diff --git a/Cargo.lock b/Cargo.lock
index a77dcf9..2f0dadb 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -17,6 +17,15 @@ version = "0.2.21"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "683d7910e743518b0e34f1186f92494becacb047c7b6bf616c96772180fef923"
 
+[[package]]
+name = "android_system_properties"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "819e7219dbd41043ac279b19830f2efc897156490d7fd6ea916720117ee66311"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "anes"
 version = "0.1.6"
@@ -30,12 +39,30 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "55cc3b69f167a1ef2e161439aa98aed94e6028e5f9a59be9a6ffb47aef1651f9"
 
 [[package]]
-name = "arbitrary"
-version = "1.4.1"
+name = "arrayvec"
+version = "0.7.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7c02d123df017efcdfbd739ef81735b36c5ba83ec3c59c80a9d7ecc718f92e50"
+
+[[package]]
+name = "ash"
+version = "0.38.0+1.3.281"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0bb44936d800fea8f016d7f2311c6a4f97aebd5dc86f09906139ec848cf3a46f"
+dependencies = [
+ "libloading",
+]
+
+[[package]]
+name = "async-channel"
+version = "2.3.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dde20b3d026af13f561bdd0f15edf01fc734f0dafcedbaf42bba506a9517f223"
+checksum = "89b47800b0be77592da0afd425cc03468052844aff33b84e33cc696f64e77b6a"
 dependencies = [
- "derive_arbitrary",
+ "concurrent-queue",
+ "event-listener-strategy",
+ "futures-core",
+ "pin-project-lite",
 ]
 
 [[package]]
@@ -44,6 +71,21 @@ version = "1.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0c4b4d0bd25bd0b74681c0ad21497610ce1b7c91b1022cd21c80c6fbdd9476b0"
 
+[[package]]
+name = "bit-set"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "08807e080ed7f9d5433fa9b275196cfc35414f66a0c79d864dc51a0d825231a3"
+dependencies = [
+ "bit-vec",
+]
+
+[[package]]
+name = "bit-vec"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5e764a1d40d510daf35e07be9eb06e75770908c27d411ee6c92109c9840eaaf7"
+
 [[package]]
 name = "bitflags"
 version = "1.3.2"
@@ -52,9 +94,18 @@ checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
 
 [[package]]
 name = "bitflags"
-version = "2.5.0"
+version = "2.9.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "cf4b9d6a944f767f8e5e0db018570623c85f3d925ac718db4e06d0187adb21c1"
+checksum = "5c8214115b7bf84099f1309324e63141d4c5d7cc26862f97a0a857dbefe165bd"
+dependencies = [
+ "serde",
+]
+
+[[package]]
+name = "block"
+version = "0.1.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0d8c1fef690941d3e7788d328517591fecc684c084084702d6ff1641e993699a"
 
 [[package]]
 name = "bumpalo"
@@ -88,28 +139,6 @@ version = "1.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
 
-[[package]]
-name = "candle-core"
-version = "0.8.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bef79e00300acaaa6b0b33168e2ed5050f06125d1add87e7103c40a119083581"
-dependencies = [
- "byteorder",
- "gemm 0.17.1",
- "half",
- "memmap2",
- "num-traits",
- "num_cpus",
- "rand 0.8.5",
- "rand_distr 0.4.3",
- "rayon",
- "safetensors",
- "thiserror",
- "ug",
- "yoke",
- "zip",
-]
-
 [[package]]
 name = "cast"
 version = "0.3.0"
@@ -122,6 +151,12 @@ version = "1.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
 
+[[package]]
+name = "cfg_aliases"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "613afe47fcd5fac7ccf1db93babcb082c5994d996f20b8b159f2ad1658eb5724"
+
 [[package]]
 name = "ciborium"
 version = "0.2.2"
@@ -174,31 +209,69 @@ version = "0.7.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f46ad14479a25103f283c0f10005961cf086d8dc42205bb44c46ac563475dca6"
 
+[[package]]
+name = "codespan-reporting"
+version = "0.12.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fe6d2e5af09e8c8ad56c969f2157a3d4238cebc7c55f0a517728c38f7b200f81"
+dependencies = [
+ "serde",
+ "termcolor",
+ "unicode-width",
+]
+
+[[package]]
+name = "concurrent-queue"
+version = "2.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4ca0197aee26d1ae37445ee532fefce43251d24cc7c166799f4d46817f1d3973"
+dependencies = [
+ "crossbeam-utils",
+]
+
 [[package]]
 name = "constensor-core"
 version = "0.1.1"
 dependencies = [
- "candle-core",
  "criterion",
+ "cubecl",
  "cudarc",
- "dirs",
- "gemm 0.18.2",
+ "dirs 5.0.1",
+ "gemm",
  "half",
  "num_cpus",
  "petgraph",
- "rand 0.9.1",
- "rand_distr 0.5.1",
+ "rand",
+ "rand_distr",
  "rayon",
- "thiserror",
+ "thiserror 1.0.61",
 ]
 
 [[package]]
-name = "crc32fast"
-version = "1.4.2"
+name = "core-foundation"
+version = "0.9.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a97769d94ddab943e4510d138150169a2758b5ef3eb191a9ee688de3e23ef7b3"
+checksum = "91e195e091a93c46f7102ec7818a2aa394e1e1771c3ab4825963fa03e45afb8f"
 dependencies = [
- "cfg-if",
+ "core-foundation-sys",
+ "libc",
+]
+
+[[package]]
+name = "core-foundation-sys"
+version = "0.8.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "773648b94d0e5d620f64f280777445740e61fe701025087ec8b57f45c791888b"
+
+[[package]]
+name = "core-graphics-types"
+version = "0.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "45390e6114f68f718cc7a830514a96f903cccd70d02a8f6d9f643ac4ba45afaf"
+dependencies = [
+ "bitflags 1.3.2",
+ "core-foundation",
+ "libc",
 ]
 
 [[package]]
@@ -237,6 +310,12 @@ dependencies = [
  "itertools",
 ]
 
+[[package]]
+name = "critical-section"
+version = "1.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "790eea4361631c5e7d22598ecd5723ff611904e3344ce8720784c93e3d83d40b"
+
 [[package]]
 name = "crossbeam-deque"
 version = "0.8.5"
@@ -268,25 +347,384 @@ version = "0.2.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7a81dae078cea95a014a339291cec439d2f232ebe854a9d672b796c6afafa9b7"
 
+[[package]]
+name = "cubecl"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b418aabe526efb2d2714b9282a583bccbc555fe2e42546101259b2042bc133fa"
+dependencies = [
+ "cubecl-convolution",
+ "cubecl-core",
+ "cubecl-cuda",
+ "cubecl-hip",
+ "cubecl-matmul",
+ "cubecl-runtime",
+ "cubecl-std",
+ "cubecl-wgpu",
+ "half",
+]
+
+[[package]]
+name = "cubecl-common"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ff2bed7d9aa1a2fe640696067cab6b0b1b40dd30cdb8d12f91ae43ca6392498e"
+dependencies = [
+ "bytemuck",
+ "cfg-if",
+ "derive-new",
+ "derive_more",
+ "dirs 6.0.0",
+ "embassy-futures",
+ "embassy-time",
+ "futures-lite",
+ "half",
+ "hashbrown",
+ "log",
+ "num-traits",
+ "portable-atomic",
+ "rand",
+ "sanitize-filename",
+ "serde",
+ "serde_json",
+ "spin",
+ "wasm-bindgen-futures",
+ "web-time",
+]
+
+[[package]]
+name = "cubecl-convolution"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a324c7e684c627a2746fb4700a29b5881970c7f93cc06c66bcf3c2c5184af97b"
+dependencies = [
+ "bytemuck",
+ "cubecl-common",
+ "cubecl-core",
+ "cubecl-matmul",
+ "cubecl-random",
+ "cubecl-reduce",
+ "cubecl-runtime",
+ "cubecl-std",
+ "half",
+ "serde",
+]
+
+[[package]]
+name = "cubecl-core"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bc4c683bb3ed715eed12cfeca6599dc554efb9371e85f783eb7c3ae0d3781274"
+dependencies = [
+ "bitflags 2.9.0",
+ "bytemuck",
+ "cubecl-common",
+ "cubecl-ir",
+ "cubecl-macros",
+ "cubecl-runtime",
+ "derive-new",
+ "derive_more",
+ "half",
+ "hashbrown",
+ "log",
+ "num-traits",
+ "paste",
+ "serde",
+ "serde_json",
+ "variadics_please",
+]
+
+[[package]]
+name = "cubecl-cpp"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0d06ed3e34f0a9f0a1ca225fc3351865a2ddf66966c73a7953dd27f1bbaed9b9"
+dependencies = [
+ "bytemuck",
+ "cubecl-common",
+ "cubecl-core",
+ "cubecl-runtime",
+ "derive-new",
+ "half",
+ "log",
+]
+
+[[package]]
+name = "cubecl-cuda"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bf8898be518bd3df9f7d0df8c25e95f07cfc087c6485dda42e99add432e3463b"
+dependencies = [
+ "bytemuck",
+ "cubecl-common",
+ "cubecl-core",
+ "cubecl-cpp",
+ "cubecl-runtime",
+ "cudarc",
+ "derive-new",
+ "half",
+ "log",
+ "serde",
+]
+
+[[package]]
+name = "cubecl-hip"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d9afb1c085f8d715fd8686ad77ada69e364c9cc16a51cf10a7d65ad88d760b6b"
+dependencies = [
+ "bytemuck",
+ "cubecl-common",
+ "cubecl-core",
+ "cubecl-cpp",
+ "cubecl-hip-sys",
+ "cubecl-runtime",
+ "derive-new",
+ "half",
+ "log",
+ "paste",
+ "serde",
+]
+
+[[package]]
+name = "cubecl-hip-sys"
+version = "6.4.4348201"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "678a20e5e38ce9c772bdd53596f2801ef210ae735ec2d7d46b5d5b675c09d929"
+dependencies = [
+ "libc",
+ "regex",
+]
+
+[[package]]
+name = "cubecl-ir"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bbb72454fb17da66e60f98a011cce199ff0bb78c4e465236459af619039b6cd5"
+dependencies = [
+ "cubecl-common",
+ "cubecl-macros-internal",
+ "derive_more",
+ "float-ord",
+ "fnv",
+ "half",
+ "hashbrown",
+ "num-traits",
+ "portable-atomic",
+ "serde",
+ "variadics_please",
+]
+
+[[package]]
+name = "cubecl-macros"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "72d61316b8a2ec26bb554745591174022bb048dbcc508091b15dd4f6fcb4793a"
+dependencies = [
+ "cubecl-common",
+ "darling",
+ "derive-new",
+ "ident_case",
+ "prettyplease",
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "cubecl-macros-internal"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b7c27cf02393d78a41ad125ef6d6274cd3930385f6945779f4f21a2c66e718db"
+dependencies = [
+ "darling",
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "cubecl-matmul"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "66af2716f458eed3bd14d810d9a90bc43dd35c5464b2fe4630a711a074892fb8"
+dependencies = [
+ "bytemuck",
+ "cubecl-common",
+ "cubecl-core",
+ "cubecl-random",
+ "cubecl-reduce",
+ "cubecl-runtime",
+ "cubecl-std",
+ "half",
+ "serde",
+]
+
+[[package]]
+name = "cubecl-random"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "648731639463df215c01ee209477a2f233fb5168fb50e8df755e658ddaaf29f4"
+dependencies = [
+ "cubecl-common",
+ "cubecl-core",
+ "cubecl-runtime",
+ "cubecl-std",
+ "half",
+ "num-traits",
+ "rand",
+ "serde",
+]
+
+[[package]]
+name = "cubecl-reduce"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "729e3de15783aed5c617c8c4de633765c6d0cf4b9564bb51d21dd67b4fdeeebf"
+dependencies = [
+ "cubecl-core",
+ "cubecl-runtime",
+ "cubecl-std",
+ "half",
+ "num-traits",
+ "serde",
+]
+
+[[package]]
+name = "cubecl-runtime"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "68ab7c56cbb83e43fd1b52ee3d521bcb816068e0bb6f4aeb56c9bd63218b1147"
+dependencies = [
+ "async-channel",
+ "bytemuck",
+ "cfg-if",
+ "cfg_aliases",
+ "cubecl-common",
+ "cubecl-ir",
+ "derive-new",
+ "dirs 6.0.0",
+ "foldhash",
+ "hashbrown",
+ "log",
+ "md5",
+ "serde",
+ "serde_json",
+ "spin",
+ "toml",
+ "variadics_please",
+ "wasm-bindgen-futures",
+]
+
+[[package]]
+name = "cubecl-std"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3536c4c1c878bd77ceda959a3f157bbb69f957269e3f138c7edd72d9deb533fd"
+dependencies = [
+ "cubecl-core",
+ "cubecl-runtime",
+ "half",
+ "serde",
+]
+
+[[package]]
+name = "cubecl-wgpu"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1d3f068c9a81201f43cdd3f6385201b330f2499f6c24d688d82a70d705261a32"
+dependencies = [
+ "async-channel",
+ "bytemuck",
+ "cfg-if",
+ "cfg_aliases",
+ "cubecl-common",
+ "cubecl-core",
+ "cubecl-cpp",
+ "cubecl-runtime",
+ "derive-new",
+ "derive_more",
+ "half",
+ "hashbrown",
+ "log",
+ "wgpu",
+]
+
 [[package]]
 name = "cudarc"
-version = "0.16.1"
+version = "0.16.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "13f3807b96899cd88fb9feecc41316cd0879df6f2726616ab88a4bc919af7663"
+checksum = "17200eb07e7d85a243aa1bf4569a7aa998385ba98d14833973a817a63cc86e92"
 dependencies = [
  "half",
  "libloading",
 ]
 
 [[package]]
-name = "derive_arbitrary"
-version = "1.4.1"
+name = "darling"
+version = "0.21.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "30542c1ad912e0e3d22a1935c290e12e8a29d704a420177a31faad4a601a0800"
+checksum = "08440b3dd222c3d0433e63e097463969485f112baff337dfdaca043a0d760570"
+dependencies = [
+ "darling_core",
+ "darling_macro",
+]
+
+[[package]]
+name = "darling_core"
+version = "0.21.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d25b7912bc28a04ab1b7715a68ea03aaa15662b43a1a4b2c480531fd19f8bf7e"
+dependencies = [
+ "fnv",
+ "ident_case",
+ "proc-macro2",
+ "quote",
+ "strsim",
+ "syn",
+]
+
+[[package]]
+name = "darling_macro"
+version = "0.21.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ce154b9bea7fb0c8e8326e62d00354000c36e79770ff21b8c84e3aa267d9d531"
+dependencies = [
+ "darling_core",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "derive-new"
+version = "0.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2cdc8d50f426189eef89dac62fabfa0abb27d5cc008f25bf4156a0203325becc"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "derive_more"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "093242cf7570c207c83073cf82f79706fe7b8317e98620a47d5be7c3d8497678"
+dependencies = [
+ "derive_more-impl",
+]
+
+[[package]]
+name = "derive_more-impl"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bda628edc44c4bb645fbe0f758797143e4e07926f7ebf4e9bdfbd3d2ce621df3"
 dependencies = [
  "proc-macro2",
  "quote",
  "syn",
+ "unicode-xid",
 ]
 
 [[package]]
@@ -295,7 +733,16 @@ version = "5.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "44c45a9d03d6676652bcb5e724c7e988de1acad23a711b5217ab9cbecbec2225"
 dependencies = [
- "dirs-sys",
+ "dirs-sys 0.4.1",
+]
+
+[[package]]
+name = "dirs"
+version = "6.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c3e8aa94d75141228480295a7d0e7feb620b1a5ad9f12bc40be62411e38cce4e"
+dependencies = [
+ "dirs-sys 0.5.0",
 ]
 
 [[package]]
@@ -306,29 +753,29 @@ checksum = "520f05a5cbd335fae5a99ff7a6ab8627577660ee5cfd6a94a6a929b52ff0321c"
 dependencies = [
  "libc",
  "option-ext",
- "redox_users",
+ "redox_users 0.4.5",
  "windows-sys 0.48.0",
 ]
 
 [[package]]
-name = "displaydoc"
-version = "0.2.5"
+name = "dirs-sys"
+version = "0.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "97369cbbc041bc366949bc74d34658d6cda5621039731c6310521892a3a20ae0"
+checksum = "e01a3366d27ee9890022452ee61b2b63a67e6f13f58900b651ff5665f0bb1fab"
 dependencies = [
- "proc-macro2",
- "quote",
- "syn",
+ "libc",
+ "option-ext",
+ "redox_users 0.5.2",
+ "windows-sys 0.59.0",
 ]
 
 [[package]]
-name = "dyn-stack"
-version = "0.10.0"
+name = "document-features"
+version = "0.2.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "56e53799688f5632f364f8fb387488dd05db9fe45db7011be066fc20e7027f8b"
+checksum = "95249b50c6c185bee49034bcb378a49dc2b5dff0be90ff6616d31d64febab05d"
 dependencies = [
- "bytemuck",
- "reborrow",
+ "litrs",
 ]
 
 [[package]]
@@ -346,6 +793,62 @@ version = "1.12.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3dca9240753cf90908d7e4aac30f630662b02aebaa1b58a3cadabdb23385b58b"
 
+[[package]]
+name = "embassy-futures"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1f878075b9794c1e4ac788c95b728f26aa6366d32eeb10c7051389f898f7d067"
+
+[[package]]
+name = "embassy-time"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f820157f198ada183ad62e0a66f554c610cdcd1a9f27d4b316358103ced7a1f8"
+dependencies = [
+ "cfg-if",
+ "critical-section",
+ "document-features",
+ "embassy-time-driver",
+ "embedded-hal 0.2.7",
+ "embedded-hal 1.0.0",
+ "embedded-hal-async",
+ "futures-util",
+]
+
+[[package]]
+name = "embassy-time-driver"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8d45f5d833b6d98bd2aab0c2de70b18bfaa10faf661a1578fd8e5dfb15eb7eba"
+dependencies = [
+ "document-features",
+]
+
+[[package]]
+name = "embedded-hal"
+version = "0.2.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "35949884794ad573cf46071e41c9b60efb0cb311e3ca01f7af807af1debc66ff"
+dependencies = [
+ "nb 0.1.3",
+ "void",
+]
+
+[[package]]
+name = "embedded-hal"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "361a90feb7004eca4019fb28352a9465666b24f840f5c3cddf0ff13920590b89"
+
+[[package]]
+name = "embedded-hal-async"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0c4c685bbef7fe13c3c6dd4da26841ed3980ef33e841cddfa15ce8a8fb3f1884"
+dependencies = [
+ "embedded-hal 1.0.0",
+]
+
 [[package]]
 name = "enum-as-inner"
 version = "0.6.1"
@@ -364,12 +867,51 @@ version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "877a4ace8713b0bcf2a4e7eec82529c029f1d0619886d18145fea96c3ffe5c0f"
 
+[[package]]
+name = "event-listener"
+version = "5.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3492acde4c3fc54c845eaab3eed8bd00c7a7d881f78bfc801e43a93dec1331ae"
+dependencies = [
+ "concurrent-queue",
+ "parking",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "event-listener-strategy"
+version = "0.5.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8be9f3dfaaffdae2972880079a491a1a8bb7cbed0b8dd7a347f668b4150a3b93"
+dependencies = [
+ "event-listener",
+ "pin-project-lite",
+]
+
+[[package]]
+name = "fastrand"
+version = "2.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"
+
 [[package]]
 name = "fixedbitset"
 version = "0.5.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1d674e81391d1e1ab681a28d99df07927c6d4aa5b027d7da16ba32d1d21ecd99"
 
+[[package]]
+name = "float-ord"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8ce81f49ae8a0482e4c55ea62ebbd7e5a686af544c00b9d090bba3ff9be97b3d"
+
+[[package]]
+name = "fnv"
+version = "1.0.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3f9eec918d3f24069decb9af1554cad7c880e2da24a9afd88aca000531ab82c1"
+
 [[package]]
 name = "foldhash"
 version = "0.1.5"
@@ -377,123 +919,123 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d9c4f5dac5e15c24eb999c26181a6ca40b39fe946cbe4c263c7209467bc83af2"
 
 [[package]]
-name = "gemm"
-version = "0.17.1"
+name = "foreign-types"
+version = "0.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ab24cc62135b40090e31a76a9b2766a501979f3070fa27f689c27ec04377d32"
+checksum = "d737d9aa519fb7b749cbc3b962edcf310a8dd1f4b67c91c4f83975dbdd17d965"
 dependencies = [
- "dyn-stack 0.10.0",
- "gemm-c32 0.17.1",
- "gemm-c64 0.17.1",
- "gemm-common 0.17.1",
- "gemm-f16 0.17.1",
- "gemm-f32 0.17.1",
- "gemm-f64 0.17.1",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 10.7.0",
- "seq-macro",
+ "foreign-types-macros",
+ "foreign-types-shared",
 ]
 
 [[package]]
-name = "gemm"
-version = "0.18.2"
+name = "foreign-types-macros"
+version = "0.2.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ab96b703d31950f1aeddded248bc95543c9efc7ac9c4a21fda8703a83ee35451"
+checksum = "1a5c6c585bc94aaf2c7b51dd4c2ba22680844aba4c687be581871a6f518c5742"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-c32 0.18.2",
- "gemm-c64 0.18.2",
- "gemm-common 0.18.2",
- "gemm-f16 0.18.2",
- "gemm-f32 0.18.2",
- "gemm-f64 0.18.2",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 11.5.0",
- "seq-macro",
+ "proc-macro2",
+ "quote",
+ "syn",
 ]
 
 [[package]]
-name = "gemm-c32"
-version = "0.17.1"
+name = "foreign-types-shared"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "aa9a19cbb55df58761df49b23516a86d432839add4af60fc256da840f66ed35b"
+
+[[package]]
+name = "futures-core"
+version = "0.3.31"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b9c030d0b983d1e34a546b86e08f600c11696fde16199f971cd46c12e67512c0"
+checksum = "05f29059c0c2090612e8d742178b0580d2dc940c837851ad723096f87af6663e"
+
+[[package]]
+name = "futures-io"
+version = "0.3.31"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9e5c1b78ca4aae1ac06c48a526a655760685149f0d465d21f37abfe57ce075c6"
+
+[[package]]
+name = "futures-lite"
+version = "2.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f5edaec856126859abb19ed65f39e90fea3a9574b9707f13539acf4abf7eb532"
 dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 10.7.0",
- "seq-macro",
+ "fastrand",
+ "futures-core",
+ "futures-io",
+ "parking",
+ "pin-project-lite",
 ]
 
 [[package]]
-name = "gemm-c32"
-version = "0.18.2"
+name = "futures-task"
+version = "0.3.31"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f6db9fd9f40421d00eea9dd0770045a5603b8d684654816637732463f4073847"
+checksum = "f90f7dce0722e95104fcb095585910c0977252f286e354b5e3bd38902cd99988"
+
+[[package]]
+name = "futures-util"
+version = "0.3.31"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9fa08315bb612088cc391249efdc3bc77536f16c91f6cf495e6fbe85b20a4a81"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 11.5.0",
- "seq-macro",
+ "futures-core",
+ "futures-task",
+ "pin-project-lite",
+ "pin-utils",
 ]
 
 [[package]]
-name = "gemm-c64"
-version = "0.17.1"
+name = "gemm"
+version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fbb5f2e79fefb9693d18e1066a557b4546cd334b226beadc68b11a8f9431852a"
+checksum = "ab96b703d31950f1aeddded248bc95543c9efc7ac9c4a21fda8703a83ee35451"
 dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
+ "dyn-stack",
+ "gemm-c32",
+ "gemm-c64",
+ "gemm-common",
+ "gemm-f16",
+ "gemm-f32",
+ "gemm-f64",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 10.7.0",
+ "raw-cpuid",
  "seq-macro",
 ]
 
 [[package]]
-name = "gemm-c64"
+name = "gemm-c32"
 version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dfcad8a3d35a43758330b635d02edad980c1e143dc2f21e6fd25f9e4eada8edf"
+checksum = "f6db9fd9f40421d00eea9dd0770045a5603b8d684654816637732463f4073847"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
+ "dyn-stack",
+ "gemm-common",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 11.5.0",
+ "raw-cpuid",
  "seq-macro",
 ]
 
 [[package]]
-name = "gemm-common"
-version = "0.17.1"
+name = "gemm-c64"
+version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a2e7ea062c987abcd8db95db917b4ffb4ecdfd0668471d8dc54734fdff2354e8"
+checksum = "dfcad8a3d35a43758330b635d02edad980c1e143dc2f21e6fd25f9e4eada8edf"
 dependencies = [
- "bytemuck",
- "dyn-stack 0.10.0",
- "half",
+ "dyn-stack",
+ "gemm-common",
  "num-complex",
  "num-traits",
- "once_cell",
  "paste",
- "pulp 0.18.22",
- "raw-cpuid 10.7.0",
- "rayon",
+ "raw-cpuid",
  "seq-macro",
- "sysctl 0.5.5",
 ]
 
 [[package]]
@@ -503,36 +1045,18 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a352d4a69cbe938b9e2a9cb7a3a63b7e72f9349174a2752a558a8a563510d0f3"
 dependencies = [
  "bytemuck",
- "dyn-stack 0.13.0",
+ "dyn-stack",
  "half",
  "libm",
  "num-complex",
  "num-traits",
  "once_cell",
  "paste",
- "pulp 0.21.4",
- "raw-cpuid 11.5.0",
- "rayon",
- "seq-macro",
- "sysctl 0.6.0",
-]
-
-[[package]]
-name = "gemm-f16"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7ca4c06b9b11952071d317604acb332e924e817bd891bec8dfb494168c7cedd4"
-dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
- "gemm-f32 0.17.1",
- "half",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 10.7.0",
+ "pulp",
+ "raw-cpuid",
  "rayon",
  "seq-macro",
+ "sysctl",
 ]
 
 [[package]]
@@ -541,60 +1065,30 @@ version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cff95ae3259432f3c3410eaa919033cd03791d81cebd18018393dc147952e109"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
- "gemm-f32 0.18.2",
+ "dyn-stack",
+ "gemm-common",
+ "gemm-f32",
  "half",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 11.5.0",
+ "raw-cpuid",
  "rayon",
  "seq-macro",
 ]
 
-[[package]]
-name = "gemm-f32"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e9a69f51aaefbd9cf12d18faf273d3e982d9d711f60775645ed5c8047b4ae113"
-dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 10.7.0",
- "seq-macro",
-]
-
 [[package]]
 name = "gemm-f32"
 version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bc8d3d4385393304f407392f754cd2dc4b315d05063f62cf09f47b58de276864"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
- "num-complex",
- "num-traits",
- "paste",
- "raw-cpuid 11.5.0",
- "seq-macro",
-]
-
-[[package]]
-name = "gemm-f64"
-version = "0.17.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "aa397a48544fadf0b81ec8741e5c0fba0043008113f71f2034def1935645d2b0"
-dependencies = [
- "dyn-stack 0.10.0",
- "gemm-common 0.17.1",
+ "dyn-stack",
+ "gemm-common",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 10.7.0",
+ "raw-cpuid",
  "seq-macro",
 ]
 
@@ -604,12 +1098,12 @@ version = "0.18.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "35b2a4f76ce4b8b16eadc11ccf2e083252d8237c1b589558a49b0183545015bd"
 dependencies = [
- "dyn-stack 0.13.0",
- "gemm-common 0.18.2",
+ "dyn-stack",
+ "gemm-common",
  "num-complex",
  "num-traits",
  "paste",
- "raw-cpuid 11.5.0",
+ "raw-cpuid",
  "seq-macro",
 ]
 
@@ -628,37 +1122,122 @@ dependencies = [
 name = "getrandom"
 version = "0.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "73fea8450eea4bac3940448fb7ae50d91f034f941199fcd9d909a5a07aa455f0"
+checksum = "73fea8450eea4bac3940448fb7ae50d91f034f941199fcd9d909a5a07aa455f0"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "r-efi",
+ "wasi 0.14.2+wasi-0.2.4",
+]
+
+[[package]]
+name = "gl_generator"
+version = "0.14.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1a95dfc23a2b4a9a2f5ab41d194f8bfda3cabec42af4e39f08c339eb2a0c124d"
+dependencies = [
+ "khronos_api",
+ "log",
+ "xml-rs",
+]
+
+[[package]]
+name = "glow"
+version = "0.16.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c5e5ea60d70410161c8bf5da3fdfeaa1c72ed2c15f8bbb9d19fe3a4fad085f08"
+dependencies = [
+ "js-sys",
+ "slotmap",
+ "wasm-bindgen",
+ "web-sys",
+]
+
+[[package]]
+name = "glutin_wgl_sys"
+version = "0.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2c4ee00b289aba7a9e5306d57c2d05499b2e5dc427f84ac708bd2c090212cf3e"
+dependencies = [
+ "gl_generator",
+]
+
+[[package]]
+name = "gpu-alloc"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fbcd2dba93594b227a1f57ee09b8b9da8892c34d55aa332e034a228d0fe6a171"
+dependencies = [
+ "bitflags 2.9.0",
+ "gpu-alloc-types",
+]
+
+[[package]]
+name = "gpu-alloc-types"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "98ff03b468aa837d70984d55f5d3f846f6ec31fe34bbb97c4f85219caeee1ca4"
+dependencies = [
+ "bitflags 2.9.0",
+]
+
+[[package]]
+name = "gpu-allocator"
+version = "0.27.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c151a2a5ef800297b4e79efa4f4bec035c5f51d5ae587287c9b952bdf734cacd"
+dependencies = [
+ "log",
+ "presser",
+ "thiserror 1.0.61",
+ "windows",
+]
+
+[[package]]
+name = "gpu-descriptor"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dcf29e94d6d243368b7a56caa16bc213e4f9f8ed38c4d9557069527b5d5281ca"
+dependencies = [
+ "bitflags 2.9.0",
+ "gpu-descriptor-types",
+ "hashbrown",
+]
+
+[[package]]
+name = "gpu-descriptor-types"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fdf242682df893b86f33a73828fb09ca4b2d3bb6cc95249707fc684d27484b91"
 dependencies = [
- "cfg-if",
- "libc",
- "r-efi",
- "wasi 0.14.2+wasi-0.2.4",
+ "bitflags 2.9.0",
 ]
 
 [[package]]
 name = "half"
-version = "2.4.1"
+version = "2.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6dd08c532ae367adf81c312a4580bc67f1d0fe8bc9c460520283f4c0ff277888"
+checksum = "459196ed295495a68f7d7fe1d84f6c4b7ff0e21fe3017b2f283c6fac3ad803c9"
 dependencies = [
  "bytemuck",
  "cfg-if",
  "crunchy",
  "num-traits",
- "rand 0.8.5",
- "rand_distr 0.4.3",
+ "rand",
+ "rand_distr",
+ "serde",
 ]
 
 [[package]]
 name = "hashbrown"
-version = "0.15.2"
+version = "0.15.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bf151400ff0baff5465007dd2f3e717f3fe502074ca563069ce3a6629d07b289"
+checksum = "9229cfe53dfd69f0609a49f65461bd93001ea1ef889cd5529dd176593f5338a1"
 dependencies = [
  "allocator-api2",
  "equivalent",
  "foldhash",
+ "serde",
 ]
 
 [[package]]
@@ -679,6 +1258,18 @@ version = "0.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "fbd780fe5cc30f81464441920d82ac8740e2e46b29a6fad543ddd075229ce37e"
 
+[[package]]
+name = "hexf-parse"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dfa686283ad6dd069f105e5ab091b04c62850d3e4cf5d67debad1933f55023df"
+
+[[package]]
+name = "ident_case"
+version = "1.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b9e0384b61958566e926dc50660321d12159025e767c18e043daf26b70104c39"
+
 [[package]]
 name = "indexmap"
 version = "2.9.0"
@@ -715,6 +1306,12 @@ version = "1.0.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4a5f13b858c8d314ee3e8f639011f7ccefe71f97f96e50151fb991f267928e2c"
 
+[[package]]
+name = "jni-sys"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8eaf4bc02d17cbdd7ff4c7438cafcdf7fb9a4613313ad11b4f8fefe7d3fa0130"
+
 [[package]]
 name = "js-sys"
 version = "0.3.77"
@@ -725,11 +1322,28 @@ dependencies = [
  "wasm-bindgen",
 ]
 
+[[package]]
+name = "khronos-egl"
+version = "6.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6aae1df220ece3c0ada96b8153459b67eebe9ae9212258bb0134ae60416fdf76"
+dependencies = [
+ "libc",
+ "libloading",
+ "pkg-config",
+]
+
+[[package]]
+name = "khronos_api"
+version = "3.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e2db585e1d738fc771bf08a151420d3ed193d9d895a36df7f6f8a9456b911ddc"
+
 [[package]]
 name = "libc"
-version = "0.2.155"
+version = "0.2.172"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "97b3888a4aecf77e811145cadf6eef5901f4782c53886191b2f693f24761847c"
+checksum = "d750af042f7ef4f724306de029d18836c26c1765a54a6a3f094cbd23a7267ffa"
 
 [[package]]
 name = "libloading"
@@ -753,94 +1367,124 @@ version = "0.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c0ff37bd590ca25063e35af745c343cb7a0271906fb7b37e4813e8f79f00268d"
 dependencies = [
- "bitflags 2.5.0",
+ "bitflags 2.9.0",
  "libc",
 ]
 
 [[package]]
-name = "log"
-version = "0.4.27"
+name = "litrs"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "13dc2df351e3202783a1fe0d44375f7295ffb4049267b0f3018346dc122a1d94"
+checksum = "b4ce301924b7887e9d637144fdade93f9dfff9b60981d4ac161db09720d39aa5"
 
 [[package]]
-name = "memchr"
-version = "2.7.4"
+name = "lock_api"
+version = "0.4.12"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "78ca9ab1a0babb1e7d5695e3530886289c18cf2f87ec19a575a0abdce112e3a3"
+checksum = "07af8b9cdd281b7915f413fa73f29ebd5d55d0d3f0155584dade1ff18cea1b17"
+dependencies = [
+ "autocfg",
+ "scopeguard",
+]
 
 [[package]]
-name = "memmap2"
-version = "0.9.5"
+name = "log"
+version = "0.4.27"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "13dc2df351e3202783a1fe0d44375f7295ffb4049267b0f3018346dc122a1d94"
+
+[[package]]
+name = "malloc_buf"
+version = "0.0.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fd3f7eed9d3848f8b98834af67102b720745c4ec028fcd0aa0239277e7de374f"
+checksum = "62bb907fe88d54d8d9ce32a3cceab4218ed2f6b7d35617cafe9adf84e43919cb"
 dependencies = [
  "libc",
- "stable_deref_trait",
 ]
 
 [[package]]
-name = "num"
-version = "0.4.3"
+name = "md5"
+version = "0.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "35bd024e8b2ff75562e5f34e7f4905839deb4b22955ef5e73d2fea1b9813cb23"
-dependencies = [
- "num-bigint",
- "num-complex",
- "num-integer",
- "num-iter",
- "num-rational",
- "num-traits",
-]
+checksum = "ae960838283323069879657ca3de837e9f7bbb4c7bf6ea7f1b290d5e9476d2e0"
 
 [[package]]
-name = "num-bigint"
-version = "0.4.6"
+name = "memchr"
+version = "2.7.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "78ca9ab1a0babb1e7d5695e3530886289c18cf2f87ec19a575a0abdce112e3a3"
+
+[[package]]
+name = "metal"
+version = "0.31.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a5e44f723f1133c9deac646763579fdb3ac745e418f2a7af9cd0c431da1f20b9"
+checksum = "f569fb946490b5743ad69813cb19629130ce9374034abe31614a36402d18f99e"
 dependencies = [
- "num-integer",
- "num-traits",
+ "bitflags 2.9.0",
+ "block",
+ "core-graphics-types",
+ "foreign-types",
+ "log",
+ "objc",
+ "paste",
 ]
 
 [[package]]
-name = "num-complex"
-version = "0.4.6"
+name = "naga"
+version = "25.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "73f88a1307638156682bada9d7604135552957b7818057dcef22705b4d509495"
+checksum = "2b977c445f26e49757f9aca3631c3b8b836942cb278d69a92e7b80d3b24da632"
 dependencies = [
- "bytemuck",
+ "arrayvec",
+ "bit-set",
+ "bitflags 2.9.0",
+ "cfg_aliases",
+ "codespan-reporting",
+ "half",
+ "hashbrown",
+ "hexf-parse",
+ "indexmap",
+ "log",
  "num-traits",
+ "once_cell",
+ "rustc-hash",
+ "spirv",
+ "strum",
+ "thiserror 2.0.12",
+ "unicode-ident",
 ]
 
 [[package]]
-name = "num-integer"
-version = "0.1.46"
+name = "nb"
+version = "0.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7969661fd2958a5cb096e56c8e1ad0444ac2bbcd0061bd28660485a44879858f"
+checksum = "801d31da0513b6ec5214e9bf433a77966320625a37860f910be265be6e18d06f"
 dependencies = [
- "num-traits",
+ "nb 1.1.0",
 ]
 
 [[package]]
-name = "num-iter"
-version = "0.1.45"
+name = "nb"
+version = "1.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1429034a0490724d0075ebb2bc9e875d6503c3cf69e235a8941aa757d83ef5bf"
+checksum = "8d5439c4ad607c3c23abf66de8c8bf57ba8adcd1f129e699851a6e43935d339d"
+
+[[package]]
+name = "ndk-sys"
+version = "0.5.0+25.2.9519653"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8c196769dd60fd4f363e11d948139556a344e79d451aeb2fa2fd040738ef7691"
 dependencies = [
- "autocfg",
- "num-integer",
- "num-traits",
+ "jni-sys",
 ]
 
 [[package]]
-name = "num-rational"
-version = "0.4.2"
+name = "num-complex"
+version = "0.4.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f83d14da390562dca69fc84082e73e548e1ad308d24accdedd2720017cb37824"
+checksum = "73f88a1307638156682bada9d7604135552957b7818057dcef22705b4d509495"
 dependencies = [
- "num-bigint",
- "num-integer",
+ "bytemuck",
  "num-traits",
 ]
 
@@ -865,24 +1509,12 @@ dependencies = [
 ]
 
 [[package]]
-name = "num_enum"
-version = "0.7.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4e613fc340b2220f734a8595782c551f1250e969d87d3be1ae0579e8d4065179"
-dependencies = [
- "num_enum_derive",
-]
-
-[[package]]
-name = "num_enum_derive"
-version = "0.7.3"
+name = "objc"
+version = "0.2.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "af1844ef2428cc3e1cb900be36181049ef3d3193c63e43026cfe202983b27a56"
+checksum = "915b1b472bc21c53464d6c8461c9d3af805ba1ef837e1cac254428f4a77177b1"
 dependencies = [
- "proc-macro-crate",
- "proc-macro2",
- "quote",
- "syn",
+ "malloc_buf",
 ]
 
 [[package]]
@@ -903,6 +1535,44 @@ version = "0.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "04744f49eae99ab78e0d5c0b603ab218f515ea8cfe5a456d7629ad883a3b6e7d"
 
+[[package]]
+name = "ordered-float"
+version = "4.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7bb71e1b3fa6ca1c61f383464aaf2bb0e2f8e772a1f01d486832464de363b951"
+dependencies = [
+ "num-traits",
+]
+
+[[package]]
+name = "parking"
+version = "2.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f38d5652c16fde515bb1ecef450ab0f6a219d619a7274976324d5e377f7dceba"
+
+[[package]]
+name = "parking_lot"
+version = "0.12.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f1bf18183cf54e8d6059647fc3063646a1801cf30896933ec2311622cc4b9a27"
+dependencies = [
+ "lock_api",
+ "parking_lot_core",
+]
+
+[[package]]
+name = "parking_lot_core"
+version = "0.9.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e401f977ab385c9e4e3ab30627d6f26d00e2c73eef317493c4ec6d468726cf8"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "redox_syscall",
+ "smallvec",
+ "windows-targets 0.52.6",
+]
+
 [[package]]
 name = "paste"
 version = "1.0.15"
@@ -927,6 +1597,18 @@ version = "0.2.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3b3cff922bd51709b605d9ead9aa71031d81447142d828eb4a6eba76fe619f9b"
 
+[[package]]
+name = "pin-utils"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"
+
+[[package]]
+name = "pkg-config"
+version = "0.3.32"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7edddbd0b52d732b21ad9a5fab5c704c14cd949e5e9a1ec5929a24fded1b904c"
+
 [[package]]
 name = "plotters"
 version = "0.3.7"
@@ -955,6 +1637,15 @@ dependencies = [
  "plotters-backend",
 ]
 
+[[package]]
+name = "portable-atomic"
+version = "1.11.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "350e9b48cbc6b0e028b0473b114454c6316e57336ee184ceab6e53f72c178b3e"
+dependencies = [
+ "serde",
+]
+
 [[package]]
 name = "ppv-lite86"
 version = "0.2.21"
@@ -965,34 +1656,35 @@ dependencies = [
 ]
 
 [[package]]
-name = "proc-macro-crate"
-version = "3.3.0"
+name = "presser"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e8cf8e6a8aa66ce33f63993ffc4ea4271eb5b0530a9002db8455ea6050c77bfa"
+
+[[package]]
+name = "prettyplease"
+version = "0.2.32"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "edce586971a4dfaa28950c6f18ed55e0406c1ab88bbce2c6f6293a7aaba73d35"
+checksum = "664ec5419c51e34154eec046ebcba56312d5a2fc3b09a06da188e1ad21afadf6"
 dependencies = [
- "toml_edit",
+ "proc-macro2",
+ "syn",
 ]
 
 [[package]]
 name = "proc-macro2"
-version = "1.0.85"
+version = "1.0.95"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "22244ce15aa966053a896d1accb3a6e68469b97c7f33f284b99f0d576879fc23"
+checksum = "02b3e5e68a3a1a02aad3ec490a98007cbc13c37cbe84a3cd7b8e406d76e7f778"
 dependencies = [
  "unicode-ident",
 ]
 
 [[package]]
-name = "pulp"
-version = "0.18.22"
+name = "profiling"
+version = "1.0.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a0a01a0dc67cf4558d279f0c25b0962bd08fc6dec0137699eae304103e882fe6"
-dependencies = [
- "bytemuck",
- "libm",
- "num-complex",
- "reborrow",
-]
+checksum = "afbdc74edc00b6f6a218ca6a5364d6226a259d4b8ea1af4a0ea063f27e179f4d"
 
 [[package]]
 name = "pulp"
@@ -1023,35 +1715,14 @@ version = "5.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "74765f6d916ee2faa39bc8e68e4f3ed8949b48cccdac59983d287a7cb71ce9c5"
 
-[[package]]
-name = "rand"
-version = "0.8.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "34af8d1a0e25924bc5b7c43c079c942339d8f0a8b57c39049bef581b46327404"
-dependencies = [
- "libc",
- "rand_chacha 0.3.1",
- "rand_core 0.6.4",
-]
-
 [[package]]
 name = "rand"
 version = "0.9.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9fbfd9d094a40bf3ae768db9361049ace4c0e04a4fd6b359518bd7b73a73dd97"
 dependencies = [
- "rand_chacha 0.9.0",
- "rand_core 0.9.3",
-]
-
-[[package]]
-name = "rand_chacha"
-version = "0.3.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e6c10a63a0fa32252be49d21e7709d4d4baf8d231c2dbce1eaa8141b9b127d88"
-dependencies = [
- "ppv-lite86",
- "rand_core 0.6.4",
+ "rand_chacha",
+ "rand_core",
 ]
 
 [[package]]
@@ -1061,16 +1732,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d3022b5f1df60f26e1ffddd6c66e8aa15de382ae63b3a0c1bfc0e4d3e3f325cb"
 dependencies = [
  "ppv-lite86",
- "rand_core 0.9.3",
-]
-
-[[package]]
-name = "rand_core"
-version = "0.6.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ec0be4795e2f6a28069bec0b5ff3e2ac9bafc99e6a9a7dc3547996c5c816922c"
-dependencies = [
- "getrandom 0.2.15",
+ "rand_core",
 ]
 
 [[package]]
@@ -1082,16 +1744,6 @@ dependencies = [
  "getrandom 0.3.2",
 ]
 
-[[package]]
-name = "rand_distr"
-version = "0.4.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "32cb0b9bc82b0a0876c2dd994a7e7a2683d3e7390ca40e6886785ef0c7e3ee31"
-dependencies = [
- "num-traits",
- "rand 0.8.5",
-]
-
 [[package]]
 name = "rand_distr"
 version = "0.5.1"
@@ -1099,17 +1751,14 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6a8615d50dcf34fa31f7ab52692afec947c4dd0ab803cc87cb3b0b4570ff7463"
 dependencies = [
  "num-traits",
- "rand 0.9.1",
+ "rand",
 ]
 
 [[package]]
-name = "raw-cpuid"
-version = "10.7.0"
+name = "range-alloc"
+version = "0.1.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6c297679cb867470fa8c9f67dbba74a78d78e3e98d7cf2b08d6d71540f797332"
-dependencies = [
- "bitflags 1.3.2",
-]
+checksum = "c3d6831663a5098ea164f89cff59c6284e95f4e3c76ce9848d4529f5ccca9bde"
 
 [[package]]
 name = "raw-cpuid"
@@ -1117,9 +1766,15 @@ version = "11.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c6df7ab838ed27997ba19a4664507e6f82b41fe6e20be42929332156e5e85146"
 dependencies = [
- "bitflags 2.5.0",
+ "bitflags 2.9.0",
 ]
 
+[[package]]
+name = "raw-window-handle"
+version = "0.6.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "20675572f6f24e9e76ef639bc5552774ed45f1c30e2951e1e99c59888861c539"
+
 [[package]]
 name = "rayon"
 version = "1.10.0"
@@ -1146,6 +1801,15 @@ version = "0.5.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "03251193000f4bd3b042892be858ee50e8b3719f2b08e5833ac4353724632430"
 
+[[package]]
+name = "redox_syscall"
+version = "0.5.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d2f103c6d277498fbceb16e84d317e2a400f160f46904d5f5410848c829511a3"
+dependencies = [
+ "bitflags 2.9.0",
+]
+
 [[package]]
 name = "redox_users"
 version = "0.4.5"
@@ -1154,7 +1818,18 @@ checksum = "bd283d9651eeda4b2a83a43c1c91b266c40fd76ecd39a50a8c630ae69dc72891"
 dependencies = [
  "getrandom 0.2.15",
  "libredox",
- "thiserror",
+ "thiserror 1.0.61",
+]
+
+[[package]]
+name = "redox_users"
+version = "0.5.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a4e608c6638b9c18977b00b475ac1f28d14e84b27d8d42f70e0bf1e3dec127ac"
+dependencies = [
+ "getrandom 0.2.15",
+ "libredox",
+ "thiserror 2.0.12",
 ]
 
 [[package]]
@@ -1186,6 +1861,18 @@ version = "0.8.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2b15c43186be67a4fd63bee50d0303afffcef381492ebe2c5d87f324e1b8815c"
 
+[[package]]
+name = "renderdoc-sys"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "19b30a45b0cd0bcca8037f3d0dc3421eaf95327a17cad11964fb8179b4fc4832"
+
+[[package]]
+name = "rustc-hash"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "08d43f7aa6b08d49f382cde6a7982047c3426db949b1424bc4b7ec9ae12c6ce2"
+
 [[package]]
 name = "rustversion"
 version = "1.0.20"
@@ -1199,24 +1886,29 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "28d3b2b1366ec20994f1fd18c3c594f05c5dd4bc44d8bb0c1c632c8d6829481f"
 
 [[package]]
-name = "safetensors"
-version = "0.4.5"
+name = "same-file"
+version = "1.0.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "44560c11236a6130a46ce36c836a62936dc81ebf8c36a37947423571be0e55b6"
+checksum = "93fc1dc3aaa9bfed95e02e6eadabb4baf7e3078b0bd1b4d7b6b0b68378900502"
 dependencies = [
- "serde",
- "serde_json",
+ "winapi-util",
 ]
 
 [[package]]
-name = "same-file"
-version = "1.0.6"
+name = "sanitize-filename"
+version = "0.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "93fc1dc3aaa9bfed95e02e6eadabb4baf7e3078b0bd1b4d7b6b0b68378900502"
+checksum = "bc984f4f9ceb736a7bb755c3e3bd17dc56370af2600c9780dcc48c66453da34d"
 dependencies = [
- "winapi-util",
+ "regex",
 ]
 
+[[package]]
+name = "scopeguard"
+version = "1.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"
+
 [[package]]
 name = "seq-macro"
 version = "0.3.6"
@@ -1229,72 +1921,118 @@ version = "1.0.210"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c8e3592472072e6e22e0a54d5904d9febf8508f65fb8552499a1abc7d1078c3a"
 dependencies = [
- "serde_derive",
+ "serde_derive",
+]
+
+[[package]]
+name = "serde_derive"
+version = "1.0.210"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "243902eda00fad750862fc144cea25caca5e20d615af0a81bee94ca738f1df1f"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "serde_json"
+version = "1.0.140"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "20068b6e96dc6c9bd23e01df8827e6c7e1f2fddd43c21810382803c136b99373"
+dependencies = [
+ "itoa",
+ "memchr",
+ "ryu",
+ "serde",
+]
+
+[[package]]
+name = "serde_spanned"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "40734c41988f7306bb04f0ecf60ec0f3f1caa34290e4e8ea471dcd3346483b83"
+dependencies = [
+ "serde",
+]
+
+[[package]]
+name = "slotmap"
+version = "1.0.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dbff4acf519f630b3a3ddcfaea6c06b42174d9a44bc70c620e9ed1649d58b82a"
+dependencies = [
+ "version_check",
+]
+
+[[package]]
+name = "smallvec"
+version = "1.15.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8917285742e9f3e1683f0a9c4e6b57960b7314d0b08d30d1ecd426713ee2eee9"
+
+[[package]]
+name = "spin"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d5fe4ccb98d9c292d56fec89a5e07da7fc4cf0dc11e156b41793132775d3e591"
+dependencies = [
+ "lock_api",
+ "portable-atomic",
 ]
 
 [[package]]
-name = "serde_derive"
-version = "1.0.210"
+name = "spirv"
+version = "0.3.0+sdk-1.3.268.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "243902eda00fad750862fc144cea25caca5e20d615af0a81bee94ca738f1df1f"
+checksum = "eda41003dc44290527a59b13432d4a0379379fa074b70174882adfbdfd917844"
 dependencies = [
- "proc-macro2",
- "quote",
- "syn",
+ "bitflags 2.9.0",
 ]
 
 [[package]]
-name = "serde_json"
-version = "1.0.140"
+name = "static_assertions"
+version = "1.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "20068b6e96dc6c9bd23e01df8827e6c7e1f2fddd43c21810382803c136b99373"
-dependencies = [
- "itoa",
- "memchr",
- "ryu",
- "serde",
-]
+checksum = "a2eb9349b6444b326872e140eb1cf5e7c522154d69e7a0ffb0fb81c06b37543f"
 
 [[package]]
-name = "stable_deref_trait"
-version = "1.2.0"
+name = "strsim"
+version = "0.11.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a8f112729512f8e442d81f95a8a7ddf2b7c6b8a1a6f509a95864142b30cab2d3"
+checksum = "7da8b5736845d9f2fcb837ea5d9e2628564b3b043a70948a3f0b778838c5fb4f"
 
 [[package]]
-name = "syn"
-version = "2.0.66"
+name = "strum"
+version = "0.26.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c42f3f41a2de00b01c0aaad383c5a45241efc8b2d1eda5661812fda5f3cdcff5"
+checksum = "8fec0f0aef304996cf250b31b5a10dee7980c85da9d759361292b8bca5a18f06"
 dependencies = [
- "proc-macro2",
- "quote",
- "unicode-ident",
+ "strum_macros",
 ]
 
 [[package]]
-name = "synstructure"
-version = "0.13.1"
+name = "strum_macros"
+version = "0.26.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c8af7666ab7b6390ab78131fb5b0fce11d6b7a6951602017c35fa82800708971"
+checksum = "4c6bee85a5a24955dc440386795aa378cd9cf82acd5f764469152d2270e581be"
 dependencies = [
+ "heck",
  "proc-macro2",
  "quote",
+ "rustversion",
  "syn",
 ]
 
 [[package]]
-name = "sysctl"
-version = "0.5.5"
+name = "syn"
+version = "2.0.100"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ec7dddc5f0fee506baf8b9fdb989e242f17e4b11c61dfbb0635b705217199eea"
+checksum = "b09a44accad81e1ba1cd74a32461ba89dee89095ba17b32f5d03683b1b1fc2a0"
 dependencies = [
- "bitflags 2.5.0",
- "byteorder",
- "enum-as-inner",
- "libc",
- "thiserror",
- "walkdir",
+ "proc-macro2",
+ "quote",
+ "unicode-ident",
 ]
 
 [[package]]
@@ -1303,21 +2041,39 @@ version = "0.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "01198a2debb237c62b6826ec7081082d951f46dbb64b0e8c7649a452230d1dfc"
 dependencies = [
- "bitflags 2.5.0",
+ "bitflags 2.9.0",
  "byteorder",
  "enum-as-inner",
  "libc",
- "thiserror",
+ "thiserror 1.0.61",
  "walkdir",
 ]
 
+[[package]]
+name = "termcolor"
+version = "1.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "06794f8f6c5c898b3275aebefa6b8a1cb24cd2c6c79397ab15774837a0bc5755"
+dependencies = [
+ "winapi-util",
+]
+
 [[package]]
 name = "thiserror"
 version = "1.0.61"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c546c80d6be4bc6a00c0f01730c08df82eaa7a7a61f11d656526506112cc1709"
 dependencies = [
- "thiserror-impl",
+ "thiserror-impl 1.0.61",
+]
+
+[[package]]
+name = "thiserror"
+version = "2.0.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "567b8a2dae586314f7be2a752ec7474332959c6460e02bde30d702a66d488708"
+dependencies = [
+ "thiserror-impl 2.0.12",
 ]
 
 [[package]]
@@ -1331,6 +2087,17 @@ dependencies = [
  "syn",
 ]
 
+[[package]]
+name = "thiserror-impl"
+version = "2.0.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f7cf42b4507d8ea322120659672cf1b9dbb93f8f2d4ecfd6e51350ff5b17a1d"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
 [[package]]
 name = "tinytemplate"
 version = "1.2.1"
@@ -1342,73 +2109,43 @@ dependencies = [
 ]
 
 [[package]]
-name = "toml_datetime"
-version = "0.6.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3da5db5a963e24bc68be8b17b6fa82814bb22ee8660f192bb182771d498f09a3"
-
-[[package]]
-name = "toml_edit"
-version = "0.22.25"
+name = "toml"
+version = "0.9.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "10558ed0bd2a1562e630926a2d1f0b98c827da99fabd3fe20920a59642504485"
+checksum = "75129e1dc5000bfbaa9fee9d1b21f974f9fbad9daec557a521ee6e080825f6e8"
 dependencies = [
  "indexmap",
+ "serde",
+ "serde_spanned",
  "toml_datetime",
+ "toml_parser",
+ "toml_writer",
  "winnow",
 ]
 
 [[package]]
-name = "tracing"
-version = "0.1.41"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "784e0ac535deb450455cbfa28a6f0df145ea1bb7ae51b821cf5e7927fdcfbdd0"
-dependencies = [
- "pin-project-lite",
- "tracing-attributes",
- "tracing-core",
-]
-
-[[package]]
-name = "tracing-attributes"
-version = "0.1.28"
+name = "toml_datetime"
+version = "0.7.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "395ae124c09f9e6918a2310af6038fba074bcf474ac352496d5910dd59a2226d"
+checksum = "bade1c3e902f58d73d3f294cd7f20391c1cb2fbcb643b73566bc773971df91e3"
 dependencies = [
- "proc-macro2",
- "quote",
- "syn",
+ "serde",
 ]
 
 [[package]]
-name = "tracing-core"
-version = "0.1.33"
+name = "toml_parser"
+version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e672c95779cf947c5311f83787af4fa8fffd12fb27e4993211a84bdfd9610f9c"
+checksum = "b551886f449aa90d4fe2bdaa9f4a2577ad2dde302c61ecf262d80b116db95c10"
 dependencies = [
- "once_cell",
+ "winnow",
 ]
 
 [[package]]
-name = "ug"
-version = "0.1.0"
+name = "toml_writer"
+version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "03719c61a91b51541f076dfdba45caacf750b230cefaa4b32d6f5411c3f7f437"
-dependencies = [
- "gemm 0.18.2",
- "half",
- "libloading",
- "memmap2",
- "num",
- "num-traits",
- "num_cpus",
- "rayon",
- "safetensors",
- "serde",
- "thiserror",
- "tracing",
- "yoke",
-]
+checksum = "fcc842091f2def52017664b53082ecbbeb5c7731092bad69d2c63050401dfd64"
 
 [[package]]
 name = "unicode-ident"
@@ -1416,12 +2153,41 @@ version = "1.0.12"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3354b9ac3fae1ff6755cb6db53683adb661634f67557942dea4facebec0fee4b"
 
+[[package]]
+name = "unicode-width"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1fc81956842c57dac11422a97c3b8195a1ff727f06e85c84ed2e8aa277c9a0fd"
+
+[[package]]
+name = "unicode-xid"
+version = "0.2.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ebc1c04c71510c7f702b52b7c350734c9ff1295c464a03335b00bb84fc54f853"
+
+[[package]]
+name = "variadics_please"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "41b6d82be61465f97d42bd1d15bf20f3b0a3a0905018f38f9d6f6962055b0b5c"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
 [[package]]
 name = "version_check"
 version = "0.9.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
 
+[[package]]
+name = "void"
+version = "1.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6a02e4885ed3bc0f2de90ea6dd45ebcbb66dacffe03547fadbb0eeae2770887d"
+
 [[package]]
 name = "walkdir"
 version = "2.5.0"
@@ -1473,6 +2239,19 @@ dependencies = [
  "wasm-bindgen-shared",
 ]
 
+[[package]]
+name = "wasm-bindgen-futures"
+version = "0.4.50"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "555d470ec0bc3bb57890405e5d4322cc9ea83cebb085523ced7be4144dac1e61"
+dependencies = [
+ "cfg-if",
+ "js-sys",
+ "once_cell",
+ "wasm-bindgen",
+ "web-sys",
+]
+
 [[package]]
 name = "wasm-bindgen-macro"
 version = "0.2.100"
@@ -1515,6 +2294,163 @@ dependencies = [
  "wasm-bindgen",
 ]
 
+[[package]]
+name = "web-time"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5a6580f308b1fad9207618087a65c04e7a10bc77e02c8e84e9b00dd4b12fa0bb"
+dependencies = [
+ "js-sys",
+ "wasm-bindgen",
+]
+
+[[package]]
+name = "wgpu"
+version = "25.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ca6049eb2014a0e0d8689f9b787605dd71d5bbfdc74095ead499f3cff705c229"
+dependencies = [
+ "arrayvec",
+ "bitflags 2.9.0",
+ "cfg_aliases",
+ "document-features",
+ "hashbrown",
+ "js-sys",
+ "log",
+ "naga",
+ "parking_lot",
+ "portable-atomic",
+ "profiling",
+ "raw-window-handle",
+ "smallvec",
+ "static_assertions",
+ "wasm-bindgen",
+ "wasm-bindgen-futures",
+ "web-sys",
+ "wgpu-core",
+ "wgpu-hal",
+ "wgpu-types",
+]
+
+[[package]]
+name = "wgpu-core"
+version = "25.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a19813e647da7aa3cdaa84f5846e2c64114970ea7c86b1e6aae8be08091f4bdc"
+dependencies = [
+ "arrayvec",
+ "bit-set",
+ "bit-vec",
+ "bitflags 2.9.0",
+ "cfg_aliases",
+ "document-features",
+ "hashbrown",
+ "indexmap",
+ "log",
+ "naga",
+ "once_cell",
+ "parking_lot",
+ "portable-atomic",
+ "profiling",
+ "raw-window-handle",
+ "rustc-hash",
+ "smallvec",
+ "thiserror 2.0.12",
+ "wgpu-core-deps-apple",
+ "wgpu-core-deps-emscripten",
+ "wgpu-core-deps-windows-linux-android",
+ "wgpu-hal",
+ "wgpu-types",
+]
+
+[[package]]
+name = "wgpu-core-deps-apple"
+version = "25.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cfd488b3239b6b7b185c3b045c39ca6bf8af34467a4c5de4e0b1a564135d093d"
+dependencies = [
+ "wgpu-hal",
+]
+
+[[package]]
+name = "wgpu-core-deps-emscripten"
+version = "25.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f09ad7aceb3818e52539acc679f049d3475775586f3f4e311c30165cf2c00445"
+dependencies = [
+ "wgpu-hal",
+]
+
+[[package]]
+name = "wgpu-core-deps-windows-linux-android"
+version = "25.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cba5fb5f7f9c98baa7c889d444f63ace25574833df56f5b817985f641af58e46"
+dependencies = [
+ "wgpu-hal",
+]
+
+[[package]]
+name = "wgpu-hal"
+version = "25.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fb7c4a1dc42ff14c23c9b11ebf1ee85cde661a9b1cf0392f79c1faca5bc559fb"
+dependencies = [
+ "android_system_properties",
+ "arrayvec",
+ "ash",
+ "bit-set",
+ "bitflags 2.9.0",
+ "block",
+ "bytemuck",
+ "cfg-if",
+ "cfg_aliases",
+ "core-graphics-types",
+ "glow",
+ "glutin_wgl_sys",
+ "gpu-alloc",
+ "gpu-allocator",
+ "gpu-descriptor",
+ "hashbrown",
+ "js-sys",
+ "khronos-egl",
+ "libc",
+ "libloading",
+ "log",
+ "metal",
+ "naga",
+ "ndk-sys",
+ "objc",
+ "ordered-float",
+ "parking_lot",
+ "portable-atomic",
+ "profiling",
+ "range-alloc",
+ "raw-window-handle",
+ "renderdoc-sys",
+ "smallvec",
+ "thiserror 2.0.12",
+ "wasm-bindgen",
+ "web-sys",
+ "wgpu-types",
+ "windows",
+ "windows-core",
+]
+
+[[package]]
+name = "wgpu-types"
+version = "25.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2aa49460c2a8ee8edba3fca54325540d904dd85b2e086ada762767e17d06e8bc"
+dependencies = [
+ "bitflags 2.9.0",
+ "bytemuck",
+ "js-sys",
+ "log",
+ "thiserror 2.0.12",
+ "web-sys",
+]
+
 [[package]]
 name = "winapi-util"
 version = "0.1.9"
@@ -1524,6 +2460,70 @@ dependencies = [
  "windows-sys 0.59.0",
 ]
 
+[[package]]
+name = "windows"
+version = "0.58.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dd04d41d93c4992d421894c18c8b43496aa748dd4c081bac0dc93eb0489272b6"
+dependencies = [
+ "windows-core",
+ "windows-targets 0.52.6",
+]
+
+[[package]]
+name = "windows-core"
+version = "0.58.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6ba6d44ec8c2591c134257ce647b7ea6b20335bf6379a27dac5f1641fcf59f99"
+dependencies = [
+ "windows-implement",
+ "windows-interface",
+ "windows-result",
+ "windows-strings",
+ "windows-targets 0.52.6",
+]
+
+[[package]]
+name = "windows-implement"
+version = "0.58.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2bbd5b46c938e506ecbce286b6628a02171d56153ba733b6c741fc627ec9579b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "windows-interface"
+version = "0.58.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "053c4c462dc91d3b1504c6fe5a726dd15e216ba718e84a0e46a88fbe5ded3515"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "windows-result"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1d1043d8214f791817bab27572aaa8af63732e11bf84aa21a45a78d6c317ae0e"
+dependencies = [
+ "windows-targets 0.52.6",
+]
+
+[[package]]
+name = "windows-strings"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4cd9b125c486025df0eabcb585e62173c6c9eddcec5d117d3b6e8c30e2ee4d10"
+dependencies = [
+ "windows-result",
+ "windows-targets 0.52.6",
+]
+
 [[package]]
 name = "windows-sys"
 version = "0.48.0"
@@ -1665,12 +2665,9 @@ checksum = "589f6da84c646204747d1270a2a5661ea66ed1cced2631d546fdfb155959f9ec"
 
 [[package]]
 name = "winnow"
-version = "0.7.7"
+version = "0.7.12"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6cb8234a863ea0e8cd7284fcdd4f145233eb00fee02bbdd9861aec44e6477bc5"
-dependencies = [
- "memchr",
-]
+checksum = "f3edebf492c8125044983378ecb5766203ad3b4c2f7a922bd7dd207f6d443e95"
 
 [[package]]
 name = "wit-bindgen-rt"
@@ -1678,32 +2675,14 @@ version = "0.39.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6f42320e61fe2cfd34354ecb597f86f413484a798ba44a8ca1165c58d42da6c1"
 dependencies = [
- "bitflags 2.5.0",
-]
-
-[[package]]
-name = "yoke"
-version = "0.7.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "120e6aef9aa629e3d4f52dc8cc43a015c7724194c97dfaf45180d2daf2b77f40"
-dependencies = [
- "serde",
- "stable_deref_trait",
- "yoke-derive",
- "zerofrom",
+ "bitflags 2.9.0",
 ]
 
 [[package]]
-name = "yoke-derive"
-version = "0.7.5"
+name = "xml-rs"
+version = "0.8.26"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2380878cad4ac9aac1e2435f3eb4020e8374b5f13c296cb75b4620ff8e229154"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
- "synstructure",
-]
+checksum = "a62ce76d9b56901b19a74f19431b0d8b3bc7ca4ad685a746dfd78ca8f4fc6bda"
 
 [[package]]
 name = "zerocopy"
@@ -1724,39 +2703,3 @@ dependencies = [
  "quote",
  "syn",
 ]
-
-[[package]]
-name = "zerofrom"
-version = "0.1.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "50cc42e0333e05660c3587f3bf9d0478688e15d870fab3346451ce7f8c9fbea5"
-dependencies = [
- "zerofrom-derive",
-]
-
-[[package]]
-name = "zerofrom-derive"
-version = "0.1.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d71e5d6e06ab090c67b5e44993ec16b72dcbaabc526db883a360057678b48502"
-dependencies = [
- "proc-macro2",
- "quote",
- "syn",
- "synstructure",
-]
-
-[[package]]
-name = "zip"
-version = "1.1.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9cc23c04387f4da0374be4533ad1208cbb091d5c11d070dfef13676ad6497164"
-dependencies = [
- "arbitrary",
- "crc32fast",
- "crossbeam-utils",
- "displaydoc",
- "indexmap",
- "num_enum",
- "thiserror",
-]
diff --git a/Cargo.toml b/Cargo.toml
index c4078c6..521cff8 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -24,3 +24,4 @@ gemm = "0.18"
 num_cpus = "1.16.0"
 rand = "0.9.1"
 rand_distr = "0.5.1"
+cubecl = "0.6.0"
diff --git a/README.md b/README.md
index 1df586c..faafcc0 100644
--- a/README.md
+++ b/README.md
@@ -16,6 +16,11 @@ Experimental machine learning framework featuring a graph-based JIT compiler.
 - Multi-device support (CPU, optional CUDA)
 - Graph visualization (requires Graphviz)
 - Zero-cost abstractions with idiomatic Rust API
+- Broad GPU support:
+  - CUDA (NVIDIA): `--features cuda`
+  - HIP (AMD): `--features hip`
+  - Metal (Apple): `--features metal`
+  - WGPU (recommended for [all others](https://github.com/gfx-rs/wgpu?tab=readme-ov-file#supported-platforms)): `--features wgpu`
 
 ## Installation
 
diff --git a/constensor-core/Cargo.toml b/constensor-core/Cargo.toml
index 122fa0d..0a33b91 100644
--- a/constensor-core/Cargo.toml
+++ b/constensor-core/Cargo.toml
@@ -20,22 +20,19 @@ gemm.workspace = true
 num_cpus.workspace = true
 rand.workspace = true
 rand_distr.workspace = true
+cubecl = { workspace = true, features = ["compilation-cache", "wgpu"] }
 
 [features]
-default = []
-all = ["cuda", "half", "bfloat"]
-cuda = ["cudarc"]
+default = ["half", "bfloat"]
+cuda = ["cudarc", "cubecl/cuda"]
+hip = ["cubecl/hip"]
+metal = ["cubecl/wgpu-msl"]
 half = ["dep:half"]
 bfloat = ["dep:half"]
 slow_integral_fma_cuda = []
 
-[[example]]
-name = "hello_world"
-required-features = []
-  
 [dev-dependencies]
 criterion = "0.5"
-candle-core = "0.8"
 
 [[bench]]
 name = "cpu_graph"
diff --git a/constensor-core/benches/cpu_graph.rs b/constensor-core/benches/cpu_graph.rs
index 904e1e3..129b361 100644
--- a/constensor-core/benches/cpu_graph.rs
+++ b/constensor-core/benches/cpu_graph.rs
@@ -1,4 +1,4 @@
-use candle_core::{Device, Tensor};
+// use candle_core::{Device, Tensor};
 use constensor_core::{Cpu, Graph, GraphTensor, R3};
 use criterion::{criterion_group, criterion_main, Criterion};
 
@@ -44,46 +44,46 @@ fn bench_cpu_graph_matmul_256(c: &mut Criterion) {
     });
 }
 
-fn bench_candle_matmul_64(c: &mut Criterion) {
-    const N: usize = 64;
-    c.bench_function("candle_matmul_64x64", |bencher| {
-        bencher.iter(|| {
-            let a = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
-            let b = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
-            let _ = a.matmul(&b).unwrap();
-        });
-    });
-}
+// fn bench_candle_matmul_64(c: &mut Criterion) {
+//     const N: usize = 64;
+//     let a = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
+//     let b = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
+//     c.bench_function("candle_matmul_64x64", |bencher| {
+//         bencher.iter(|| {
+//             let _ = a.matmul(&b).unwrap();
+//         });
+//     });
+// }
 
-fn bench_candle_matmul_128(c: &mut Criterion) {
-    const N: usize = 128;
-    c.bench_function("candle_matmul_128x128", |bencher| {
-        bencher.iter(|| {
-            let a = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
-            let b = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
-            let _ = a.matmul(&b).unwrap();
-        });
-    });
-}
+// fn bench_candle_matmul_128(c: &mut Criterion) {
+//     const N: usize = 128;
+//     let a = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
+//     let b = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
+//     c.bench_function("candle_matmul_128x128", |bencher| {
+//         bencher.iter(|| {
+//             let _ = a.matmul(&b).unwrap();
+//         });
+//     });
+// }
 
-fn bench_candle_matmul_256(c: &mut Criterion) {
-    const N: usize = 256;
-    c.bench_function("candle_matmul_256x256", |bencher| {
-        bencher.iter(|| {
-            let a = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
-            let b = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
-            let _ = a.matmul(&b).unwrap();
-        });
-    });
-}
+// fn bench_candle_matmul_256(c: &mut Criterion) {
+//     const N: usize = 256;
+//     let a = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
+//     let b = Tensor::rand(0f32, 1f32, &[1, N, N], &Device::Cpu).unwrap();
+//     c.bench_function("candle_matmul_256x256", |bencher| {
+//         bencher.iter(|| {
+//             let _ = a.matmul(&b).unwrap();
+//         });
+//     });
+// }
 
 criterion_group!(
     benches,
     bench_cpu_graph_matmul_64,
     bench_cpu_graph_matmul_128,
     bench_cpu_graph_matmul_256,
-    bench_candle_matmul_64,
-    bench_candle_matmul_128,
-    bench_candle_matmul_256
+    // bench_candle_matmul_64,
+    // bench_candle_matmul_128,
+    // bench_candle_matmul_256
 );
 criterion_main!(benches);
diff --git a/constensor-core/examples/test/main.rs b/constensor-core/examples/test/main.rs
new file mode 100644
index 0000000..a49910d
--- /dev/null
+++ b/constensor-core/examples/test/main.rs
@@ -0,0 +1,18 @@
+use constensor_core::{Graph, GraphTensor, Tensor, Wgpu, R2};
+
+fn main() {
+    let mut graph: Graph<f32> = Graph::empty();
+    let a = GraphTensor::<R2<3, 4>, f32, Wgpu>::fill(&mut graph, 1.0);
+    let b = GraphTensor::<R2<3, 4>, f32, Wgpu>::fill(&mut graph, 2.0);
+    let c = GraphTensor::<R2<3, 4>, f32, Wgpu>::fill(&mut graph, 3.0);
+    let _out = a + b + c;
+
+    graph.visualize("graph.png").unwrap();
+
+    let compiled: constensor_core::CompiledGraph<R2<3, 4>, f32, Wgpu> = graph.compile().unwrap();
+    let res = compiled.run().unwrap();
+
+    let tensor: Tensor<R2<3, 4>, f32, Wgpu> = res;
+
+    assert_eq!(tensor.data().unwrap().to_vec(), vec![vec![6.0; 4]; 3],);
+}
diff --git a/constensor-core/src/cpu_storage/mod.rs b/constensor-core/src/backends/cpu_backend/mod.rs
similarity index 88%
rename from constensor-core/src/cpu_storage/mod.rs
rename to constensor-core/src/backends/cpu_backend/mod.rs
index 31982ca..803d10a 100644
--- a/constensor-core/src/cpu_storage/mod.rs
+++ b/constensor-core/src/backends/cpu_backend/mod.rs
@@ -1,5 +1,4 @@
-use petgraph::algo::toposort;
-use petgraph::graphmap::DiGraphMap;
+use super::scheduler::topo_order;
 use std::{borrow::Cow, marker::PhantomData};
 
 use pool::{BufferPool, PooledBuffer};
@@ -31,7 +30,7 @@ impl<T: DType> BackendStorage<T> for CpuStorage<T> {
         Ok(Cow::Borrowed(self))
     }
     fn cast<U: DType>(&self) -> Result<Storage<U>> {
-        let new = self.0.iter().map(|x| U::from_f64(x.to_f64()));
+        let new = self.0.iter().map(|x| U::from_f64(x.cast_f64()));
         Ok(Storage::Cpu(CpuStorage(new.collect())))
     }
 }
@@ -43,47 +42,8 @@ impl BackendDevice for CpuDevice {
         &self,
         graph: Vec<GraphNode<T>>,
     ) -> Result<CompiledGraph<S, T, D>> {
-        // Build a dependency graph of tensor indices
-        let mut dep_graph = DiGraphMap::<usize, ()>::new();
-        for id in 0..graph.len() {
-            dep_graph.add_node(id);
-        }
-
-        for node in graph.iter() {
-            let idx = node.id.get();
-            match &node.op {
-                Op::BinaryOp { l_id, r_id, .. } => {
-                    dep_graph.add_edge(l_id.get(), idx, ());
-                    dep_graph.add_edge(r_id.get(), idx, ());
-                }
-                Op::UnaryOp { v_id, .. } => {
-                    dep_graph.add_edge(v_id.get(), idx, ());
-                }
-                Op::FusedMulAdd { a_id, b_id, c_id } => {
-                    dep_graph.add_edge(a_id.get(), idx, ());
-                    dep_graph.add_edge(b_id.get(), idx, ());
-                    dep_graph.add_edge(c_id.get(), idx, ());
-                }
-                Op::MatMul {
-                    l_id, r_id, o_id, ..
-                } => {
-                    dep_graph.add_edge(l_id.get(), idx, ());
-                    dep_graph.add_edge(r_id.get(), idx, ());
-                    if let Some(o_id) = o_id {
-                        dep_graph.add_edge(o_id.get(), idx, ());
-                    }
-                }
-                Op::Permute { v_id } => {
-                    dep_graph.add_edge(v_id.get(), idx, ());
-                }
-                // NoOp, Fill/Arange, Rand/Randn don’t create incoming edges
-                Op::NoOp | Op::Fill { .. } | Op::Arange { .. } | Op::Rand | Op::Randn { .. } => {}
-            }
-        }
-
-        // Compute topological order
-        let order = toposort(&dep_graph, None).expect("Cycle detected in graph!");
-
+        // Compute topological order using shared scheduler
+        let order = topo_order(&graph);
         Ok(CompiledGraph::Cpu {
             order,
             graph,
@@ -240,10 +200,10 @@ fn eval_node<T: DType + Send + Sync + 'static>(
         }
         Op::Arange { start, step, stop } => {
             let mut buf = pool.lock().unwrap().get_empty_buffer(out_elem_count);
-            let mut x = start.to_f64();
-            while x < stop.to_f64() {
+            let mut x = start.cast_f64();
+            while x < stop.cast_f64() {
                 buf.push(T::from_f64(x));
-                x += step.to_f64();
+                x += step.cast_f64();
             }
             PooledBuffer::new(buf, pool.clone())
         }
@@ -255,8 +215,8 @@ fn eval_node<T: DType + Send + Sync + 'static>(
             PooledBuffer::new(buf, pool.clone())
         }
         Op::Randn { mean, std } => {
-            let mean_f = mean.to_f64();
-            let std_f = std.to_f64();
+            let mean_f = mean.cast_f64();
+            let std_f = std.cast_f64();
             let normal = Normal::new(mean_f, std_f).unwrap();
             let mut buf = pool.lock().unwrap().get_buffer(out_elem_count);
             for elt in &mut buf {
diff --git a/constensor-core/src/cpu_storage/pool.rs b/constensor-core/src/backends/cpu_backend/pool.rs
similarity index 98%
rename from constensor-core/src/cpu_storage/pool.rs
rename to constensor-core/src/backends/cpu_backend/pool.rs
index 84154b2..35ce628 100644
--- a/constensor-core/src/cpu_storage/pool.rs
+++ b/constensor-core/src/backends/cpu_backend/pool.rs
@@ -128,7 +128,7 @@ impl<T: DType> BufferPool<T> {
                 self.pool
                     .iter()
                     .map(|b| b.capacity() * size_of::<T>())
-                    .sum()
+                    .sum::<usize>()
             );
 
             buf
@@ -161,7 +161,7 @@ impl<T: DType> BufferPool<T> {
                 self.pool
                     .iter()
                     .map(|b| b.capacity() * size_of::<T>())
-                    .sum()
+                    .sum::<usize>()
             );
 
             self.trim_excess();
@@ -195,7 +195,7 @@ impl<T: DType> BufferPool<T> {
             self.pool
                 .iter()
                 .map(|b| b.capacity() * size_of::<T>())
-                .sum()
+                .sum::<usize>()
         );
     }
 
diff --git a/constensor-core/src/cuda_backend/error.rs b/constensor-core/src/backends/cuda_backend/error.rs
similarity index 100%
rename from constensor-core/src/cuda_backend/error.rs
rename to constensor-core/src/backends/cuda_backend/error.rs
diff --git a/constensor-core/src/cuda_backend/mod.rs b/constensor-core/src/backends/cuda_backend/mod.rs
similarity index 93%
rename from constensor-core/src/cuda_backend/mod.rs
rename to constensor-core/src/backends/cuda_backend/mod.rs
index 29897f2..5ec5a0f 100644
--- a/constensor-core/src/cuda_backend/mod.rs
+++ b/constensor-core/src/backends/cuda_backend/mod.rs
@@ -1,3 +1,4 @@
+use super::scheduler::topo_order;
 use cudarc::{
     cublas::CudaBlas,
     driver::{
@@ -6,7 +7,6 @@ use cudarc::{
     nvrtc::{CompileOptions, Ptx},
 };
 use error::WrapErr;
-use petgraph::{algo::toposort, prelude::DiGraphMap};
 use std::sync::{
     atomic::{AtomicUsize, Ordering},
     Arc, Mutex, RwLock,
@@ -474,45 +474,8 @@ impl BackendDevice for CudaDevice {
         &self,
         graph: Vec<GraphNode<T>>,
     ) -> Result<CompiledGraph<S, T, D>> {
-        // Build a dependency graph of tensor indices
-        let mut dep_graph = DiGraphMap::<usize, ()>::new();
-        for idx in 0..graph.len() {
-            dep_graph.add_node(idx);
-        }
-
-        for (idx, node) in graph.iter().enumerate() {
-            match &node.op {
-                Op::BinaryOp { l_id, r_id, .. } => {
-                    dep_graph.add_edge(l_id.get(), idx, ());
-                    dep_graph.add_edge(r_id.get(), idx, ());
-                }
-                Op::UnaryOp { v_id, .. } => {
-                    dep_graph.add_edge(v_id.get(), idx, ());
-                }
-                Op::FusedMulAdd { a_id, b_id, c_id } => {
-                    dep_graph.add_edge(a_id.get(), idx, ());
-                    dep_graph.add_edge(b_id.get(), idx, ());
-                    dep_graph.add_edge(c_id.get(), idx, ());
-                }
-                Op::MatMul {
-                    l_id, r_id, o_id, ..
-                } => {
-                    dep_graph.add_edge(l_id.get(), idx, ());
-                    dep_graph.add_edge(r_id.get(), idx, ());
-                    if let Some(o_id) = o_id {
-                        dep_graph.add_edge(o_id.get(), idx, ());
-                    }
-                }
-                Op::Permute { v_id } => {
-                    dep_graph.add_edge(v_id.get(), idx, ());
-                }
-                // These don’t create incoming edges
-                Op::NoOp | Op::Fill { .. } | Op::Rand | Op::Randn { .. } | Op::Arange { .. } => {}
-            }
-        }
-
-        // Compute topological order
-        let order = toposort(&dep_graph, None).expect("Cycle detected in graph!");
+        // Compute topological order using shared scheduler
+        let order = topo_order(&graph);
 
         // New kernel and grouping logic with matmul input tracking
         let mut kernels = Vec::<CudaCompiledKernel<T>>::new();
diff --git a/constensor-core/src/cuda_backend/util.rs b/constensor-core/src/backends/cuda_backend/util.rs
similarity index 100%
rename from constensor-core/src/cuda_backend/util.rs
rename to constensor-core/src/backends/cuda_backend/util.rs
diff --git a/constensor-core/src/backends/mod.rs b/constensor-core/src/backends/mod.rs
new file mode 100644
index 0000000..784a95a
--- /dev/null
+++ b/constensor-core/src/backends/mod.rs
@@ -0,0 +1,5 @@
+pub mod cpu_backend;
+#[cfg(feature = "cuda")]
+pub mod cuda_backend;
+pub mod scheduler;
+pub mod wgpu_backend;
diff --git a/constensor-core/src/backends/scheduler.rs b/constensor-core/src/backends/scheduler.rs
new file mode 100644
index 0000000..03c5b89
--- /dev/null
+++ b/constensor-core/src/backends/scheduler.rs
@@ -0,0 +1,50 @@
+use petgraph::algo::toposort;
+use petgraph::graphmap::DiGraphMap;
+
+use crate::{DType, GraphNode, Op};
+
+/// Compute a topological ordering of the computation graph nodes.
+///
+/// # Panics
+/// Panics if the graph contains a cycle.
+pub fn topo_order<T: DType>(graph: &[GraphNode<T>]) -> Vec<usize> {
+    // Build dependency graph
+    let mut dep_graph = DiGraphMap::<usize, ()>::new();
+    for node in graph.iter() {
+        let idx = node.id.get();
+        dep_graph.add_node(idx);
+    }
+    for node in graph.iter() {
+        let dst = node.id.get();
+        match &node.op {
+            Op::BinaryOp { l_id, r_id, .. } => {
+                dep_graph.add_edge(l_id.get(), dst, ());
+                dep_graph.add_edge(r_id.get(), dst, ());
+            }
+            Op::UnaryOp { v_id, .. } => {
+                dep_graph.add_edge(v_id.get(), dst, ());
+            }
+            Op::FusedMulAdd { a_id, b_id, c_id } => {
+                dep_graph.add_edge(a_id.get(), dst, ());
+                dep_graph.add_edge(b_id.get(), dst, ());
+                dep_graph.add_edge(c_id.get(), dst, ());
+            }
+            Op::MatMul {
+                l_id, r_id, o_id, ..
+            } => {
+                dep_graph.add_edge(l_id.get(), dst, ());
+                dep_graph.add_edge(r_id.get(), dst, ());
+                if let Some(o) = o_id {
+                    dep_graph.add_edge(o.get(), dst, ());
+                }
+            }
+            Op::Permute { v_id } => {
+                dep_graph.add_edge(v_id.get(), dst, ());
+            }
+            // No incoming edges for other ops
+            _ => {}
+        }
+    }
+    // Compute topological order
+    toposort(&dep_graph, None).expect("Cycle detected in graph!")
+}
diff --git a/constensor-core/src/backends/wgpu_backend/kernels.rs b/constensor-core/src/backends/wgpu_backend/kernels.rs
new file mode 100644
index 0000000..1c6e1b3
--- /dev/null
+++ b/constensor-core/src/backends/wgpu_backend/kernels.rs
@@ -0,0 +1,260 @@
+//! Instantiate cubecl kernels for operations on all supported dtypes.
+
+use crate::{
+    dtype::DTypeOps,
+    graph::{BinaryOpType, UnaryOpType},
+};
+use cubecl::{channel::MutexComputeChannel, cube, prelude::*, wgpu::WgpuServer};
+
+use super::RT;
+
+pub trait UnaryKernelLaunch: CubeType + CubePrimitive + Send + Sync + Cast {
+    fn launch(
+        client: &ComputeClient<WgpuServer, MutexComputeChannel<WgpuServer>>,
+        count: CubeCount,
+        dim: CubeDim,
+        a: ArrayArg<'_, RT>,
+        out: ArrayArg<'_, RT>,
+        numel: u32,
+        ops: Sequence<UnaryOpType>,
+    );
+}
+
+// Integer impl → forwards to unary_int
+macro_rules! impl_launch_int {
+    ($($t:ty),* $(,)?) => {
+        $(
+            impl UnaryKernelLaunch for $t {
+                #[inline(always)]
+                fn launch(
+                    client: &ComputeClient<WgpuServer, MutexComputeChannel<WgpuServer>>,
+                    count: CubeCount,
+                    dim: CubeDim,
+                    a: ArrayArg<'_, RT>,
+                    out: ArrayArg<'_, RT>,
+                    numel: u32,
+                    ops: Sequence<UnaryOpType>,
+                ) {
+                    unsafe { unary_int::launch_unchecked::<Self, RT>(client, count, dim, a, out, numel, ops) };
+                }
+            }
+        )*
+    };
+}
+
+impl_launch_int!(i32, i64, u8, u32);
+
+impl UnaryKernelLaunch for f32 {
+    #[inline(always)]
+    fn launch(
+        client: &ComputeClient<WgpuServer, MutexComputeChannel<WgpuServer>>,
+        count: CubeCount,
+        dim: CubeDim,
+        a: ArrayArg<'_, RT>,
+        out: ArrayArg<'_, RT>,
+        numel: u32,
+        ops: Sequence<UnaryOpType>,
+    ) {
+        unsafe {
+            unary_float::launch_unchecked::<Self, RT>(client, count, dim, a, out, numel, ops)
+        };
+    }
+}
+
+impl UnaryKernelLaunch for f64 {
+    #[inline(always)]
+    fn launch(
+        client: &ComputeClient<WgpuServer, MutexComputeChannel<WgpuServer>>,
+        count: CubeCount,
+        dim: CubeDim,
+        a: ArrayArg<'_, RT>,
+        out: ArrayArg<'_, RT>,
+        numel: u32,
+        ops: Sequence<UnaryOpType>,
+    ) {
+        unsafe {
+            unary_float::launch_unchecked::<Self, RT>(client, count, dim, a, out, numel, ops)
+        };
+    }
+}
+
+#[cfg(feature = "half")]
+impl UnaryKernelLaunch for half::f16 {
+    #[inline(always)]
+    fn launch(
+        client: &ComputeClient<WgpuServer, MutexComputeChannel<WgpuServer>>,
+        count: CubeCount,
+        dim: CubeDim,
+        a: ArrayArg<'_, RT>,
+        out: ArrayArg<'_, RT>,
+        numel: u32,
+        ops: Sequence<UnaryOpType>,
+    ) {
+        unsafe {
+            unary_float::launch_unchecked::<Self, RT>(client, count, dim, a, out, numel, ops)
+        };
+    }
+}
+
+#[cfg(feature = "bfloat")]
+impl UnaryKernelLaunch for half::bf16 {
+    #[inline(always)]
+    fn launch(
+        client: &ComputeClient<WgpuServer, MutexComputeChannel<WgpuServer>>,
+        count: CubeCount,
+        dim: CubeDim,
+        a: ArrayArg<'_, RT>,
+        out: ArrayArg<'_, RT>,
+        numel: u32,
+        ops: Sequence<UnaryOpType>,
+    ) {
+        unsafe {
+            unary_float::launch_unchecked::<Self, RT>(client, count, dim, a, out, numel, ops)
+        };
+    }
+}
+
+/// Convenience wrapper that launches the *right* kernel for `T`.
+///
+/// `numel` and `ops` are compile‑time parameters exactly like the inner
+/// kernels, so you call this with the same `cube!` launch macro you
+/// already use.
+///
+/// ```ignore
+/// cube!(
+///     ctx,
+///     unary_auto::<u32>( &a, &mut out, comptime numel, comptime ops_seq )
+/// );
+/// ```
+pub(super) fn unary_auto<T: UnaryKernelLaunch>(
+    client: &ComputeClient<WgpuServer, MutexComputeChannel<WgpuServer>>,
+    count: CubeCount,
+    dim: CubeDim,
+    a: ArrayArg<'_, RT>,
+    out: ArrayArg<'_, RT>,
+    numel: u32,
+    ops: Sequence<UnaryOpType>,
+) {
+    T::launch(client, count, dim, a, out, numel, ops);
+}
+
+#[cube(launch_unchecked)]
+pub(super) fn binary<T: CubeType + CubePrimitive + Send + Sync + DTypeOps>(
+    a: &Array<Line<T>>,
+    b: &Sequence<Array<Line<T>>>,
+    out: &mut Array<Line<T>>,
+    #[comptime] numel: u32,
+    #[comptime] ops: Sequence<BinaryOpType>,
+) {
+    if ABSOLUTE_POS < numel {
+        let op = comptime! { ops.index(0) };
+        let bv = b.index(0);
+        match op {
+            BinaryOpType::Add => out[ABSOLUTE_POS] = a[ABSOLUTE_POS] + bv[ABSOLUTE_POS],
+            BinaryOpType::Sub => out[ABSOLUTE_POS] = a[ABSOLUTE_POS] - bv[ABSOLUTE_POS],
+            BinaryOpType::Mul => out[ABSOLUTE_POS] = a[ABSOLUTE_POS] * bv[ABSOLUTE_POS],
+            BinaryOpType::Div => out[ABSOLUTE_POS] = a[ABSOLUTE_POS] / bv[ABSOLUTE_POS],
+        }
+
+        #[unroll]
+        for index in 1..ops.len() {
+            let op = comptime! { ops.index(index.clone()) };
+            let bv = b.index(index);
+            match op {
+                BinaryOpType::Add => out[ABSOLUTE_POS] = out[ABSOLUTE_POS] + bv[ABSOLUTE_POS],
+                BinaryOpType::Sub => out[ABSOLUTE_POS] = out[ABSOLUTE_POS] - bv[ABSOLUTE_POS],
+                BinaryOpType::Mul => out[ABSOLUTE_POS] = out[ABSOLUTE_POS] * bv[ABSOLUTE_POS],
+                BinaryOpType::Div => out[ABSOLUTE_POS] = out[ABSOLUTE_POS] / bv[ABSOLUTE_POS],
+            }
+        }
+    }
+}
+
+#[cube(launch_unchecked)]
+pub(super) fn unary_float<F: Float>(
+    a: &Array<F>,
+    out: &mut Array<F>,
+    #[comptime] numel: u32,
+    #[comptime] ops: Sequence<UnaryOpType>,
+) {
+    if ABSOLUTE_POS < numel {
+        let op = comptime! { ops.index(0) };
+        match op {
+            UnaryOpType::Neg => out[ABSOLUTE_POS] = F::from_int(0) - a[ABSOLUTE_POS],
+            UnaryOpType::Sqrt => out[ABSOLUTE_POS] = F::sqrt(a[ABSOLUTE_POS]),
+            UnaryOpType::Exp => out[ABSOLUTE_POS] = F::exp(a[ABSOLUTE_POS]),
+            UnaryOpType::Exp2 => out[ABSOLUTE_POS] = F::powf(F::from_int(2), a[ABSOLUTE_POS]),
+            UnaryOpType::Log => out[ABSOLUTE_POS] = F::log(a[ABSOLUTE_POS]),
+            UnaryOpType::Log1p => out[ABSOLUTE_POS] = F::log1p(a[ABSOLUTE_POS]),
+        }
+
+        #[unroll]
+        for index in 1..ops.len() {
+            let op = comptime! { ops.index(index.clone()) };
+            match op {
+                UnaryOpType::Neg => out[ABSOLUTE_POS] = F::from_int(0) - out[ABSOLUTE_POS],
+                UnaryOpType::Sqrt => out[ABSOLUTE_POS] = F::sqrt(out[ABSOLUTE_POS]),
+                UnaryOpType::Exp => out[ABSOLUTE_POS] = F::exp(out[ABSOLUTE_POS]),
+                UnaryOpType::Exp2 => out[ABSOLUTE_POS] = F::powf(F::from_int(2), out[ABSOLUTE_POS]),
+                UnaryOpType::Log => out[ABSOLUTE_POS] = F::log(out[ABSOLUTE_POS]),
+                UnaryOpType::Log1p => out[ABSOLUTE_POS] = F::log1p(out[ABSOLUTE_POS]),
+            }
+        }
+    }
+}
+
+#[cube(launch_unchecked)]
+pub(super) fn unary_int<I: CubeType + CubePrimitive + Send + Sync + DTypeOps + Cast>(
+    a: &Array<I>,
+    out: &mut Array<I>,
+    #[comptime] numel: u32,
+    #[comptime] ops: Sequence<UnaryOpType>,
+) {
+    if ABSOLUTE_POS < numel {
+        // ---- first op -----------------------------------------------------
+        let op = comptime! { ops.index(0) };
+        let mut tmp: f32 = f32::cast_from(a[ABSOLUTE_POS]);
+
+        match op {
+            UnaryOpType::Neg => tmp = 0. - tmp,
+            UnaryOpType::Sqrt => tmp = f32::sqrt(tmp),
+            UnaryOpType::Exp => tmp = f32::exp(tmp),
+            UnaryOpType::Exp2 => tmp = Powf::powf(2.0_f32, tmp),
+            UnaryOpType::Log => tmp = Log::log(tmp),
+            UnaryOpType::Log1p => tmp = Log1p::log1p(tmp),
+        }
+
+        out[ABSOLUTE_POS] = I::cast_from(tmp);
+
+        // ---- remaining ops (if any) --------------------------------------
+        #[unroll]
+        for index in 1..ops.len() {
+            let op = comptime! { ops.index(index.clone()) };
+            let mut tmp: f32 = f32::cast_from(out[ABSOLUTE_POS]);
+
+            match op {
+                UnaryOpType::Neg => tmp = 0. - tmp,
+                UnaryOpType::Sqrt => tmp = f32::sqrt(tmp),
+                UnaryOpType::Exp => tmp = f32::exp(tmp),
+                UnaryOpType::Exp2 => tmp = Powf::powf(2.0_f32, tmp),
+                UnaryOpType::Log => tmp = Log::log(tmp),
+                UnaryOpType::Log1p => tmp = Log1p::log1p(tmp),
+            }
+
+            out[ABSOLUTE_POS] = I::cast_from(tmp);
+        }
+    }
+}
+
+#[cube(launch_unchecked)]
+pub(super) fn fill<
+    T: CubeType + CubePrimitive + Send + Sync + LaunchArgExpand + Numeric + DTypeOps,
+>(
+    out: &mut Array<T>,
+    value: T,
+    #[comptime] numel: u32,
+) {
+    if ABSOLUTE_POS < numel {
+        out[ABSOLUTE_POS] = value;
+    }
+}
diff --git a/constensor-core/src/backends/wgpu_backend/mod.rs b/constensor-core/src/backends/wgpu_backend/mod.rs
new file mode 100644
index 0000000..d5020fd
--- /dev/null
+++ b/constensor-core/src/backends/wgpu_backend/mod.rs
@@ -0,0 +1,212 @@
+use std::{borrow::Cow, collections::HashMap, marker::PhantomData};
+
+use super::scheduler::topo_order;
+use crate::Op;
+use cubecl::{
+    channel::MutexComputeChannel,
+    prelude::*,
+    server::Handle,
+    wgpu::{WgpuRuntime, WgpuServer},
+};
+
+use crate::{
+    device::Dev,
+    storage::{BackendDevice, BackendStorage, Storage},
+    CompiledGraph, DType, GraphNode, Result, Shape,
+};
+
+pub(crate) mod kernels;
+
+use super::cpu_backend::CpuStorage;
+
+type RT = WgpuRuntime;
+
+#[cfg(any(feature = "cuda", feature = "hip"))]
+const DEVICE: cubecl::wgpu::WgpuDevice = cubecl::wgpu::WgpuDevice::DiscreteGpu(0);
+#[cfg(feature = "metal")]
+const DEVICE: cubecl::wgpu::WgpuDevice = cubecl::wgpu::WgpuDevice::IntegratedGpu(0);
+#[cfg(not(any(feature = "cuda", feature = "hip", feature = "metal")))]
+const DEVICE: cubecl::wgpu::WgpuDevice = cubecl::wgpu::WgpuDevice::DefaultDevice;
+
+fn client() -> ComputeClient<WgpuServer, MutexComputeChannel<WgpuServer>> {
+    RT::client(&DEVICE)
+}
+
+const VECTORIZATION: u32 = 4;
+
+pub struct WgpuDevice;
+
+pub struct WgpuStorage<X: DType> {
+    handle: Handle,
+    ghost: PhantomData<X>,
+}
+
+impl<X: DType> BackendStorage<X> for WgpuStorage<X> {
+    fn to_cpu_storage(&self) -> Result<Cow<'_, CpuStorage<X>>> {
+        let client = client();
+
+        let bytes = client.read_one(self.handle.clone().binding());
+        let output = X::from_bytes(&bytes);
+
+        Ok(Cow::Owned(CpuStorage(output.to_vec())))
+    }
+
+    fn cast<U: DType>(&self) -> Result<Storage<U>> {
+        todo!()
+    }
+}
+
+impl BackendDevice for WgpuDevice {
+    type Storage<X: DType> = WgpuStorage<X>;
+
+    fn compile<S: Shape, T: DType, D: Dev>(
+        &self,
+        graph: Vec<GraphNode<T>>,
+    ) -> Result<CompiledGraph<S, T, D>> {
+        // Compute topological order using shared scheduler
+        let order = topo_order(&graph);
+        Ok(CompiledGraph::Wgpu {
+            order,
+            graph,
+            ghost: PhantomData,
+        })
+    }
+
+    fn run_graph<S: Shape, T: DType, D: Dev>(
+        &self,
+        comp: &CompiledGraph<S, T, D>,
+    ) -> Result<Self::Storage<T>> {
+        // Sequential execution of nodes in topological order.
+        #[allow(irrefutable_let_patterns)]
+        let CompiledGraph::Wgpu {
+            order,
+            graph,
+            ghost: _,
+        } = comp
+        else {
+            unreachable!("Expected Wgpu compiled graph");
+        };
+
+        let client = client();
+
+        let mut handles = HashMap::new();
+        for &idx in order.iter() {
+            let node = &graph[idx];
+            let out_elem_count: usize = node.shape.iter().product();
+
+            match &node.op {
+                Op::Fill { v } => {
+                    let output_handle = client.empty(out_elem_count * core::mem::size_of::<T>());
+
+                    let out: ArrayArg<'_, RT> = unsafe {
+                        ArrayArg::from_raw_parts::<T>(
+                            &output_handle,
+                            out_elem_count,
+                            VECTORIZATION as u8,
+                        )
+                    };
+                    unsafe {
+                        kernels::fill::launch_unchecked::<T, WgpuRuntime>(
+                            &client,
+                            CubeCount::Static(VECTORIZATION, 1, 1),
+                            CubeDim::new((out_elem_count as u32).div_ceil(VECTORIZATION), 1, 1),
+                            out,
+                            ScalarArg::new(*v),
+                            out_elem_count as u32,
+                        );
+                    };
+
+                    handles.insert(idx, output_handle.clone());
+                }
+
+                Op::BinaryOp {
+                    l_id,
+                    r_id,
+                    operator,
+                } => {
+                    let a_handle = &handles[&l_id.get()];
+                    let b_handle = &handles[&r_id.get()];
+                    let output_handle = client.empty(out_elem_count * core::mem::size_of::<T>());
+
+                    unsafe {
+                        let a: ArrayArg<'_, RT> = ArrayArg::from_raw_parts::<T>(
+                            &a_handle,
+                            out_elem_count,
+                            VECTORIZATION as u8,
+                        );
+
+                        let mut b_seq: SequenceArg<'_, RT, Array<Line<T>>> = SequenceArg::new();
+                        b_seq.push(ArrayArg::from_raw_parts::<T>(
+                            &b_handle,
+                            out_elem_count,
+                            VECTORIZATION as u8,
+                        ));
+
+                        let out: ArrayArg<'_, RT> = ArrayArg::from_raw_parts::<T>(
+                            &output_handle,
+                            out_elem_count,
+                            VECTORIZATION as u8,
+                        );
+
+                        let mut ops = Sequence::new();
+                        ops.push(*operator);
+                        kernels::binary::launch_unchecked(
+                            &client,
+                            CubeCount::Static(VECTORIZATION, 1, 1),
+                            CubeDim::new((out_elem_count as u32).div_ceil(VECTORIZATION), 1, 1),
+                            a,
+                            b_seq,
+                            out,
+                            out_elem_count as u32,
+                            ops,
+                        );
+                    };
+
+                    handles.insert(idx, output_handle.clone());
+                }
+
+                Op::UnaryOp { v_id, operator } => {
+                    let v_handle = &handles[&v_id.get()];
+                    let output_handle = client.empty(out_elem_count * core::mem::size_of::<T>());
+
+                    unsafe {
+                        let a: ArrayArg<'_, RT> = ArrayArg::from_raw_parts::<T>(
+                            &v_handle,
+                            out_elem_count,
+                            VECTORIZATION as u8,
+                        );
+
+                        let out: ArrayArg<'_, RT> = ArrayArg::from_raw_parts::<T>(
+                            &output_handle,
+                            out_elem_count,
+                            VECTORIZATION as u8,
+                        );
+
+                        let mut ops = Sequence::new();
+                        ops.push(*operator);
+                        kernels::unary_auto::<T>(
+                            &client,
+                            CubeCount::Static(VECTORIZATION, 1, 1),
+                            CubeDim::new((out_elem_count as u32).div_ceil(VECTORIZATION), 1, 1),
+                            a,
+                            out,
+                            out_elem_count as u32,
+                            ops,
+                        );
+                    };
+
+                    handles.insert(idx, output_handle.clone());
+                }
+
+                _ => todo!(),
+            }
+        }
+
+        let key = *handles.keys().max().unwrap();
+        let final_handle = handles.remove(&key).expect("No output");
+        Ok(WgpuStorage {
+            handle: final_handle,
+            ghost: PhantomData,
+        })
+    }
+}
diff --git a/constensor-core/src/device.rs b/constensor-core/src/device.rs
index fe03c45..d18011c 100644
--- a/constensor-core/src/device.rs
+++ b/constensor-core/src/device.rs
@@ -1,7 +1,8 @@
 #[cfg(feature = "cuda")]
 use crate::cuda_backend::CudaDevice;
 use crate::{
-    cpu_storage::CpuDevice,
+    backends::wgpu_backend::WgpuDevice,
+    cpu_backend::CpuDevice,
     storage::{BackendDevice, Storage},
     CompiledGraph, DType, GraphNode, Result, Shape,
 };
@@ -19,6 +20,14 @@ impl Dev for Cpu {
         Ok(Device::Cpu)
     }
 }
+#[derive(Clone)]
+pub struct Wgpu;
+
+impl Dev for Wgpu {
+    fn resolve() -> Result<Device> {
+        Ok(Device::Wgpu)
+    }
+}
 
 #[cfg(feature = "cuda")]
 #[derive(Clone)]
@@ -68,6 +77,7 @@ pub enum Device {
     #[cfg(feature = "cuda")]
     Cuda(CudaDevice),
     Cpu,
+    Wgpu,
 }
 
 impl Device {
@@ -79,6 +89,7 @@ impl Device {
             #[cfg(feature = "cuda")]
             Self::Cuda(cuda) => Ok(Storage::Cuda(cuda.run_graph::<S, T, D>(graph)?)),
             Self::Cpu => Ok(Storage::Cpu(CpuDevice.run_graph::<S, T, D>(graph)?)),
+            Self::Wgpu => Ok(Storage::Wgpu(WgpuDevice.run_graph::<S, T, D>(graph)?)),
         }
     }
 
@@ -90,6 +101,7 @@ impl Device {
             #[cfg(feature = "cuda")]
             Self::Cuda(cuda) => cuda.compile::<S, T, D>(graph),
             Self::Cpu => CpuDevice.compile::<S, T, D>(graph),
+            Self::Wgpu => WgpuDevice.compile::<S, T, D>(graph),
         }
     }
 }
diff --git a/constensor-core/src/dtype/mod.rs b/constensor-core/src/dtype/mod.rs
index 3529fc2..36b3f88 100644
--- a/constensor-core/src/dtype/mod.rs
+++ b/constensor-core/src/dtype/mod.rs
@@ -3,6 +3,10 @@ use std::{
     ops::{Add, Div, Mul, Sub},
 };
 
+use cubecl::{
+    prelude::{CubePrimitive, CubeType, Numeric},
+    CubeElement,
+};
 #[cfg(feature = "bfloat")]
 use half::bf16;
 #[cfg(feature = "half")]
@@ -273,34 +277,6 @@ impl Loggable for f16 {
     }
 }
 
-pub trait DTypeOps:
-    Copy
-    + Add<Output = Self>
-    + Div<Output = Self>
-    + Sub<Output = Self>
-    + Mul<Output = Self>
-    + Sqrtable
-    + Expable
-    + Loggable
-    + SimdSupported
-    + GemmDispatch
-    + RandDispatch
-{
-}
-
-#[cfg(feature = "cuda")]
-pub trait DeviceReprLike: DeviceRepr {}
-
-#[cfg(not(feature = "cuda"))]
-pub trait DeviceReprLike {}
-
-impl DeviceReprLike for u8 {}
-impl DeviceReprLike for i32 {}
-impl DeviceReprLike for u32 {}
-impl DeviceReprLike for i64 {}
-impl DeviceReprLike for f32 {}
-impl DeviceReprLike for f64 {}
-
 pub trait MaybeNeg {
     const NAME: &'static str;
 
@@ -339,9 +315,49 @@ maybe_neg!(i64);
 maybe_neg!(f32);
 maybe_neg!(f64);
 
+pub trait DTypeOps:
+    Copy
+    + Add<Output = Self>
+    + Div<Output = Self>
+    + Sub<Output = Self>
+    + Mul<Output = Self>
+    + MaybeNeg
+    + Sqrtable
+    + Expable
+    + Loggable
+    + SimdSupported
+    + GemmDispatch
+    + RandDispatch
+    + CubeType
+    + crate::wgpu_backend::kernels::UnaryKernelLaunch
+{
+}
+
+#[cfg(feature = "cuda")]
+pub trait DeviceReprLike: DeviceRepr {}
+
+#[cfg(not(feature = "cuda"))]
+pub trait DeviceReprLike {}
+
+impl DeviceReprLike for u8 {}
+impl DeviceReprLike for i32 {}
+impl DeviceReprLike for u32 {}
+impl DeviceReprLike for i64 {}
+impl DeviceReprLike for f32 {}
+impl DeviceReprLike for f64 {}
+
 /// Marker trait for tensor datatypes.
 pub trait DType:
-    Debug + Clone + DTypeOps + Send + Sync + MaybeNeg + DeviceReprLike + 'static
+    Debug
+    + Clone
+    + DTypeOps
+    + Send
+    + Sync
+    + DeviceReprLike
+    + CubePrimitive
+    + CubeElement
+    + Numeric
+    + 'static
 {
     const ZERO: Self;
     const ONE: Self;
@@ -349,7 +365,7 @@ pub trait DType:
     const C_DEP: Option<&'static str>;
     const INTEGRAL: bool;
 
-    fn to_f64(&self) -> f64;
+    fn cast_f64(&self) -> f64;
     fn from_f64(x: f64) -> Self;
 }
 
@@ -363,7 +379,7 @@ macro_rules! dtype {
             const C_DEP: Option<&'static str> = None;
             const INTEGRAL: bool = $integral;
 
-            fn to_f64(&self) -> f64 {
+            fn cast_f64(&self) -> f64 {
                 *self as f64
             }
             fn from_f64(x: f64) -> Self {
@@ -394,7 +410,7 @@ impl DType for f16 {
     const C_DEP: Option<&'static str> = Some("#include \"cuda_fp16.h\"");
     const INTEGRAL: bool = false;
 
-    fn to_f64(&self) -> f64 {
+    fn cast_f64(&self) -> f64 {
         self.to_f64_const()
     }
     fn from_f64(x: f64) -> Self {
@@ -415,7 +431,7 @@ impl DType for bf16 {
     const C_DEP: Option<&'static str> = Some("#include \"cuda_bf16.h\"");
     const INTEGRAL: bool = false;
 
-    fn to_f64(&self) -> f64 {
+    fn cast_f64(&self) -> f64 {
         self.to_f64_const()
     }
     fn from_f64(x: f64) -> Self {
diff --git a/constensor-core/src/graph.rs b/constensor-core/src/graph.rs
index efb2773..fc2a9f8 100644
--- a/constensor-core/src/graph.rs
+++ b/constensor-core/src/graph.rs
@@ -13,6 +13,7 @@ use std::{
 
 use crate::{device::Dev, tensor::concretetensor::from_storage, DType, Result, Shape, Tensor};
 
+use cubecl::CubeType;
 use petgraph::Graph as PetGraph;
 use petgraph::{dot::Dot, graph::NodeIndex};
 
@@ -663,6 +664,12 @@ pub enum CompiledGraph<S: Shape, T: DType, D: Dev> {
         graph: Vec<GraphNode<T>>,
         ghost: PhantomData<(S, T, D)>,
     },
+    /// Compiled graph for WGPU backend.
+    Wgpu {
+        order: Vec<usize>,
+        graph: Vec<GraphNode<T>>,
+        ghost: PhantomData<(S, T, D)>,
+    },
     #[cfg(feature = "cuda")]
     Cuda {
         kernels: Vec<crate::cuda_backend::CudaCompiledKernel<T>>,
@@ -679,7 +686,7 @@ impl<S: Shape, T: DType, D: Dev> CompiledGraph<S, T, D> {
     }
 }
 
-#[derive(PartialEq, Debug, Clone, Copy)]
+#[derive(CubeType, Clone, Copy, Debug, Hash, PartialEq, Eq)]
 pub enum BinaryOpType {
     Add,
     Div,
@@ -707,7 +714,7 @@ impl BinaryOpType {
     }
 }
 
-#[derive(PartialEq, Debug, Clone)]
+#[derive(CubeType, Clone, Copy, Debug, Hash, PartialEq, Eq)]
 pub enum UnaryOpType {
     Neg,
     Sqrt,
diff --git a/constensor-core/src/lib.rs b/constensor-core/src/lib.rs
index 0733715..cff8235 100644
--- a/constensor-core/src/lib.rs
+++ b/constensor-core/src/lib.rs
@@ -43,9 +43,8 @@
 //! assert_eq!(tensor.data().unwrap().to_vec(), vec![vec![9.0; 4]; 3],);
 //! ```
 
-mod cpu_storage;
-#[cfg(feature = "cuda")]
-mod cuda_backend;
+mod backends;
+use backends::*;
 mod device;
 mod dtype;
 mod error;
@@ -56,7 +55,7 @@ mod tensor;
 
 #[cfg(feature = "cuda")]
 pub use device::Cuda;
-pub use device::{BestDevice, Cpu};
+pub use device::{BestDevice, Cpu, Wgpu};
 pub use dtype::DType;
 pub use error::{Context, Error, Result};
 pub use graph::{CompiledGraph, Graph, GraphNode, Op};
diff --git a/constensor-core/src/storage.rs b/constensor-core/src/storage.rs
index a3e0851..d772ae5 100644
--- a/constensor-core/src/storage.rs
+++ b/constensor-core/src/storage.rs
@@ -2,18 +2,23 @@ use std::borrow::Cow;
 
 #[cfg(feature = "cuda")]
 use crate::cuda_backend::CudaStorage;
-use crate::{cpu_storage::CpuStorage, device::Dev, CompiledGraph, DType, GraphNode, Result, Shape};
+use crate::{
+    backends::wgpu_backend::WgpuStorage, cpu_backend::CpuStorage, device::Dev, CompiledGraph,
+    DType, GraphNode, Result, Shape,
+};
 
 pub enum Storage<T: DType> {
     #[cfg(feature = "cuda")]
     Cuda(CudaStorage<T>),
     Cpu(CpuStorage<T>),
+    Wgpu(WgpuStorage<T>),
 }
 
 impl<T: DType> Storage<T> {
     pub(crate) fn to_cpu_storage(&self) -> Result<Cow<'_, CpuStorage<T>>> {
         match self {
             Self::Cpu(cpu) => cpu.to_cpu_storage(),
+            Self::Wgpu(wgpu) => wgpu.to_cpu_storage(),
             #[cfg(feature = "cuda")]
             Self::Cuda(cuda) => cuda.to_cpu_storage(),
         }
@@ -22,6 +27,7 @@ impl<T: DType> Storage<T> {
     pub(crate) fn cast<U: DType>(&self) -> Result<Storage<U>> {
         match self {
             Self::Cpu(cpu) => cpu.cast::<U>(),
+            Self::Wgpu(wgpu) => wgpu.cast::<U>(),
             #[cfg(feature = "cuda")]
             Self::Cuda(cuda) => cuda.cast::<U>(),
         }
diff --git a/constensor-core/src/tensor/concretetensor.rs b/constensor-core/src/tensor/concretetensor.rs
index c27f557..63842dd 100644
--- a/constensor-core/src/tensor/concretetensor.rs
+++ b/constensor-core/src/tensor/concretetensor.rs
@@ -1,5 +1,5 @@
 use crate::{
-    device::{Cpu, Dev},
+    device::{Cpu, Dev, Wgpu},
     storage::Storage,
     DType, Result, Shape, R1, R2, R3,
 };
@@ -108,6 +108,7 @@ macro_rules! tensor_api {
 }
 
 tensor_api!(Cpu);
+tensor_api!(Wgpu);
 #[cfg(feature = "cuda")]
 tensor_api!(Cuda<0>);
 
diff --git a/constensor-core/src/tensor/graphtensor.rs b/constensor-core/src/tensor/graphtensor.rs
index cf6a9d9..3faeed9 100644
--- a/constensor-core/src/tensor/graphtensor.rs
+++ b/constensor-core/src/tensor/graphtensor.rs
@@ -257,7 +257,7 @@ impl<const A: usize, T: DType, D: Dev> GraphTensor<R1<A>, T, D> {
     /// A GraphTensor representing a vector ranging from `start` to `stop` with `step` computed using A.
     pub fn arange(graph: &mut Graph<T>, start: T, stop: T) -> Self {
         let id = graph.next_id();
-        let step = (stop.to_f64() - start.to_f64()) / (A as f64);
+        let step = (stop.cast_f64() - start.cast_f64()) / (A as f64);
         let strides = contiguous_strides(&[A]);
         graph.add_op::<R1<A>>(
             Op::Arange {
diff --git a/graph.png b/graph.png
index 584fddd..90ad13d 100644
Binary files a/graph.png and b/graph.png differ