Replaced artifacts with JLL artifacts

.pkg/platform_augmentation.jl

@@ -0,0 +1,35 @@
+using Libdl, Base.BinaryPlatforms
+
+function augment_platform!(p::Platform, tag::Union{String,Nothing} = nothing)
+    if tag === nothing
+        return p
+    end
+
+    if tag === "cuda"
+        if haskey(p, tag)
+            return p
+        end
+
+        # Open libcuda explicitly, so it gets `dlclose()`'ed after we're done
+        try
+            dlopen("libcuda") do lib
+                # find symbol to ask for driver version; if we can't find it, just silently continue
+                cuDriverGetVersion = dlsym(lib, "cuDriverGetVersion"; throw_error=false)
+                if cuDriverGetVersion !== nothing
+                    # Interrogate CUDA driver for driver version:
+                    driverVersion = Ref{Cint}()
+                    ccall(cuDriverGetVersion, UInt32, (Ptr{Cint},), driverVersion)
+
+                    # Store only the major version
+                    p[tag] = div(driverVersion, 1000)
+                end
+            end
+        catch
+        end
+
+        return p
+    else
+        @warn "Unexpected tag $tag for $p"
+        return p
+    end
+end

.pkg/select_artifacts.jl

@@ -0,0 +1,19 @@
+using TOML, Artifacts, Base.BinaryPlatforms
+
+import ONNXRuntime_jll
+
+include("platform_augmentation.jl")
+
+artifacts_toml = Artifacts.find_artifacts_toml(pathof(ONNXRuntime_jll))
+
+# Get "target triplet" from ARGS, if given (defaulting to the host triplet otherwise)
+target_triplet = get(ARGS, 1, Base.BinaryPlatforms.host_triplet())
+
+# Augment this platform object with any special tags we require
+platform = augment_platform!(HostPlatform(parse(Platform, target_triplet)))
+
+# Select all downloadable artifacts that match that platform
+artifacts = select_downloadable_artifacts(artifacts_toml; platform)
+
+# Output the result to `stdout` as a TOML dictionary
+TOML.print(stdout, artifacts)

Project.toml

@@ -5,12 +5,15 @@ version = "1.3.1"
 
 [deps]
 ArgCheck = "dce04be8-c92d-5529-be00-80e4d2c0e197"
+Artifacts = "56f22d72-fd6d-98f1-02f0-08ddc0907c33"
 CEnum = "fa961155-64e5-5f13-b03f-caf6b980ea82"
 DataStructures = "864edb3b-99cc-5e75-8d2d-829cb0a9cfe8"
 DocStringExtensions = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
 LazyArtifacts = "4af54fe1-eca0-43a8-85a7-787d91b784e3"
 Libdl = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
+ONNXRuntime_jll = "09e6dd1b-8208-5c7e-a336-6e9061773d0b"
 Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
+TOML = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
 
 [compat]
 ArgCheck = "2"

src/ONNXRunTime.jl

@@ -1,5 +1,9 @@
 module ONNXRunTime
 
+using Artifacts
+using LazyArtifacts
+using ONNXRuntime_jll
+
 function _perm(arr::AbstractArray{T,N}) where {T,N}
     ntuple(i->N+1-i, N)
 end

src/capi.jl

@@ -5,20 +5,19 @@ This module closely follows the offical onnxruntime [C-API](https://github.com/m
 See [here](https://github.com/microsoft/onnxruntime-inference-examples/blob/d031f879c9a8d33c8b7dc52c5bc65fe8b9e3960d/c_cxx/fns_candy_style_transfer/fns_candy_style_transfer.c) for a C code example.
 """
 module CAPI
-using ONNXRunTime: reversedims_lazy
+
+using ONNXRunTime: EXECUTION_PROVIDERS, artifact_dir_map, reversedims_lazy
+
+using ONNXRuntime_jll
 
 using DocStringExtensions
 using Libdl
 using CEnum: @cenum
 using ArgCheck
-using LazyArtifacts
-using Pkg.Artifacts: artifact_path, ensure_artifact_installed, find_artifacts_toml
 
 const LIB_CPU  = Ref(C_NULL)
 const LIB_CUDA = Ref(C_NULL)
 
-const EXECUTION_PROVIDERS = [:cpu, :cuda]
-
 # For model_path on windows ONNX uses wchar_t while on linux + mac char is used.
 # Other strings use char on any platform it seems
 # https://github.com/microsoft/onnxruntime/issues/9568#issuecomment-952951564
@@ -45,36 +44,9 @@ end
 
 function make_lib!(execution_provider)
     @argcheck execution_provider in EXECUTION_PROVIDERS
-    artifact_name = if execution_provider === :cpu
-        "onnxruntime_cpu"
-    elseif execution_provider === :cuda
-        "onnxruntime_gpu"
-    else
-        error("Unreachable")
-    end
-    artifacts_toml = find_artifacts_toml(joinpath(@__DIR__ , "ONNXRunTime.jl"))
-    h = artifact_hash(artifact_name, artifacts_toml)
-    if h === nothing
-        msg = """
-        Unsupported execution_provider = $(repr(execution_provider)) for
-        this architectur.
-        """
-        error(msg)
-    end
-    ensure_artifact_installed(artifact_name, artifacts_toml)
-    root = artifact_path(h)
-    @check isdir(root)
-    dir = joinpath(root, only(readdir(root)))
-    @check isdir(dir)
-    libname = if Sys.iswindows()
-        "onnxruntime.dll"
-    elseif Sys.isapple()
-        "libonnxruntime.dylib"
-    else
-        "libonnxruntime.so"
-    end
-    path = joinpath(dir, "lib", libname)
-    @check isfile(path)
+    path = ONNXRuntime_jll.libonnxruntime_path
+    rel_path = joinpath(basename(dirname(path)), basename(path))
+    path = joinpath(artifact_dir_map[execution_provider], rel_path)
     set_lib!(path, execution_provider)
 end
 

src/highlevel.jl

@@ -1,5 +1,4 @@
 using ArgCheck
-using LazyArtifacts
 using DataStructures: OrderedDict
 using DocStringExtensions
 import CEnum
@@ -12,7 +11,7 @@ end
 
 
 using .CAPI
-using .CAPI: juliatype, EXECUTION_PROVIDERS
+using .CAPI: juliatype
 export InferenceSession, load_inference, release
 
 """