Ext.jl:255
Some type information was truncated. Use `show(err)` to see complete types.

wsmoses · 2024-12-16T22:57:24Z

pushed again, retry?

yolhan83 · 2024-12-16T23:06:16Z

I still get the same error for now, maybe I need to gc some stuff wait a sec

yolhan83 · 2024-12-16T23:15:59Z

ok so back to scalar indexing even in reverse

LoadError: Scalar indexing is disallowed.
Invocation of getindex resulted in scalar indexing of a GPU array.
This is typically caused by calling an iterating implementation of a method.
Such implementations *do not* execute on the GPU, but very slowly on the CPU,
and therefore should be avoided.

If you want to allow scalar iteration, use `allowscalar` or `@allowscalar`
to enable scalar iteration globally or for the operations in question.
Stacktrace:
  [1] error(s::String)
    @ Base .\error.jl:35
  [2] errorscalar(op::String)
    @ GPUArraysCore C:\Users\yolha\.julia\packages\GPUArraysCore\GMsgk\src\GPUArraysCore.jl:155
  [3] _assertscalar(op::String, behavior::GPUArraysCore.ScalarIndexing)
    @ GPUArraysCore C:\Users\yolha\.julia\packages\GPUArraysCore\GMsgk\src\GPUArraysCore.jl:128
  [4] assertscalar(op::String)
    @ GPUArraysCore C:\Users\yolha\.julia\packages\GPUArraysCore\GMsgk\src\GPUArraysCore.jl:116
  [5] getindex(A::CuArray{Float32, 1, CUDA.DeviceMemory}, I::Int64)
    @ GPUArrays C:\Users\yolha\.julia\packages\GPUArrays\qt4ax\src\host\indexing.jl:50
  [6] #130
    @ C:\Users\yolha\.julia\packages\Enzyme\Kzk9s\src\sugar.jl:930 [inlined]
  [7] macro expansion
    @ .\ntuple.jl:72 [inlined]
  [8] ntuple
    @ .\ntuple.jl:69 [inlined]
  [9] jacobian(mode::ReverseMode{false, false, FFIABI, false, false}, f::typeof(Diffusion), x::CuArray{Float32, 1, CUDA.DeviceMemory}; n_outs::Val{(1000,)}, chunk::Nothing)
    @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\Kzk9s\src\sugar.jl:924
 [10] jacobian
    @ C:\Users\yolha\.julia\packages\Enzyme\Kzk9s\src\sugar.jl:841 [inlined]
 [11] #jacobian#129
    @ C:\Users\yolha\.julia\packages\Enzyme\Kzk9s\src\sugar.jl:856 [inlined]
 [12] jacobian(mode::ReverseMode{false, false, FFIABI, false, false}, f::typeof(Diffusion), x::CuArray{Float32, 1, CUDA.DeviceMemory})
    @ Enzyme C:\Users\yolha\.julia\packages\Enzyme\Kzk9s\src\sugar.jl:841
 [13] top-level scope
    @ c:\Users\yolha\Desktop\juju_tests\AutoDiff\main.jl:29
 [14] eval
    @ .\boot.jl:385 [inlined]
 [15] include_string(mapexpr::typeof(REPL.softscope), mod::Module, code::String, filename::String)
    @ Base .\loading.jl:2139
 [16] invokelatest(::Any, ::Any, ::Vararg{Any}; kwargs::@Kwargs{})
    @ Base .\essentials.jl:892
 [17] invokelatest(::Any, ::Any, ::Vararg{Any})
    @ Base .\essentials.jl:889
 [18] inlineeval(m::Module, code::String, code_line::Int64, code_column::Int64, file::String; softscope::Bool)
    @ VSCodeServer c:\Users\yolha\.vscode\extensions\julialang.language-julia-1.127.2\scripts\packages\VSCodeServer\src\eval.jl:271
 [19] (::VSCodeServer.var"#69#74"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer c:\Users\yolha\.vscode\extensions\julialang.language-julia-1.127.2\scripts\packages\VSCodeServer\src\eval.jl:181
 [20] withpath(f::VSCodeServer.var"#69#74"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams}, path::String)
    @ VSCodeServer c:\Users\yolha\.vscode\extensions\julialang.language-julia-1.127.2\scripts\packages\VSCodeServer\src\repl.jl:276
 [21] (::VSCodeServer.var"#68#73"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer c:\Users\yolha\.vscode\extensions\julialang.language-julia-1.127.2\scripts\packages\VSCodeServer\src\eval.jl:179
 [22] hideprompt(f::VSCodeServer.var"#68#73"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})
    @ VSCodeServer c:\Users\yolha\.vscode\extensions\julialang.language-julia-1.127.2\scripts\packages\VSCodeServer\src\repl.jl:38
 [23] (::VSCodeServer.var"#67#72"{Bool, Bool, Bool, Module, String, Int64, Int64, String, VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer c:\Users\yolha\.vscode\extensions\julialang.language-julia-1.127.2\scripts\packages\VSCodeServer\src\eval.jl:150
 [24] with_logstate(f::Function, logstate::Any)
    @ Base.CoreLogging .\logging.jl:515
 [25] with_logger
    @ .\logging.jl:627 [inlined]
 [26] (::VSCodeServer.var"#66#71"{VSCodeServer.ReplRunCodeRequestParams})()
    @ VSCodeServer c:\Users\yolha\.vscode\extensions\julialang.language-julia-1.127.2\scripts\packages\VSCodeServer\src\eval.jl:263
 [27] #invokelatest#2
    @ .\essentials.jl:892 [inlined]
 [28] invokelatest(::Any)
    @ Base .\essentials.jl:889
 [29] (::VSCodeServer.var"#64#65")()
    @ VSCodeServer c:\Users\yolha\.vscode\extensions\julialang.language-julia-1.127.2\scripts\packages\VSCodeServer\src\eval.jl:34
in expression starting at c:\Users\yolha\Desktop\juju_tests\AutoDiff\main.jl:29

wsmoses · 2024-12-17T00:01:36Z

ok awesome, yeah these are resolvable as well, we just should override the equivalent of onehot

wsmoses · 2024-12-23T06:08:14Z

resolved by #2220

wsmoses changed the title ~~Enzyme on KA kernel~~ CUDA: add support for onehot Dec 23, 2024

wsmoses closed this as completed Dec 23, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

CUDA: add support for onehot #2203

CUDA: add support for onehot #2203

yolhan83 commented Dec 16, 2024 •

edited

Loading

yolhan83 commented Dec 16, 2024

wsmoses commented Dec 16, 2024

yolhan83 commented Dec 16, 2024 •

edited

Loading

yolhan83 commented Dec 16, 2024

yolhan83 commented Dec 16, 2024 •

edited

Loading

wsmoses commented Dec 16, 2024

yolhan83 commented Dec 16, 2024 •

edited

Loading

wsmoses commented Dec 16, 2024

yolhan83 commented Dec 16, 2024

wsmoses commented Dec 16, 2024

yolhan83 commented Dec 16, 2024

yolhan83 commented Dec 16, 2024

wsmoses commented Dec 17, 2024

wsmoses commented Dec 23, 2024

CUDA: add support for onehot #2203

CUDA: add support for onehot #2203

Comments

yolhan83 commented Dec 16, 2024 • edited Loading

yolhan83 commented Dec 16, 2024

wsmoses commented Dec 16, 2024

yolhan83 commented Dec 16, 2024 • edited Loading

yolhan83 commented Dec 16, 2024

yolhan83 commented Dec 16, 2024 • edited Loading

wsmoses commented Dec 16, 2024

yolhan83 commented Dec 16, 2024 • edited Loading

wsmoses commented Dec 16, 2024

yolhan83 commented Dec 16, 2024

wsmoses commented Dec 16, 2024

yolhan83 commented Dec 16, 2024

yolhan83 commented Dec 16, 2024

wsmoses commented Dec 17, 2024

wsmoses commented Dec 23, 2024

yolhan83 commented Dec 16, 2024 •

edited

Loading

yolhan83 commented Dec 16, 2024 •

edited

Loading

yolhan83 commented Dec 16, 2024 •

edited

Loading

yolhan83 commented Dec 16, 2024 •

edited

Loading