Skip to content

One way to integrate into existing CUDA infrastructure #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
DhairyaLGandhi wants to merge 10 commits into
base: master
Choose a base branch
from
Open

One way to integrate into existing CUDA infrastructure #2

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
2ad706f
use cuptrs instead of normal pointers
Mar 4, 2020
1138070
add using
Mar 4, 2020
2b63147
ptr{cvoid} -> ptr{T}
Mar 4, 2020
25df3fa
converting to a CuArray
Mar 4, 2020
fa013f2
add Artifacts
Mar 11, 2020
8afd2fb
Merge branch 'master' into dg/cu
DhairyaLGandhi Mar 11, 2020
5639e48
Merge branch 'master' into dg/cu
DhairyaLGandhi Mar 11, 2020
5777545
Merge branch 'master' of https://github.com/dhairyagandhi96/Torch.jl …
Mar 11, 2020
8e04282
add arch to artifacts
Mar 11, 2020
0fa5bdf
add correct checksums
Mar 11, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 6 additions & 4 deletions Artifacts.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,17 +1,19 @@
[[torch]]
git-tree-sha1 = "21fba0f64d7e6dce93146b84ba73226f4f0f0983"
os = "macos"
arch = "x86_64"
git-tree-sha1 = "21fba0f64d7e6dce93146b84ba73226f4f0f0983"
os = "macos"

[[torch.download]]
sha256 = "21fba0f64d7e6dce93146b84ba73226f4f0f0983"
sha256 = "84e9112b442ee1e3dc9e078d9066a855a2344ec566616cffbff1662e08cd8bf7"
url = "https://download.pytorch.org/libtorch/cpu/libtorch-macos-1.4.0.zip"

[[torch]]
arch = "x86_64"
git-tree-sha1 = "42bc9814d06d949098f5ab218be61adaa3e29c06"
os = "linux"

[[torch.download]]
sha256 = "42bc9814d06d949098f5ab218be61adaa3e29c06"
sha256 = "f214bfde532877aa5d4e0803e51a28fa8edd97b6a44b6615f75a70352b6b542e"
url = "https://download.pytorch.org/libtorch/cu101/libtorch-cxx11-abi-shared-with-deps-1.4.0.zip"


17 changes: 17 additions & 0 deletions Manifest.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,23 @@ git-tree-sha1 = "5b08ed6036d9d3f0ee6369410b830f8873d4024c"
uuid = "b99e7846-7c00-51b0-8f62-c81ae34c0232"
version = "0.5.8"

[[CEnum]]
git-tree-sha1 = "62847acab40e6855a9b5905ccb99c2b5cf6b3ebb"
uuid = "fa961155-64e5-5f13-b03f-caf6b980ea82"
version = "0.2.0"

[[CUDAapi]]
deps = ["Libdl", "Logging"]
git-tree-sha1 = "d7ceadd8f821177d05b897c0517e94633db535fe"
uuid = "3895d2a7-ec45-59b8-82bb-cfc6a382f9b3"
version = "3.1.0"

[[CUDAdrv]]
deps = ["CEnum", "CUDAapi", "Printf"]
git-tree-sha1 = "01e90fa34e25776bc7c8661183d4519149ebfe59"
uuid = "c5f51814-7f29-56b8-a69c-e4d8f6be1fde"
version = "6.0.0"

[[Codecs]]
deps = ["Test"]
git-tree-sha1 = "70885e5e038cba1c4c17a84ad6c40756e10a4fb5"
Expand Down
1 change: 1 addition & 0 deletions Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@ authors = ["dhairyagandhi96 "]
version = "0.1.0"

[deps]
CUDAdrv = "c5f51814-7f29-56b8-a69c-e4d8f6be1fde"
Libdl = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
NNlib = "872c559c-99b0-510c-b3b7-b6c96a88d5cd"
Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
Expand Down
5 changes: 5 additions & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,5 +22,10 @@ tip = tensor(ip, dev = 0) # 0 => GPU:0 in Torch
tresnet(tip);
```

## Making a CuArray
```julia
CuArray{Float32,4}(tip.ptr, size(tip), false)::CuArray
```

## Acknowledgements
Takes a lot of inspiration from existing such projects - ocaml-torch for generating the wrappers.
1 change: 1 addition & 0 deletions src/Torch.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@ export Tensor, tensor, Scalar

using YAML
using Clang
using CUDAdrv
using ZygoteRules
using ZygoteRules: @adjoint
using NNlib
Expand Down
8 changes: 5 additions & 3 deletions src/tensor.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,11 +18,13 @@ async_free!(x) = let x = x, ptr = x.ptr, oid = objectid(x)
end

mutable struct Tensor{T, N} <: AbstractArray{T,N}
ptr::Ptr
ptr::Union{Ptr,CuPtr}
device::Int

function Tensor{T,N}(ptr::Ptr, dev::Int) where {T,N}
obj = new(ptr, dev)
function Tensor{T,N}(ptr::Union{Ptr,CuPtr}, dev::Int) where {T,N}
cuptr = convert(CuPtr{Cvoid}, Base.bitcast(UInt, ptr))
cuptr = convert(CuPtr{T}, cuptr)
obj = new(cuptr, dev)
finalizer(async_free!, obj)
# TURN_ON_LOGGING == true && (logdict[ptr] = (size(obj), stacktrace()))
obj
Expand Down
1,900 changes: 950 additions & 950 deletions src/wrap/libdoeye_caml_generated.jl
View file
Open in desktop

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion src/wrap/libtorch_common.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
# Automatically generated using Clang.jl


const tensor = Ptr{Cvoid}
const tensor = CuPtr{Cvoid}
const optimizer = Ptr{Cvoid}
const scalar = Ptr{Cvoid}
const module_t = Ptr{Cvoid}
Expand Down