ci: update version string in docs

2026-01-13 08:17:57 -05:00 · 2025-01-17 14:19:32 +00:00
19 changed files with 179 additions and 471 deletions
--- a/.github/workflows/benchmarks.yml
+++ b/.github/workflows/benchmarks.yml
@@ -8,8 +8,6 @@ on:
 jobs:

  bench_poseidon:
-    permissions:
-      contents: read
    runs-on: self-hosted
    steps:
      - uses: actions/checkout@v4
@@ -24,8 +22,6 @@ jobs:
        run: cargo bench --verbose --bench poseidon

  bench_einsum_accum_matmul:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
@@ -41,8 +37,6 @@ jobs:
        run: cargo bench --verbose --bench accum_einsum_matmul

  bench_accum_matmul_relu:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
@@ -58,8 +52,6 @@ jobs:
        run: cargo bench --verbose --bench accum_matmul_relu

  bench_accum_matmul_relu_overflow:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
@@ -75,8 +67,6 @@ jobs:
        run: cargo bench --verbose --bench accum_matmul_relu_overflow

  bench_relu:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
@@ -92,8 +82,6 @@ jobs:
        run: cargo bench --verbose --bench relu

  bench_accum_dot:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
@@ -109,8 +97,6 @@ jobs:
        run: cargo bench --verbose --bench accum_dot

  bench_accum_conv:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
@@ -126,8 +112,6 @@ jobs:
        run: cargo bench --verbose --bench accum_conv

  bench_accum_sumpool:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
@@ -143,8 +127,6 @@ jobs:
        run: cargo bench --verbose --bench accum_sumpool

  bench_pairwise_add:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
@@ -160,8 +142,6 @@ jobs:
        run: cargo bench --verbose --bench pairwise_add

  bench_accum_sum:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
@@ -177,8 +157,6 @@ jobs:
        run: cargo bench --verbose --bench accum_sum

  bench_pairwise_pow:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [bench_poseidon]
    steps:
--- a/.github/workflows/engine.yml
+++ b/.github/workflows/engine.yml
@@ -15,9 +15,6 @@ defaults:
    working-directory: .
 jobs:
  publish-wasm-bindings:
-    permissions:
-      contents: read
-      packages: write
    name: publish-wasm-bindings
    runs-on: ubuntu-latest
    if: startsWith(github.ref, 'refs/tags/')
@@ -189,9 +186,6 @@ jobs:


  in-browser-evm-ver-publish:
-    permissions:
-      contents: read
-      packages: write
    name: publish-in-browser-evm-verifier-package
    needs: [publish-wasm-bindings]
    runs-on: ubuntu-latest
--- a/.github/workflows/large-tests.yml
+++ b/.github/workflows/large-tests.yml
@@ -6,8 +6,6 @@ on:
        description: "Test scenario tags"
 jobs:
  large-tests:
-    permissions:
-      contents: read
    runs-on: kaiju
    steps:
      - uses: actions/checkout@v4
--- a/.github/workflows/pypi-gpu.yml
+++ b/.github/workflows/pypi-gpu.yml
@@ -18,9 +18,6 @@ defaults:
 jobs:

  linux:
-    permissions:
-      contents: read
-      packages: write
    runs-on: GPU
    strategy:
      matrix:
--- a/.github/workflows/pypi.yml
+++ b/.github/workflows/pypi.yml
@@ -16,8 +16,6 @@ defaults:

 jobs:
  macos:
-    permissions:
-      contents: read
    runs-on: macos-latest
    if: startsWith(github.ref, 'refs/tags/')
    strategy:
@@ -49,13 +47,6 @@ jobs:
          components: rustfmt, clippy

      - name: Build wheels
-        if: matrix.target == 'universal2-apple-darwin'
-        uses: PyO3/maturin-action@v1
-        with:
-          target: ${{ matrix.target }}
-          args: --release --out dist --features python-bindings
-      - name: Build wheels
-        if: matrix.target == 'x86_64'
        uses: PyO3/maturin-action@v1
        with:
          target: ${{ matrix.target }}
@@ -73,8 +64,6 @@ jobs:
          path: dist

  windows:
-    permissions:
-      contents: read
    runs-on: windows-latest
    if: startsWith(github.ref, 'refs/tags/')
    strategy:
@@ -122,8 +111,6 @@ jobs:
          path: dist

  linux:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest
    if: startsWith(github.ref, 'refs/tags/')
    strategy:
@@ -239,8 +226,6 @@ jobs:
  #       path: dist

  musllinux:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest
    if: startsWith(github.ref, 'refs/tags/')
    strategy:
@@ -306,8 +291,6 @@ jobs:
          path: dist

  musllinux-cross:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest
    if: startsWith(github.ref, 'refs/tags/')
    strategy:
@@ -395,8 +378,6 @@ jobs:
          packages-dir: ./

  doc-publish:
-    permissions:
-      contents: read
    name: Trigger ReadTheDocs Build
    runs-on: ubuntu-latest
    needs: pypi-publish
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -10,9 +10,6 @@ on:
      - "*"
 jobs:
  create-release:
-    permissions:
-      contents: read
-      packages: write
    name: create-release
    runs-on: ubuntu-22.04
    if: startsWith(github.ref, 'refs/tags/')
@@ -36,9 +33,6 @@ jobs:
          tag_name: ${{ env.EZKL_VERSION }}

  build-release-gpu:
-    permissions:
-      contents: read
-      packages: write
    name: build-release-gpu
    needs: ["create-release"]
    runs-on: GPU
@@ -100,10 +94,6 @@ jobs:
          asset_content_type: application/octet-stream

  build-release:
-    permissions:
-      contents: read
-      packages: write
-      issues: write
    name: build-release
    needs: ["create-release"]
    runs-on: ${{ matrix.os }}
@@ -196,18 +186,14 @@ jobs:
          echo "target flag is: ${{ env.TARGET_FLAGS }}"
          echo "target dir is: ${{ env.TARGET_DIR }}"

-      - name: Build release binary (no asm or metal) 
-        if: matrix.build != 'linux-gnu' && matrix.build != 'macos-aarch64'
+      - name: Build release binary (no asm)
+        if: matrix.build != 'linux-gnu'
        run: ${{ env.CARGO }} build --release ${{ env.TARGET_FLAGS }} -Z sparse-registry

      - name: Build release binary (asm)
        if: matrix.build == 'linux-gnu'
        run: ${{ env.CARGO }} build --release ${{ env.TARGET_FLAGS }} -Z sparse-registry --features asm

-      - name: Build release binary (metal)
-        if: matrix.build == 'macos-aarch64'
-        run: ${{ env.CARGO }} build --release ${{ env.TARGET_FLAGS }} -Z sparse-registry --features macos-metal
-
      - name: Strip release binary
        if: matrix.build != 'windows-msvc' && matrix.build != 'linux-aarch64'
        run: strip "target/${{ matrix.target }}/release/ezkl"
--- a/.github/workflows/rust.yml
+++ b/.github/workflows/rust.yml
@@ -21,8 +21,6 @@ env:
 jobs:

  fr-age-test:
-    permissions:
-      contents: read
    runs-on: large-self-hosted
    steps:
      - uses: actions/checkout@v4
@@ -37,9 +35,8 @@ jobs:
        run: cargo test --release --verbose tests::large_mock_::large_tests_6_expects -- --include-ignored

  build:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest
+
    steps:
      - uses: actions/checkout@v4
        with:
@@ -53,8 +50,6 @@ jobs:
        run: cargo build --verbose

  docs:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
@@ -69,8 +64,6 @@ jobs:
        run: cargo doc --verbose

  library-tests:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest-32-cores
    steps:
      - uses: actions/checkout@v4
@@ -131,8 +124,6 @@ jobs:
  #       run: cargo nextest run conv_relu_col_ultra_overflow --no-capture --features icicle -- --include-ignored

  ultra-overflow-tests_og-lookup:
-    permissions:
-      contents: read
    runs-on: non-gpu
    steps:
      - uses: actions/checkout@v4
@@ -168,8 +159,6 @@ jobs:
        run: cargo nextest run  --release conv_relu_col_ultra_overflow --no-capture --no-default-features --features ezkl -- --include-ignored

  ultra-overflow-tests:
-    permissions:
-      contents: read
    runs-on: non-gpu
    steps:
      - uses: actions/checkout@v4
@@ -205,8 +194,6 @@ jobs:
        run: cargo nextest run  --release conv_relu_col_ultra_overflow --no-capture -- --include-ignored

  model-serialization:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest-16-cores
    steps:
      - uses: actions/checkout@v4
@@ -225,8 +212,6 @@ jobs:
        run: cargo nextest run native_tests::tests::model_serialization_different_binaries_  --test-threads 1

  wasm32-tests:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
@@ -254,8 +239,6 @@ jobs:
        run: wasm-pack test --chrome --headless -- -Z build-std="panic_abort,std" --features web

  mock-proving-tests:
-    permissions:
-      contents: read
    runs-on: non-gpu
    needs: [build, library-tests, docs, python-tests, python-integration-tests]
    steps:
@@ -321,8 +304,6 @@ jobs:
        run: cargo nextest run --release --verbose tests::mock_fixed_params_ --test-threads 32

  prove-and-verify-evm-tests:
-    permissions:
-      contents: read
    runs-on: non-gpu
    needs: [build, library-tests, docs, python-tests, python-integration-tests]
    steps:
@@ -405,44 +386,7 @@ jobs:
      - name: KZG prove and verify tests (EVM + hashed outputs)
        run: cargo nextest run --release --verbose tests_evm::kzg_evm_hashed_output_prove_and_verify --test-threads 1

-  # prove-and-verify-tests-metal:
-  #   permissions:
-  #     contents: read
-  #   runs-on: macos-13
-  #   # needs: [build, library-tests, docs]
-  #   steps:
-  #     - uses: actions/checkout@v4
-  #       with:
-  #         persist-credentials: false
-  #     - uses: actions-rs/toolchain@v1
-  #       with:
-  #         toolchain: nightly-2024-07-18
-  #         override: true
-  #         components: rustfmt, clippy
-  #     - uses: jetli/wasm-pack-action@v0.4.0
-  #       with:
-  #         # Pin to version 0.12.1
-  #         version: 'v0.12.1'
-  #     - name: Add rust-src
-  #       run: rustup component add rust-src --toolchain nightly-2024-07-18
-  #     - uses: actions/checkout@v3
-  #       with:
-  #         persist-credentials: false
-  #     - name: Use pnpm 8
-  #       uses: pnpm/action-setup@v2
-  #       with:
-  #         version: 8
-  #     - uses: baptiste0928/cargo-install@v1
-  #       with:
-  #         crate: cargo-nextest
-  #         locked: true
-  #     - name: KZG prove and verify tests (public outputs)
-  #       run: cargo nextest run --release --features macos-metal --verbose tests::kzg_prove_and_verify_::t --no-capture
-
-
  prove-and-verify-tests:
-    permissions:
-      contents: read
    runs-on: non-gpu
    needs: [build, library-tests, docs]
    steps:
@@ -560,8 +504,6 @@ jobs:
  #       run: cargo nextest run --release --verbose tests::kzg_prove_and_verify_hashed --features icicle  --test-threads 1

  prove-and-verify-mock-aggr-tests:
-    permissions:
-      contents: read
    runs-on: self-hosted
    needs: [build, library-tests, docs, python-tests, python-integration-tests]
    steps:
@@ -601,8 +543,6 @@ jobs:
  #       run: cargo nextest run --verbose tests_aggr::kzg_aggr_prove_and_verify_  --features icicle --test-threads 1 -- --include-ignored

  prove-and-verify-aggr-tests:
-    permissions:
-      contents: read
    runs-on: large-self-hosted
    needs: [build, library-tests, docs, python-tests, python-integration-tests]
    steps:
@@ -622,8 +562,6 @@ jobs:
        run: cargo nextest run --release --verbose tests_aggr::kzg_aggr_prove_and_verify_  --test-threads 4 -- --include-ignored

  prove-and-verify-aggr-evm-tests:
-    permissions:
-      contents: read
    runs-on: large-self-hosted
    needs: [build, library-tests, docs, python-tests, python-integration-tests]
    steps:
@@ -647,8 +585,6 @@ jobs:
        run: cargo nextest run  --release --verbose tests_evm::kzg_evm_aggr_prove_and_verify_::t --test-threads 4 -- --include-ignored

  examples:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest-32-cores
    needs: [build, library-tests, docs]
    steps:
@@ -668,8 +604,6 @@ jobs:
        run: cargo nextest run --release tests_examples

  python-tests:
-    permissions:
-      contents: read
    runs-on: non-gpu
    needs: [build, library-tests, docs]
    steps:
@@ -698,8 +632,6 @@ jobs:
        run: source .env/bin/activate; pip install pytest-asyncio; pytest -vv

  accuracy-measurement-tests:
-    permissions:
-      contents: read
    runs-on: non-gpu
    needs: [build, library-tests, docs, python-tests, python-integration-tests]
    steps:
@@ -732,8 +664,6 @@ jobs:
        run: source .env/bin/activate; cargo nextest run --release --verbose tests::resources_accuracy_measurement_public_outputs_

  python-integration-tests:
-    permissions:
-      contents: read
    runs-on: large-self-hosted
    services:
      # Label used to access the service container
@@ -779,8 +709,6 @@ jobs:
        run: python -m venv .env --clear; source .env/bin/activate; pip install -r requirements.txt; python -m ensurepip --upgrade
      - name: Build python ezkl
        run: source .env/bin/activate; unset CONDA_PREFIX; maturin develop --features python-bindings --release
-      - name: Voice tutorial
-        run: source .env/bin/activate; cargo nextest run py_tests::tests::voice_
      - name: Neural bow
        run: source .env/bin/activate; cargo nextest run py_tests::tests::neural_bag_of_words_ --no-capture
      - name: Felt conversion
@@ -800,14 +728,14 @@ jobs:
      #     chmod 600 /home/ubuntu/.kaggle/kaggle.json
      - name: All notebooks
        run: source .env/bin/activate; cargo nextest run py_tests::tests::run_notebook_ --test-threads 1
+      - name: Voice tutorial
+        run: source .env/bin/activate; cargo nextest run py_tests::tests::voice_
      - name: NBEATS tutorial
        run: source .env/bin/activate; cargo nextest run py_tests::tests::nbeats_
      # - name: Reusable verifier tutorial
        # run: source .env/bin/activate; cargo nextest run py_tests::tests::reusable_

  ios-integration-tests:
-      permissions:
-        contents: read
      runs-on: macos-latest
      steps:
        - uses: actions/checkout@v4
@@ -826,8 +754,6 @@ jobs:
          run: CARGO_BUILD_TARGET=aarch64-apple-darwin RUSTUP_TOOLCHAIN=nightly-2024-07-18-aarch64-apple-darwin cargo test --test ios_integration_tests --features ios-bindings-test --no-default-features

  swift-package-tests:
-      permissions:
-        contents: read
      runs-on: macos-latest
      needs: [ios-integration-tests]

--- a/.github/workflows/static-analysis.yml
+++ b/.github/workflows/static-analysis.yml
@@ -8,9 +8,8 @@ on:

 jobs:
  analyze:
-    permissions:
-      contents: read
    runs-on: ubuntu-latest
+
    steps:
      - uses: actions/checkout@v4
        with:
--- a/.github/workflows/swift-pm.yml
+++ b/.github/workflows/swift-pm.yml
@@ -9,9 +9,6 @@ on:

 jobs:
  build-and-update:
-    permissions:
-      contents: read
-      packages: write
    runs-on: macos-latest
    env:
      EZKL_SWIFT_PACKAGE_REPO: github.com/zkonduit/ezkl-swift-package.git
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -1835,16 +1835,6 @@ dependencies = [
 "syn 2.0.90",
 ]

-[[package]]
-name = "env_filter"
-version = "0.1.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "186e05a59d4c50738528153b83b0b0194d3a29507dfec16eccd4b342903397d0"
-dependencies = [
- "log",
- "regex",
-]
-
 [[package]]
 name = "env_logger"
 version = "0.10.2"
@@ -1858,19 +1848,6 @@ dependencies = [
 "termcolor",
 ]

-[[package]]
-name = "env_logger"
-version = "0.11.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dcaee3d8e3cfc3fd92428d477bc97fc29ec8716d180c0d74c643bb26166660e0"
-dependencies = [
- "anstream",
- "anstyle",
- "env_filter",
- "humantime",
- "log",
-]
-
 [[package]]
 name = "equivalent"
 version = "1.0.1"
@@ -1946,7 +1923,7 @@ dependencies = [
 "console_error_panic_hook",
 "criterion 0.5.1",
 "ecc",
- "env_logger 0.10.2",
+ "env_logger",
 "ethabi",
 "foundry-compilers",
 "gag",
@@ -1954,7 +1931,7 @@ dependencies = [
 "halo2_gadgets",
 "halo2_proofs",
 "halo2_solidity_verifier",
- "halo2curves 0.7.0 (git+https://github.com/privacy-scaling-explorations/halo2curves?rev=b753a832e92d5c86c5c997327a9cf9de86a18851)",
+ "halo2curves 0.7.0",
 "hex",
 "indicatif",
 "instant",
@@ -1962,6 +1939,7 @@ dependencies = [
 "lazy_static",
 "log",
 "maybe-rayon",
+ "metal",
 "mimalloc",
 "mnist",
 "num",
@@ -2416,14 +2394,14 @@ dependencies = [
 [[package]]
 name = "halo2_proofs"
 version = "0.3.0"
-source = "git+https://github.com/zkonduit/halo2#bf9d0057a82443be48c4779bbe14961c18fb5996#bf9d0057a82443be48c4779bbe14961c18fb5996"
+source = "git+https://github.com/zkonduit/halo2#ee4e1a09ebdb1f79f797685b78951c6034c430a6#ee4e1a09ebdb1f79f797685b78951c6034c430a6"
 dependencies = [
 "bincode",
 "blake2b_simd",
- "env_logger 0.10.2",
+ "env_logger",
 "ff",
 "group",
- "halo2curves 0.7.0 (git+https://github.com/privacy-scaling-explorations/halo2curves?rev=b753a832e92d5c86c5c997327a9cf9de86a18851)",
+ "halo2curves 0.7.0",
 "icicle-bn254",
 "icicle-core",
 "icicle-cuda-runtime",
@@ -2431,7 +2409,6 @@ dependencies = [
 "lazy_static",
 "log",
 "maybe-rayon",
- "mopro-msm",
 "rand_chacha",
 "rand_core 0.6.4",
 "rustc-hash 2.0.0",
@@ -2517,36 +2494,6 @@ dependencies = [
 "subtle",
 ]

-[[package]]
-name = "halo2curves"
-version = "0.7.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d380afeef3f1d4d3245b76895172018cfb087d9976a7cabcd5597775b2933e07"
-dependencies = [
- "blake2",
- "digest 0.10.7",
- "ff",
- "group",
- "halo2derive 0.1.0 (registry+https://github.com/rust-lang/crates.io-index)",
- "hex",
- "lazy_static",
- "num-bigint",
- "num-integer",
- "num-traits",
- "pairing",
- "pasta_curves",
- "paste",
- "rand 0.8.5",
- "rand_core 0.6.4",
- "rayon",
- "serde",
- "serde_arrays",
- "sha2",
- "static_assertions",
- "subtle",
- "unroll",
-]
-
 [[package]]
 name = "halo2curves"
 version = "0.7.0"
@@ -2556,7 +2503,7 @@ dependencies = [
 "digest 0.10.7",
 "ff",
 "group",
- "halo2derive 0.1.0 (git+https://github.com/privacy-scaling-explorations/halo2curves?rev=b753a832e92d5c86c5c997327a9cf9de86a18851)",
+ "halo2derive",
 "hex",
 "lazy_static",
 "num-bigint",
@@ -2576,20 +2523,6 @@ dependencies = [
 "unroll",
 ]

-[[package]]
-name = "halo2derive"
-version = "0.1.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bdb99e7492b4f5ff469d238db464131b86c2eaac814a78715acba369f64d2c76"
-dependencies = [
- "num-bigint",
- "num-integer",
- "num-traits",
- "proc-macro2",
- "quote",
- "syn 1.0.109",
-]
-
 [[package]]
 name = "halo2derive"
 version = "0.1.0"
@@ -3350,8 +3283,7 @@ dependencies = [
 [[package]]
 name = "metal"
 version = "0.29.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7ecfd3296f8c56b7c1f6fbac3c71cefa9d78ce009850c45000015f206dc7fa21"
+source = "git+https://github.com/gfx-rs/metal-rs#0e1918b34689c4b8cd13a43372f9898680547ee9"
 dependencies = [
 "bitflags 2.5.0",
 "block",
@@ -3422,28 +3354,6 @@ dependencies = [
 "byteorder",
 ]

-[[package]]
-name = "mopro-msm"
-version = "0.1.0"
-source = "git+https://github.com/zkonduit/metal-msm-gpu-acceleration.git#be5f647b1a6c1a6ea9024390744a2b4d87f5d002"
-dependencies = [
- "bincode",
- "env_logger 0.11.6",
- "halo2curves 0.7.0 (registry+https://github.com/rust-lang/crates.io-index)",
- "instant",
- "itertools 0.13.0",
- "lazy_static",
- "log",
- "metal",
- "objc",
- "once_cell",
- "rand 0.8.5",
- "rayon",
- "serde",
- "thiserror",
- "walkdir",
-]
-
 [[package]]
 name = "native-tls"
 version = "0.2.11"
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -91,6 +91,7 @@ pyo3-async-runtimes = { git = "https://github.com/PyO3/pyo3-async-runtimes", ver
 pyo3-log = { version = "0.12.0", default-features = false, optional = true }
 tract-onnx = { git = "https://github.com/sonos/tract/", rev = "37132e0397d0a73e5bd3a8615d932dabe44f6736", default-features = false, optional = true }
 tabled = { version = "0.12.0", optional = true }
+metal = { git = "https://github.com/gfx-rs/metal-rs", optional = true }
 objc = { version = "0.2.4", optional = true }
 mimalloc = { version = "0.1", optional = true }
 pyo3-stub-gen = { version = "0.6.0", optional = true }
@@ -276,14 +277,13 @@ icicle = ["halo2_proofs/icicle_gpu"]
 empty-cmd = []
 no-banner = []
 no-update = []
-macos-metal = ["halo2_proofs/macos"]
-ios-metal = ["halo2_proofs/ios"]
+

 [patch.'https://github.com/zkonduit/halo2']
-halo2_proofs = { git = "https://github.com/zkonduit/halo2#bf9d0057a82443be48c4779bbe14961c18fb5996", package = "halo2_proofs" }
+halo2_proofs = { git = "https://github.com/zkonduit/halo2#ee4e1a09ebdb1f79f797685b78951c6034c430a6", package = "halo2_proofs" }

 [patch.'https://github.com/zkonduit/halo2#0654e92bdf725fd44d849bfef3643870a8c7d50b']
-halo2_proofs = { git = "https://github.com/zkonduit/halo2#bf9d0057a82443be48c4779bbe14961c18fb5996", package = "halo2_proofs" }
+halo2_proofs = { git = "https://github.com/zkonduit/halo2#ee4e1a09ebdb1f79f797685b78951c6034c430a6", package = "halo2_proofs" }

 [patch.crates-io]
 uniffi_testing = { git = "https://github.com/ElusAegis/uniffi-rs", branch = "feat/testing-feature-build-fix" }
--- a/docs/python/src/conf.py
+++ b/docs/python/src/conf.py
@@ -1,7 +1,7 @@
 import ezkl

 project = 'ezkl'
-release = '0.0.0'
+release = '18.0.0'
 version = release


--- a/src/circuit/ops/chip.rs
+++ b/src/circuit/ops/chip.rs
@@ -592,9 +592,9 @@ impl<F: PrimeField + TensorType + PartialOrd + std::hash::Hash> BaseConfig<F> {
                        // this is 0 if the index is the same as the column index (starting from 1)

                        let col_expr = sel.clone()
-                            * (table
+                            * table
                                .selector_constructor
-                                .get_expr_at_idx(col_idx, synthetic_sel));
+                                .get_expr_at_idx(col_idx, synthetic_sel);

                        let multiplier =
                            table.selector_constructor.get_selector_val_at_idx(col_idx);
@@ -626,40 +626,6 @@ impl<F: PrimeField + TensorType + PartialOrd + std::hash::Hash> BaseConfig<F> {
                        res
                    });
                }
-
-                // add a degree-k custom constraint of the following form to the range check and
-                // static lookup configuration.
-                // 𝑚𝑢𝑙𝑡𝑖𝑠𝑒𝑙 · ∏ (𝑠𝑒𝑙 − 𝑖) = 0 where 𝑠𝑒𝑙 is the synthetic_sel, and the product is over the set of overflowed columns
-                // and 𝑚𝑢𝑙𝑡𝑖𝑠𝑒𝑙 is the selector value at the column index
-                cs.create_gate("range_check_on_sel", |cs| {
-                    let synthetic_sel = match len {
-                        1 => Expression::Constant(F::from(1)),
-                        _ => match index {
-                            VarTensor::Advice { inner: advices, .. } => {
-                                cs.query_advice(advices[x][y], Rotation(0))
-                            }
-                            _ => unreachable!(),
-                        },
-                    };
-
-                    let range_check_on_synthetic_sel = match len {
-                        1 => Expression::Constant(F::from(0)),
-                        _ => {
-                            let mut initial_expr = Expression::Constant(F::from(1));
-                            for i in 0..len {
-                                initial_expr = initial_expr
-                                    * (synthetic_sel.clone()
-                                        - Expression::Constant(F::from(i as u64)))
-                            }
-                            initial_expr
-                        }
-                    };
-
-                    let sel = cs.query_selector(multi_col_selector);
-
-                    Constraints::with_selector(sel, vec![range_check_on_synthetic_sel])
-                });
-
                self.static_lookups
                    .selectors
                    .insert((nl.clone(), x, y), multi_col_selector);
@@ -938,9 +904,9 @@ impl<F: PrimeField + TensorType + PartialOrd + std::hash::Hash> BaseConfig<F> {
                        let default_x = range_check.get_first_element(col_idx);

                        let col_expr = sel.clone()
-                            * (range_check
+                            * range_check
                                .selector_constructor
-                                .get_expr_at_idx(col_idx, synthetic_sel));
+                                .get_expr_at_idx(col_idx, synthetic_sel);

                        let multiplier = range_check
                            .selector_constructor
@@ -963,40 +929,6 @@ impl<F: PrimeField + TensorType + PartialOrd + std::hash::Hash> BaseConfig<F> {
                        res
                    });
                }
-
-                // add a degree-k custom constraint of the following form to the range check and
-                // static lookup configuration.
-                // 𝑚𝑢𝑙𝑡𝑖𝑠𝑒𝑙 · ∏ (𝑠𝑒𝑙 − 𝑖) = 0 where 𝑠𝑒𝑙 is the synthetic_sel, and the product is over the set of overflowed columns
-                // and 𝑚𝑢𝑙𝑡𝑖𝑠𝑒𝑙 is the selector value at the column index
-                cs.create_gate("range_check_on_sel", |cs| {
-                    let synthetic_sel = match len {
-                        1 => Expression::Constant(F::from(1)),
-                        _ => match index {
-                            VarTensor::Advice { inner: advices, .. } => {
-                                cs.query_advice(advices[x][y], Rotation(0))
-                            }
-                            _ => unreachable!(),
-                        },
-                    };
-
-                    let range_check_on_synthetic_sel = match len {
-                        1 => Expression::Constant(F::from(0)),
-                        _ => {
-                            let mut initial_expr = Expression::Constant(F::from(1));
-                            for i in 0..len {
-                                initial_expr = initial_expr
-                                    * (synthetic_sel.clone()
-                                        - Expression::Constant(F::from(i as u64)))
-                            }
-                            initial_expr
-                        }
-                    };
-
-                    let sel = cs.query_selector(multi_col_selector);
-
-                    Constraints::with_selector(sel, vec![range_check_on_synthetic_sel])
-                });
-
                self.range_checks
                    .selectors
                    .insert((range, x, y), multi_col_selector);
--- a/src/circuit/table.rs
+++ b/src/circuit/table.rs
@@ -132,16 +132,21 @@ impl<F: PrimeField + TensorType + PartialOrd + std::hash::Hash> Table<F> {
        (first_element, op_f.output[0])
    }

-    /// calculates the column size given the number of rows and reserved blinding rows
+    ///
    pub fn cal_col_size(logrows: usize, reserved_blinding_rows: usize) -> usize {
        2usize.pow(logrows as u32) - reserved_blinding_rows
    }
+
+    ///
+    pub fn cal_bit_range(bits: usize, reserved_blinding_rows: usize) -> usize {
+        2usize.pow(bits as u32) - reserved_blinding_rows
+    }
 }

 ///
 pub fn num_cols_required(range_len: IntegerRep, col_size: usize) -> usize {
    // number of cols needed to store the range
-    (range_len / col_size as IntegerRep) as usize + 1
+    (range_len / (col_size as IntegerRep)) as usize + 1
 }

 impl<F: PrimeField + TensorType + PartialOrd + std::hash::Hash> Table<F> {
@@ -350,11 +355,16 @@ impl<F: PrimeField + TensorType + PartialOrd + std::hash::Hash> RangeCheck<F> {
        integer_rep_to_felt(chunk * (self.col_size as IntegerRep) + self.range.0)
    }

-    /// calculates the column size
+    ///
    pub fn cal_col_size(logrows: usize, reserved_blinding_rows: usize) -> usize {
        2usize.pow(logrows as u32) - reserved_blinding_rows
    }

+    ///
+    pub fn cal_bit_range(bits: usize, reserved_blinding_rows: usize) -> usize {
+        2usize.pow(bits as u32) - reserved_blinding_rows
+    }
+
    /// get column index given input
    pub fn get_col_index(&self, input: F) -> F {
        //    range is split up into chunks of size col_size, find the chunk that input is in
--- a/src/graph/errors.rs
+++ b/src/graph/errors.rs
@@ -11,12 +11,6 @@ pub enum GraphError {
    /// Shape mismatch in circuit construction
    #[error("invalid dimensions used for node {0} ({1})")]
    InvalidDims(usize, String),
-    /// Non scalar power
-    #[error("we only support scalar powers")]
-    NonScalarPower,
-    /// Non scalar base for exponentiation
-    #[error("we only support scalar bases for exponentiation")]
-    NonScalarBase,
    /// Wrong method was called to configure an op
    #[error("wrong method was called to configure node {0} ({1})")]
    WrongMethod(usize, String),
@@ -149,7 +143,4 @@ pub enum GraphError {
    /// Invalid RunArg
    #[error("invalid RunArgs: {0}")]
    InvalidRunArgs(String),
-    /// Only nearest neighbor interpolation is supported
-    #[error("only nearest neighbor interpolation is supported")]
-    InvalidInterpolation,
 }
--- a/src/graph/utilities.rs
+++ b/src/graph/utilities.rs
@@ -44,10 +44,11 @@ use tract_onnx::tract_hir::{
    tract_core::ops::cnn::{conv::KernelFormat, MaxPool, SumPool},
 };

-/// Quantizes an iterable of f64 to a [Tensor] of IntegerRep using a fixed point representation.
+/// Quantizes an iterable of f32s to a [Tensor] of i32s using a fixed point representation.
 /// Arguments
 ///
-/// * `elem` - the element to quantize.
+/// * `vec` - the vector to quantize.
+/// * `dims` - the dimensionality of the resulting [Tensor].
 /// * `shift` - offset used in the fixed point representation.
 /// * `scale` - `2^scale` used in the fixed point representation.
 pub fn quantize_float(
@@ -84,7 +85,7 @@ pub fn scale_to_multiplier(scale: crate::Scale) -> f64 {
    f64::powf(2., scale as f64)
 }

-/// Converts a fixed point multiplier to a scale (log base 2).
+/// Converts a scale (log base 2) to a fixed point multiplier.
 pub fn multiplier_to_scale(mult: f64) -> crate::Scale {
    mult.log2().round() as crate::Scale
 }
@@ -311,9 +312,6 @@ pub fn new_op_from_onnx(
    let mut deleted_indices = vec![];
    let node = match node.op().name().as_ref() {
        "ShiftLeft" => {
-            if inputs.len() != 2 {
-                return Err(GraphError::InvalidDims(idx, "shift left".to_string()));
-            };
            // load shift amount
            if let Some(c) = inputs[1].opkind().get_mutable_constant() {
                inputs[1].decrement_use();
@@ -326,13 +324,10 @@ pub fn new_op_from_onnx(
                    out_scale: Some(input_scales[0] - raw_values[0] as i32),
                })
            } else {
-                return Err(GraphError::OpMismatch(idx, "shift left".to_string()));
+                return Err(GraphError::OpMismatch(idx, "ShiftLeft".to_string()));
            }
        }
        "ShiftRight" => {
-            if inputs.len() != 2 {
-                return Err(GraphError::InvalidDims(idx, "shift right".to_string()));
-            };
            // load shift amount
            if let Some(c) = inputs[1].opkind().get_mutable_constant() {
                inputs[1].decrement_use();
@@ -345,7 +340,7 @@ pub fn new_op_from_onnx(
                    out_scale: Some(input_scales[0] + raw_values[0] as i32),
                })
            } else {
-                return Err(GraphError::OpMismatch(idx, "shift right".to_string()));
+                return Err(GraphError::OpMismatch(idx, "ShiftRight".to_string()));
            }
        }
        "MultiBroadcastTo" => {
@@ -368,10 +363,7 @@ pub fn new_op_from_onnx(
                }
            }

-            if input_ops.len() != 3 {
-                return Err(GraphError::InvalidDims(idx, "range".to_string()));
-            }
-
+            assert_eq!(input_ops.len(), 3, "Range requires 3 inputs");
            let input_ops = input_ops
                .iter()
                .map(|x| x.get_constant().ok_or(GraphError::NonConstantRange))
@@ -427,10 +419,6 @@ pub fn new_op_from_onnx(
            if let Some(c) = inputs[1].opkind().get_mutable_constant() {
                inputs[1].decrement_use();
                deleted_indices.push(inputs.len() - 1);
-                if inputs[0].out_dims().is_empty() || inputs[0].out_dims()[0].len() <= axis {
-                    return Err(GraphError::InvalidDims(idx, "gather".to_string()));
-                }
-
                op = SupportedOp::Hybrid(crate::circuit::ops::hybrid::HybridOp::Gather {
                    dim: axis,
                    constant_idx: Some(c.raw_values.map(|x| {
@@ -459,17 +447,8 @@ pub fn new_op_from_onnx(
        "Topk" => {
            let op = load_op::<Topk>(node.op(), idx, node.op().name().to_string())?;
            let axis = op.axis;
-
-            if inputs.len() != 2 {
-                return Err(GraphError::InvalidDims(idx, "topk".to_string()));
-            };
-
            // if param_visibility.is_public() {
            let k = if let Some(c) = inputs[1].opkind().get_mutable_constant() {
-                if c.raw_values.len() != 1 {
-                    return Err(GraphError::InvalidDims(idx, "topk".to_string()));
-                }
-
                inputs[1].decrement_use();
                deleted_indices.push(inputs.len() - 1);
                c.raw_values.map(|x| x as usize)[0]
@@ -509,10 +488,6 @@ pub fn new_op_from_onnx(
            if let Some(c) = inputs[1].opkind().get_mutable_constant() {
                inputs[1].decrement_use();
                deleted_indices.push(1);
-                if c.raw_values.is_empty() {
-                    return Err(GraphError::InvalidDims(idx, "scatter elements".to_string()));
-                }
-
                op = SupportedOp::Linear(crate::circuit::ops::poly::PolyOp::ScatterElements {
                    dim: axis,
                    constant_idx: Some(c.raw_values.map(|x| x as usize)),
@@ -547,9 +522,6 @@ pub fn new_op_from_onnx(
            if let Some(c) = inputs[1].opkind().get_mutable_constant() {
                inputs[1].decrement_use();
                deleted_indices.push(1);
-                if c.raw_values.is_empty() {
-                    return Err(GraphError::InvalidDims(idx, "scatter nd".to_string()));
-                }
                op = SupportedOp::Linear(crate::circuit::ops::poly::PolyOp::ScatterND {
                    constant_idx: Some(c.raw_values.map(|x| x as usize)),
                })
@@ -583,9 +555,6 @@ pub fn new_op_from_onnx(
            if let Some(c) = inputs[1].opkind().get_mutable_constant() {
                inputs[1].decrement_use();
                deleted_indices.push(1);
-                if c.raw_values.is_empty() {
-                    return Err(GraphError::InvalidDims(idx, "gather nd".to_string()));
-                }
                op = SupportedOp::Linear(crate::circuit::ops::poly::PolyOp::GatherND {
                    batch_dims,
                    indices: Some(c.raw_values.map(|x| x as usize)),
@@ -620,9 +589,6 @@ pub fn new_op_from_onnx(
            if let Some(c) = inputs[1].opkind().get_mutable_constant() {
                inputs[1].decrement_use();
                deleted_indices.push(1);
-                if c.raw_values.is_empty() {
-                    return Err(GraphError::InvalidDims(idx, "gather elements".to_string()));
-                }
                op = SupportedOp::Linear(crate::circuit::ops::poly::PolyOp::GatherElements {
                    dim: axis,
                    constant_idx: Some(c.raw_values.map(|x| x as usize)),
@@ -718,9 +684,7 @@ pub fn new_op_from_onnx(
            };
            let op = load_op::<Reduce>(node.op(), idx, node.op().name().to_string())?;
            let axes: Vec<usize> = op.axes.into_iter().collect();
-            if axes.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "argmax".to_string()));
-            }
+            assert_eq!(axes.len(), 1, "only support argmax over one axis");

            SupportedOp::Hybrid(HybridOp::ReduceArgMax { dim: axes[0] })
        }
@@ -730,9 +694,7 @@ pub fn new_op_from_onnx(
            };
            let op = load_op::<Reduce>(node.op(), idx, node.op().name().to_string())?;
            let axes: Vec<usize> = op.axes.into_iter().collect();
-            if axes.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "argmin".to_string()));
-            }
+            assert_eq!(axes.len(), 1, "only support argmin over one axis");

            SupportedOp::Hybrid(HybridOp::ReduceArgMin { dim: axes[0] })
        }
@@ -841,9 +803,6 @@ pub fn new_op_from_onnx(
            }
        }
        "Recip" => {
-            if inputs.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "recip".to_string()));
-            };
            let in_scale = input_scales[0];
            let max_scale = std::cmp::max(scales.get_max(), in_scale);
            // If the input scale is larger than the params scale
@@ -887,9 +846,6 @@ pub fn new_op_from_onnx(
            scale: scale_to_multiplier(input_scales[0]).into(),
        }),
        "Rsqrt" => {
-            if input_scales.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "rsqrt".to_string()));
-            };
            let in_scale = input_scales[0];
            let max_scale = std::cmp::max(scales.get_max(), in_scale);
            SupportedOp::Hybrid(HybridOp::Rsqrt {
@@ -977,9 +933,7 @@ pub fn new_op_from_onnx(
            let op = load_op::<Cast>(node.op(), idx, node.op().name().to_string())?;
            let dt = op.to;

-            if input_scales.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "cast".to_string()));
-            };
+            assert_eq!(input_scales.len(), 1);

            match dt {
                DatumType::Bool
@@ -1029,11 +983,6 @@ pub fn new_op_from_onnx(

            if const_idx.len() == 1 {
                let const_idx = const_idx[0];
-
-                if inputs.len() <= const_idx {
-                    return Err(GraphError::InvalidDims(idx, "mul".to_string()));
-                }
-
                if let Some(c) = inputs[const_idx].opkind().get_mutable_constant() {
                    if c.raw_values.len() == 1 && c.raw_values[0] < 1. {
                        // if not divisible by 2 then we need to add a range check
@@ -1108,9 +1057,6 @@ pub fn new_op_from_onnx(
                    return Err(GraphError::OpMismatch(idx, "softmax".to_string()));
                }
            };
-            if input_scales.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "softmax".to_string()));
-            }

            let in_scale = input_scales[0];
            let max_scale = std::cmp::max(scales.get_max(), in_scale);
@@ -1150,42 +1096,22 @@ pub fn new_op_from_onnx(
                pool_dims: kernel_shape.to_vec(),
            })
        }
-        "Ceil" => {
-            if input_scales.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "ceil".to_string()));
-            }
-            SupportedOp::Hybrid(HybridOp::Ceil {
-                scale: scale_to_multiplier(input_scales[0]).into(),
-                legs: run_args.decomp_legs,
-            })
-        }
-        "Floor" => {
-            if input_scales.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "floor".to_string()));
-            }
-            SupportedOp::Hybrid(HybridOp::Floor {
-                scale: scale_to_multiplier(input_scales[0]).into(),
-                legs: run_args.decomp_legs,
-            })
-        }
-        "Round" => {
-            if input_scales.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "round".to_string()));
-            }
-            SupportedOp::Hybrid(HybridOp::Round {
-                scale: scale_to_multiplier(input_scales[0]).into(),
-                legs: run_args.decomp_legs,
-            })
-        }
-        "RoundHalfToEven" => {
-            if input_scales.len() != 1 {
-                return Err(GraphError::InvalidDims(idx, "roundhalftoeven".to_string()));
-            }
-            SupportedOp::Hybrid(HybridOp::RoundHalfToEven {
-                scale: scale_to_multiplier(input_scales[0]).into(),
-                legs: run_args.decomp_legs,
-            })
-        }
+        "Ceil" => SupportedOp::Hybrid(HybridOp::Ceil {
+            scale: scale_to_multiplier(input_scales[0]).into(),
+            legs: run_args.decomp_legs,
+        }),
+        "Floor" => SupportedOp::Hybrid(HybridOp::Floor {
+            scale: scale_to_multiplier(input_scales[0]).into(),
+            legs: run_args.decomp_legs,
+        }),
+        "Round" => SupportedOp::Hybrid(HybridOp::Round {
+            scale: scale_to_multiplier(input_scales[0]).into(),
+            legs: run_args.decomp_legs,
+        }),
+        "RoundHalfToEven" => SupportedOp::Hybrid(HybridOp::RoundHalfToEven {
+            scale: scale_to_multiplier(input_scales[0]).into(),
+            legs: run_args.decomp_legs,
+        }),
        "Sign" => SupportedOp::Linear(PolyOp::Sign),
        "Pow" => {
            // Extract the slope layer hyperparams from a const
@@ -1195,9 +1121,7 @@ pub fn new_op_from_onnx(
                inputs[1].decrement_use();
                deleted_indices.push(1);
                if c.raw_values.len() > 1 {
-                    return Err(GraphError::NonScalarPower);
-                } else if c.raw_values.is_empty() {
-                    return Err(GraphError::InvalidDims(idx, "pow".to_string()));
+                    unimplemented!("only support scalar pow")
                }

                let exponent = c.raw_values[0];
@@ -1214,9 +1138,7 @@ pub fn new_op_from_onnx(
                inputs[0].decrement_use();
                deleted_indices.push(0);
                if c.raw_values.len() > 1 {
-                    return Err(GraphError::NonScalarBase);
-                } else if c.raw_values.is_empty() {
-                    return Err(GraphError::InvalidDims(idx, "pow".to_string()));
+                    unimplemented!("only support scalar base")
                }

                let base = c.raw_values[0];
@@ -1226,14 +1148,10 @@ pub fn new_op_from_onnx(
                    base: base.into(),
                })
            } else {
-                return Err(GraphError::InvalidDims(idx, "pow".to_string()));
+                unimplemented!("only support constant base or pow for now")
            }
        }
        "Div" => {
-            if inputs.len() != 2 {
-                return Err(GraphError::InvalidDims(idx, "div".to_string()));
-            }
-
            let const_idx = inputs
                .iter()
                .enumerate()
@@ -1241,15 +1159,14 @@ pub fn new_op_from_onnx(
                .map(|(i, _)| i)
                .collect::<Vec<_>>();

-            if const_idx.len() > 1 || const_idx.is_empty() {
+            if const_idx.len() > 1 {
                return Err(GraphError::InvalidDims(idx, "div".to_string()));
            }

            let const_idx = const_idx[0];
+
            if const_idx != 1 {
-                return Err(GraphError::MisformedParams(
-                    "only support div with constant as second input".to_string(),
-                ));
+                unimplemented!("only support div with constant as second input")
            }

            if let Some(c) = inputs[const_idx].opkind().get_mutable_constant() {
@@ -1263,14 +1180,10 @@ pub fn new_op_from_onnx(
                        denom: denom.into(),
                    })
                } else {
-                    return Err(GraphError::MisformedParams(
-                        "only support non zero divisors of size 1".to_string(),
-                    ));
+                    unimplemented!("only support non zero divisors of size 1")
                }
            } else {
-                return Err(GraphError::MisformedParams(
-                    "only support div with constant as second input".to_string(),
-                ));
+                unimplemented!("only support div with constant as second input")
            }
        }
        "Cube" => SupportedOp::Linear(PolyOp::Pow(3)),
@@ -1410,7 +1323,7 @@ pub fn new_op_from_onnx(
            if !resize_node.contains("interpolator: Nearest")
                && !resize_node.contains("nearest: Floor")
            {
-                return Err(GraphError::InvalidInterpolation);
+                unimplemented!("Only nearest neighbor interpolation is supported")
            }
            // check if optional scale factor is present
            if inputs.len() != 2 && inputs.len() != 3 {
@@ -1514,10 +1427,6 @@ pub fn new_op_from_onnx(
            SupportedOp::Linear(PolyOp::Reshape(output_shape))
        }
        "Flatten" => {
-            if inputs.len() != 1 || inputs[0].out_dims().is_empty() {
-                return Err(GraphError::InvalidDims(idx, "flatten".to_string()));
-            };
-
            let new_dims: Vec<usize> = vec![inputs[0].out_dims()[0].iter().product::<usize>()];
            SupportedOp::Linear(PolyOp::Flatten(new_dims))
        }
@@ -1637,7 +1546,6 @@ pub fn homogenize_input_scales(
 }

 #[cfg(test)]
-/// tests for the utility module
 pub mod tests {

    use super::*;
--- a/src/tensor/mod.rs
+++ b/src/tensor/mod.rs
@@ -24,6 +24,9 @@ use std::path::PathBuf;
 pub use val::*;
 pub use var::*;

+#[cfg(feature = "metal")]
+use instant::Instant;
+
 use crate::{
    circuit::utils,
    fieldutils::{integer_rep_to_felt, IntegerRep},
@@ -37,6 +40,8 @@ use halo2_proofs::{
    poly::Rotation,
 };
 use itertools::Itertools;
+#[cfg(feature = "metal")]
+use metal::{Device, MTLResourceOptions, MTLSize};
 use std::error::Error;
 use std::fmt::Debug;
 use std::io::Read;
@@ -44,6 +49,31 @@ use std::iter::Iterator;
 use std::ops::{Add, Deref, DerefMut, Div, Mul, Neg, Range, Sub};
 use std::{cmp::max, ops::Rem};

+#[cfg(feature = "metal")]
+use std::collections::HashMap;
+
+#[cfg(feature = "metal")]
+const LIB_DATA: &[u8] = include_bytes!("metal/tensor_ops.metallib");
+
+#[cfg(feature = "metal")]
+lazy_static::lazy_static! {
+    static ref DEVICE: Device = Device::system_default().expect("no device found");
+
+    static ref LIB: metal::Library = DEVICE.new_library_with_data(LIB_DATA).unwrap();
+
+    static ref QUEUE: metal::CommandQueue = DEVICE.new_command_queue();
+
+    static ref PIPELINES: HashMap<String, metal::ComputePipelineState> = {
+        let mut map = HashMap::new();
+        for name in ["add", "sub", "mul"] {
+            let function = LIB.get_function(name, None).unwrap();
+            let pipeline = DEVICE.new_compute_pipeline_state_with_function(&function).unwrap();
+            map.insert(name.to_string(), pipeline);
+        }
+        map
+    };
+}
+
 /// The (inner) type of tensor elements.
 pub trait TensorType: Clone + Debug + 'static {
    /// Returns the zero value.
@@ -1374,6 +1404,10 @@ impl<T: TensorType + Add<Output = T> + std::marker::Send + std::marker::Sync> Ad
        let lhs = self.expand(&broadcasted_shape).unwrap();
        let rhs = rhs.expand(&broadcasted_shape).unwrap();

+        #[cfg(feature = "metal")]
+        let res = metal_tensor_op(&lhs, &rhs, "add");
+
+        #[cfg(not(feature = "metal"))]
        let res = {
            let mut res: Tensor<T> = lhs
                .par_iter()
@@ -1471,6 +1505,10 @@ impl<T: TensorType + Sub<Output = T> + std::marker::Send + std::marker::Sync> Su
        let lhs = self.expand(&broadcasted_shape).unwrap();
        let rhs = rhs.expand(&broadcasted_shape).unwrap();

+        #[cfg(feature = "metal")]
+        let res = metal_tensor_op(&lhs, &rhs, "sub");
+
+        #[cfg(not(feature = "metal"))]
        let res = {
            let mut res: Tensor<T> = lhs
                .par_iter()
@@ -1538,6 +1576,10 @@ impl<T: TensorType + Mul<Output = T> + std::marker::Send + std::marker::Sync> Mu
        let lhs = self.expand(&broadcasted_shape).unwrap();
        let rhs = rhs.expand(&broadcasted_shape).unwrap();

+        #[cfg(feature = "metal")]
+        let res = metal_tensor_op(&lhs, &rhs, "mul");
+
+        #[cfg(not(feature = "metal"))]
        let res = {
            let mut res: Tensor<T> = lhs
                .par_iter()
@@ -1769,4 +1811,66 @@ mod tests {
        let b = Tensor::<IntegerRep>::new(Some(&[1, 4]), &[2, 1]).unwrap();
        assert_eq!(a.get_slice(&[0..2, 0..1]).unwrap(), b);
    }
+
+    #[test]
+    #[cfg(feature = "metal")]
+    fn tensor_metal_int() {
+        let a = Tensor::<i64>::new(Some(&[1, 2, 3, 4]), &[2, 2]).unwrap();
+        let b = Tensor::<i64>::new(Some(&[1, 2, 3, 4]), &[2, 2]).unwrap();
+        let c = metal_tensor_op(&a, &b, "add");
+        assert_eq!(c, Tensor::new(Some(&[2, 4, 6, 8]), &[2, 2]).unwrap());
+
+        let c = metal_tensor_op(&a, &b, "sub");
+        assert_eq!(c, Tensor::new(Some(&[0, 0, 0, 0]), &[2, 2]).unwrap());
+
+        let c = metal_tensor_op(&a, &b, "mul");
+        assert_eq!(c, Tensor::new(Some(&[1, 4, 9, 16]), &[2, 2]).unwrap());
+    }
+
+    #[test]
+    #[cfg(feature = "metal")]
+    fn tensor_metal_felt() {
+        use halo2curves::bn256::Fr;
+
+        let a = Tensor::<Fr>::new(
+            Some(&[Fr::from(1), Fr::from(2), Fr::from(3), Fr::from(4)]),
+            &[2, 2],
+        )
+        .unwrap();
+        let b = Tensor::<Fr>::new(
+            Some(&[Fr::from(1), Fr::from(2), Fr::from(3), Fr::from(4)]),
+            &[2, 2],
+        )
+        .unwrap();
+
+        let c = metal_tensor_op(&a, &b, "add");
+        assert_eq!(
+            c,
+            Tensor::<Fr>::new(
+                Some(&[Fr::from(2), Fr::from(4), Fr::from(6), Fr::from(8)]),
+                &[2, 2],
+            )
+            .unwrap()
+        );
+
+        let c = metal_tensor_op(&a, &b, "sub");
+        assert_eq!(
+            c,
+            Tensor::<Fr>::new(
+                Some(&[Fr::from(0), Fr::from(0), Fr::from(0), Fr::from(0)]),
+                &[2, 2],
+            )
+            .unwrap()
+        );
+
+        let c = metal_tensor_op(&a, &b, "mul");
+        assert_eq!(
+            c,
+            Tensor::<Fr>::new(
+                Some(&[Fr::from(1), Fr::from(4), Fr::from(9), Fr::from(16)]),
+                &[2, 2],
+            )
+            .unwrap()
+        );
+    }
 }
--- a/tests/integration_tests.rs
+++ b/tests/integration_tests.rs
@@ -2795,10 +2795,7 @@ mod native_tests {
            "--features",
            "icicle",
        ];
-        #[cfg(feature = "macos-metal")]
-        let args = ["build", "--release", "--bin", "ezkl", "--features", "macos-metal"];
-        // not macos-metal and not icicle
-        #[cfg(all(not(feature = "icicle"), not(feature = "macos-metal")))]
+        #[cfg(not(feature = "icicle"))]
        let args = ["build", "--release", "--bin", "ezkl"];
        #[cfg(not(feature = "mv-lookup"))]
        let args = [
--- a/tests/py_integration_tests.rs
+++ b/tests/py_integration_tests.rs
@@ -72,10 +72,11 @@ mod py_tests {
                    "torchtext==0.17.2",
                    "torchvision==0.17.2",
                    "pandas==2.2.1",
+                    "numpy==1.26.4",
                    "seaborn==0.13.2",
                    "notebook==7.1.2",
                    "nbconvert==7.16.3",
-                    "onnx==1.17.0",
+                    "onnx==1.16.0",
                    "kaggle==1.6.8",
                    "py-solc-x==2.0.3",
                    "web3==7.5.0",
@@ -89,13 +90,12 @@ mod py_tests {
                    "xgboost==2.0.3",
                    "hummingbird-ml==0.4.11",
                    "lightgbm==4.3.0",
-                    "numpy==1.26.4",
                ])
                .status()
                .expect("failed to execute process");
            assert!(status.success());
            let status = Command::new("pip")
-                .args(["install", "numpy==1.26.4"])
+                .args(["install", "numpy==1.23"])
                .status()
                .expect("failed to execute process");