Merge pull request #253 from avik-pal/ap/autosparse

New High Level Interface for Sparse Jacobian Computation

.gitignore

@@ -6,3 +6,5 @@ Manifest.toml
 
 docs/build/
 docs/site/
+
+.vscode

Project.toml

@@ -1,7 +1,7 @@
 name = "SparseDiffTools"
 uuid = "47a9eef4-7e08-11e9-0b38-333d64bd3804"
 authors = ["Pankaj Mishra <[email protected]>", "Chris Rackauckas <[email protected]>"]
-version = "2.4.1"
+version = "2.5.0"
 
 [deps]
 ADTypes = "47edcb42-4c32-4615-8424-f2b9edc5f35b"
@@ -13,38 +13,45 @@ FiniteDiff = "6a86dc24-6348-571c-b903-95158fe2bd41"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 Graphs = "86223c79-3864-5bf0-83f7-82e725a168b6"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+PackageExtensionCompat = "65ce6f38-6b18-4e1d-a461-8949797d7930"
 Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
-Requires = "ae029012-a4dd-5104-9daa-d747884805df"
 SciMLOperators = "c0aeaf25-5076-4817-a8d5-81caf7dfa961"
 Setfield = "efcf1570-3423-57d1-acb7-fd33fddbac46"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 StaticArrayInterface = "0d7ed370-da01-4f52-bd93-41d350b8b718"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
 Tricks = "410a4b4d-49e4-4fbc-ab6d-cb71b17b3775"
+UnPack = "3a884ed6-31ef-47d7-9d2a-63182c4928ed"
 VertexSafeGraphs = "19fa3120-7c27-5ec5-8db8-b0b0aa330d6f"
 
 [weakdeps]
+Enzyme = "7da242da-08ed-463a-9acd-ee780be4f1d9"
+Symbolics = "0c5d862f-8b57-4792-8d23-62f2024744c7"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [extensions]
+SparseDiffToolsEnzymeExt = "Enzyme"
+SparseDiffToolsSymbolicsExt = "Symbolics"
 SparseDiffToolsZygoteExt = "Zygote"
 
 [compat]
-ADTypes = "0.1"
+ADTypes = "0.2.1"
 Adapt = "3.0"
 ArrayInterface = "7.4.2"
 Compat = "4"
 DataStructures = "0.18"
+Enzyme = "0.11"
 FiniteDiff = "2.8.1"
 ForwardDiff = "0.10"
 Graphs = "1"
 Reexport = "1"
-Requires = "1"
 SciMLOperators = "0.2.11, 0.3"
 Setfield = "1"
 StaticArrayInterface = "1.3"
 StaticArrays = "1"
+Symbolics = "5.5"
 Tricks = "0.1.6"
+UnPack = "1"
 VertexSafeGraphs = "0.2"
 Zygote = "0.6"
 julia = "1.6"
@@ -54,6 +61,7 @@ ArrayInterfaceBandedMatrices = "2e50d22c-5be1-4042-81b1-c572ed69783d"
 ArrayInterfaceBlockBandedMatrices = "5331f1e9-51c7-46b0-a9b0-df4434785e0a"
 BandedMatrices = "aae01518-5342-5314-be14-df237901396f"
 BlockBandedMatrices = "ffab5731-97b5-5995-9138-79e8c1846df0"
+Enzyme = "7da242da-08ed-463a-9acd-ee780be4f1d9"
 IterativeSolvers = "42fd0dbc-a981-5370-80f2-aaf504508153"
 Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
@@ -64,4 +72,4 @@ Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [targets]
-test = ["Test", "ArrayInterfaceBandedMatrices", "ArrayInterfaceBlockBandedMatrices", "BandedMatrices", "BlockBandedMatrices", "IterativeSolvers", "Pkg", "Random", "SafeTestsets", "Symbolics", "Zygote", "StaticArrays"]
+test = ["Test", "ArrayInterfaceBandedMatrices", "ArrayInterfaceBlockBandedMatrices", "BandedMatrices", "BlockBandedMatrices", "Enzyme", "IterativeSolvers", "Pkg", "Random", "SafeTestsets", "Symbolics", "Zygote", "StaticArrays"]

README.md

@@ -47,6 +47,67 @@ function g(x) # out-of-place
 end
 ```
 
+## High Level API
+
+We need to perform the following steps to utilize SparseDiffTools:
+
+1. Specify a Sparsity Detection Algorithm. There are 3 possible choices currently:
+    1. `NoSparsityDetection`: This will ignore any AD choice and compute the dense Jacobian
+    2. `JacPrototypeSparsityDetection`: If you already know the sparsity pattern, you can
+       specify it as `JacPrototypeSparsityDetection(; jac_prototype=<sparsity pattern>)`.
+    3. `SymbolicsSparsityDetection`: This will use `Symbolics.jl` to automatically detect
+       the sparsity pattern. (Note that `Symbolics.jl` must be explicitly loaded before
+       using this functionality.)
+2. Now choose an AD backend from `ADTypes.jl`:
+    1. If using a Non `*Sparse*` type, then we will not use sparsity detection.
+    2. All other sparse AD types will internally compute the proper sparsity pattern, and
+       try to exploit that.
+3. Now there are 2 options:
+    1. Precompute the cache using `sparse_jacobian_cache` and use the `sparse_jacobian` or
+       `sparse_jacobian!` functions to compute the Jacobian. This option is recommended if
+       you are repeatedly computing the Jacobian for the same function.
+    2. Directly use `sparse_jacobian` or `sparse_jacobian!` to compute the Jacobian. This
+       option should be used if you are only computing the Jacobian once.
+
+```julia
+using Symbolics
+
+sd = SymbolicsSparsityDetection()
+adtype = AutoSparseFiniteDiff()
+x = rand(30)
+y = similar(x)
+
+# Option 1
+## OOP Function
+cache = sparse_jacobian_cache(adtype, sd, g, x; fx=y) # Passing `fx` is needed if size(y) != size(x)
+J = sparse_jacobian(adtype, cache, g, x)
+### Or
+J_preallocated = similar(J)
+sparse_jacobian!(J_preallocated, adtype, cache, g, x)
+
+## IIP Function
+cache = sparse_jacobian_cache(adtype, sd, f, y, x)
+J = sparse_jacobian(adtype, cache, f, y, x)
+### Or
+J_preallocated = similar(J)
+sparse_jacobian!(J_preallocated, adtype, cache, f, y, x)
+
+# Option 2
+## OOP Function
+J = sparse_jacobian(adtype, sd, g, x)
+### Or
+J_preallocated = similar(J)
+sparse_jacobian!(J_preallocated, adtype, sd, g, x)
+
+## IIP Function
+J = sparse_jacobian(adtype, sd, f, y, x)
+### Or
+J_preallocated = similar(J)
+sparse_jacobian!(J_preallocated, adtype, sd, f, y, x)
+```
+
+## Lower Level API
+
 For this function, we know that the sparsity pattern of the Jacobian is a
 `Tridiagonal` matrix. However, if we didn't know the sparsity pattern for
 the Jacobian, we could use the `Symbolics.jacobian_sparsity` function to automatically

docs/make.jl

@@ -3,13 +3,13 @@ using Documenter, SparseDiffTools
 include("pages.jl")
 
 makedocs(sitename = "SparseDiffTools.jl",
-         authors = "Chris Rackauckas",
-         modules = [SparseDiffTools],
-         clean = true,
-         doctest = false,
-         format = Documenter.HTML(assets = ["assets/favicon.ico"],
-                                  canonical = "https://docs.sciml.ai/SparseDiffTools/stable/"),
-         pages = pages)
+    authors = "Chris Rackauckas",
+    modules = [SparseDiffTools],
+    clean = true,
+    doctest = false,
+    format = Documenter.HTML(assets = ["assets/favicon.ico"],
+        canonical = "https://docs.sciml.ai/SparseDiffTools/stable/"),
+    pages = pages)
 
 deploydocs(repo = "github.com/JuliaDiff/SparseDiffTools.jl.git";
-           push_preview = true)
+    push_preview = true)

docs/src/index.md

@@ -37,6 +37,67 @@ function g(x) # out-of-place
 end
 ```
 
+## High Level API
+
+We need to perform the following steps to utilize SparseDiffTools:
+
+1. Specify a Sparsity Detection Algorithm. There are 3 possible choices currently:
+    1. `NoSparsityDetection`: This will ignore any AD choice and compute the dense Jacobian
+    2. `JacPrototypeSparsityDetection`: If you already know the sparsity pattern, you can
+       specify it as `JacPrototypeSparsityDetection(; jac_prototype=<sparsity pattern>)`.
+    3. `SymbolicsSparsityDetection`: This will use `Symbolics.jl` to automatically detect
+       the sparsity pattern. (Note that `Symbolics.jl` must be explicitly loaded before
+       using this functionality.)
+2. Now choose an AD backend from `ADTypes.jl`:
+    1. If using a Non `*Sparse*` type, then we will not use sparsity detection.
+    2. All other sparse AD types will internally compute the proper sparsity pattern, and
+       try to exploit that.
+3. Now there are 2 options:
+    1. Precompute the cache using `sparse_jacobian_cache` and use the `sparse_jacobian` or
+       `sparse_jacobian!` functions to compute the Jacobian. This option is recommended if
+       you are repeatedly computing the Jacobian for the same function.
+    2. Directly use `sparse_jacobian` or `sparse_jacobian!` to compute the Jacobian. This
+       option should be used if you are only computing the Jacobian once.
+
+```julia
+using Symbolics
+
+sd = SymbolicsSparsityDetection()
+adtype = AutoSparseFiniteDiff()
+x = rand(30)
+y = similar(x)
+
+# Option 1
+## OOP Function
+cache = sparse_jacobian_cache(adtype, sd, g, x; fx=y) # Passing `fx` is needed if size(y) != size(x)
+J = sparse_jacobian(adtype, cache, g, x)
+### Or
+J_preallocated = similar(J)
+sparse_jacobian!(J_preallocated, adtype, cache, g, x)
+
+## IIP Function
+cache = sparse_jacobian_cache(adtype, sd, f, y, x)
+J = sparse_jacobian(adtype, cache, f, y, x)
+### Or
+J_preallocated = similar(J)
+sparse_jacobian!(J_preallocated, adtype, cache, f, y, x)
+
+# Option 2
+## OOP Function
+J = sparse_jacobian(adtype, sd, g, x)
+### Or
+J_preallocated = similar(J)
+sparse_jacobian!(J_preallocated, adtype, sd, g, x)
+
+## IIP Function
+J = sparse_jacobian(adtype, sd, f, y, x)
+### Or
+J_preallocated = similar(J)
+sparse_jacobian!(J_preallocated, adtype, sd, f, y, x)
+```
+
+## Lower Level API
+
 For this function, we know that the sparsity pattern of the Jacobian is a
 `Tridiagonal` matrix. However, if we didn't know the sparsity pattern for
 the Jacobian, we could use the `Symbolics.jacobian_sparsity` function to automatically

ext/SparseDiffToolsEnzymeExt.jl

@@ -0,0 +1,61 @@
+module SparseDiffToolsEnzymeExt
+
+import ArrayInterface: fast_scalar_indexing
+import SparseDiffTools: __f̂,
+    __maybe_copy_x, __jacobian!, __gradient, __gradient!, AutoSparseEnzyme
+# FIXME: For Enzyme we currently assume reverse mode
+import ADTypes: AutoEnzyme
+using Enzyme
+
+using ForwardDiff
+
+## Satisfying High-Level Interface for Sparse Jacobians
+function __gradient(::Union{AutoSparseEnzyme, AutoEnzyme}, f, x, cols)
+    dx = zero(x)
+    autodiff(Reverse, __f̂, Const(f), Duplicated(x, dx), Const(cols))
+    return vec(dx)
+end
+
+function __gradient!(::Union{AutoSparseEnzyme, AutoEnzyme}, f!, fx, x, cols)
+    dx = zero(x)
+    dfx = zero(fx)
+    autodiff(Reverse, __f̂, Active, Const(f!), Duplicated(fx, dfx), Duplicated(x, dx),
+        Const(cols))
+    return dx
+end
+
+function __jacobian!(J::AbstractMatrix, ::Union{AutoSparseEnzyme, AutoEnzyme}, f, x)
+    J .= jacobian(Reverse, f, x, Val(size(J, 1)))
+    return J
+end
+
+@views function __jacobian!(J, ad::Union{AutoSparseEnzyme, AutoEnzyme}, f!, fx, x)
+    # This version is slowish not sure how to do jacobians for inplace functions
+    @warn "Current code for computing jacobian for inplace functions in Enzyme is slow." maxlog=1
+    dfx = zero(fx)
+    dx = zero(x)
+
+    function __f_row_idx(f!, fx, x, row_idx)
+        f!(fx, x)
+        if fast_scalar_indexing(fx)
+            return fx[row_idx]
+        else
+            # Avoid the GPU Arrays scalar indexing error
+            return sum(selectdim(fx, 1, row_idx:row_idx))
+        end
+    end
+
+    for row_idx in 1:size(J, 1)
+        autodiff(Reverse, __f_row_idx, Const(f!), DuplicatedNoNeed(fx, dfx),
+            Duplicated(x, dx), Const(row_idx))
+        J[row_idx, :] .= dx
+        fill!(dfx, 0)
+        fill!(dx, 0)
+    end
+
+    return J
+end
+
+__maybe_copy_x(::Union{AutoSparseEnzyme, AutoEnzyme}, x::SubArray) = copy(x)
+
+end

ext/SparseDiffToolsSymbolicsExt.jl

@@ -0,0 +1,21 @@
+module SparseDiffToolsSymbolicsExt
+
+using SparseDiffTools, Symbolics
+import SparseDiffTools: AbstractSparseADType
+
+function (alg::SymbolicsSparsityDetection)(ad::AbstractSparseADType, f, x; fx = nothing,
+    kwargs...)
+    fx = fx === nothing ? similar(f(x)) : dx
+    f!(y, x) = (y .= f(x))
+    J = Symbolics.jacobian_sparsity(f!, fx, x)
+    _alg = JacPrototypeSparsityDetection(J, alg.alg)
+    return _alg(ad, f, x; fx, kwargs...)
+end
+
+function (alg::SymbolicsSparsityDetection)(ad::AbstractSparseADType, f!, fx, x; kwargs...)
+    J = Symbolics.jacobian_sparsity(f!, fx, x)
+    _alg = JacPrototypeSparsityDetection(J, alg.alg)
+    return _alg(ad, f!, fx, x; kwargs...)
+end
+
+end