Comments (10)
looks like a mapreduce error. related to #154?
from componentarrays.jl.
I’m testing this with JLArrays (which is the only way I can test CUDA-like arrays) and I am not getting an error or warning calling DiffEqBase.UNITLESS_ABS2
. I’ll have to see if there’s some way I can reproduce this without a GPU. Any thoughts @ChrisRackauckas or @YingboMa?
from componentarrays.jl.
try this
Base.mapreduce(f, op, x:: ComponentArray; kwargs...) = mapreduce(f,op,getdata(x); kwargs...)
from componentarrays.jl.
JLArrays
is working on my computer
julia> ps
ComponentVector{Int64, JLArray{Int64, 1}, Tuple{Axis{(a = 1:2, b = 3:4)}}}(a = [1, 2], b = [3, 4])
julia> DiffEqBase.UNITLESS_ABS2(ps)
ERROR: 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 are only permitted from the REPL for prototyping purposes.
If you did intend to index this array, annotate the caller with @allowscalar.
Stacktrace:
[1] error(s::String)
@ Base .\error.jl:33
[2] assertscalar(op::String)
@ GPUArraysCore C:\Users\Luffy\.julia\packages\GPUArraysCore\ZBmfM\src\GPUArraysCore.jl:87
[3] getindex
@ C:\Users\Luffy\.julia\packages\GPUArrays\Hyss4\src\host\indexing.jl:9 [inlined]
[4] getindex
@ C:\Users\Luffy\.julia\packages\ComponentArrays\cg6tC\src\array_interface.jl:96 [inlined]
[5] iterate
@ .\abstractarray.jl:1144 [inlined]
[6] iterate
@ .\abstractarray.jl:1142 [inlined]
[7] _foldl_impl(op::Base.MappingRF{typeof(DiffEqBase.UNITLESS_ABS2), Base.BottomRF{typeof(DiffEqBase.abs2_and_sum)}}, init::Int64, itr::ComponentVector{Int64, JLArray{Int64, 1}, Tuple{Axis{(a = 1:2, b = 3:4)}}})
@ Base .\reduce.jl:56
[8] foldl_impl(op::Base.MappingRF{typeof(DiffEqBase.UNITLESS_ABS2), Base.BottomRF{typeof(DiffEqBase.abs2_and_sum)}}, nt::Int64, itr::ComponentVector{Int64, JLArray{Int64, 1}, Tuple{Axis{(a = 1:2, b = 3:4)}}})
@ Base .\reduce.jl:48
[9] mapfoldl_impl
@ .\reduce.jl:44 [inlined]
[10] _mapreduce_dim
@ .\reducedim.jl:327 [inlined]
[11] #mapreduce#725
@ .\reducedim.jl:322 [inlined]
[12] UNITLESS_ABS2(x::ComponentVector{Int64, JLArray{Int64, 1}, Tuple{Axis{(a = 1:2, b = 3:4)}}})
@ DiffEqBase C:\Users\Luffy\.julia\packages\DiffEqBase\HDcso\src\common_defaults.jl:7
[13] top-level scope
@ REPL[15]:1
[14] top-level scope
@ C:\Users\Luffy\.julia\packages\CUDA\DfvRa\src\initialization.jl:52
from componentarrays.jl.
Interesting
julia> ca = ComponentArray(jl([1, 2, 3, 4]), Axis(a=1:2, b=3:4))
ComponentVector{Int64, JLArray{Int64, 1}, Tuple{Axis{(a = 1:2, b = 3:4)}}}(a = [1, 2], b = [3, 4])
julia> DiffEqBase.UNITLESS_ABS2(ca)
30
from componentarrays.jl.
Did you set JLArrays.allowscalar(false)
?
from componentarrays.jl.
Figured out a way to get broadcasting to work. With that and special casing overloads for map and mapreduce, it should fix this and all other broadcasting issues. I’ll finish it up tomorrow.
from componentarrays.jl.
@jonniedie thanks for getting to it so quickly. it seems like @milkshakeforreal is working on a fix at LuxDL/Lux.jl#109. Do you know if your fixes overlap?
from componentarrays.jl.
@vpuri3 Yeah, adapt_structure
and backend
overloads is how I had it. So they definitely overlap. I'll still do them here, since it makes more sense for that to live here instead of over at Lux.
from componentarrays.jl.
Okay, ended up needing any
, all
, count
, sum
, prod
, maximum
, and minimum
overloads in addition to multiple map
and mapreduce
methods. It could probably use more test coverage, but for now it is what it is.
from componentarrays.jl.
Related Issues (20)
- error on empty array of named tuples
- vcat ambiguity with SparseArrays method HOT 2
- add missing LinearAlgebra methods on GPU
- Problem with `Zygote.hessian` HOT 1
- `lmul!` fails with Component Array on GPU
- Incorrect gradient type. HOT 2
- Scalar indexing on GPU when computing `Zygote.gradient` of `dot(x::CA, x::CA)`
- Method ambiguities reported by Aqua
- getting "Only homogeneous arrays are allowed" error for Vector{SVector} HOT 4
- How to get index range of subarray? HOT 3
- Extend KeepIndex to Vector indices HOT 1
- Nine broken tests for Test Summary: | Pass Broken Total Time Broadcasting | 30 9 39 7.2s HOT 1
- The ComponentArray type that can specify the properties of the ComponentArray object
- Indexing `ComponentMatrix` with `FlatAxis` components HOT 4
- `Diagonal(ComponentArray)` scalar indexing error HOT 1
- ComponentVectors are not stack-able HOT 10
- ComponentArray errors with `findall`
- Can't index component array with shaped components with subset of keys HOT 2
- FR: Preserve CA-ness when indexing component matrix with shaped components HOT 1
- Get rid of `ComponentMatrix` and higher-order `ComponentArray`s? HOT 12
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from componentarrays.jl.