From 501dc4ceac6b29b62e56b3cae85fc8f79f1fc651 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Wed, 26 Jun 2024 00:54:51 +0900 Subject: [PATCH 01/26] =?UTF-8?q?change:=20`Onnxruntime`=E5=9E=8B=E3=82=92?= =?UTF-8?q?=E8=BF=BD=E5=8A=A0=E3=81=97=E3=80=81=E3=81=9D=E3=81=93=E3=81=8B?= =?UTF-8?q?=E3=82=89`dlopen`/`LoadLibrary*`=E3=82=92=E8=A1=8C=E3=81=86?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .github/workflows/build_and_deploy.yml | 31 +- .github/workflows/generate_document.yml | 4 +- .github/workflows/test.yml | 19 +- Cargo.lock | 125 +++++- Cargo.toml | 13 +- README.md | 2 +- crates/test_util/Cargo.toml | 2 + crates/test_util/build.rs | 34 ++ crates/test_util/src/lib.rs | 3 + crates/voicevox_core/Cargo.toml | 11 +- .../src/__internal/doctest_fixtures.rs | 12 +- crates/voicevox_core/src/blocking.rs | 11 +- crates/voicevox_core/src/devices.rs | 32 -- crates/voicevox_core/src/error.rs | 10 + crates/voicevox_core/src/infer.rs | 4 +- crates/voicevox_core/src/infer/runtimes.rs | 4 +- .../src/infer/runtimes/onnxruntime.rs | 377 +++++++++++++++++- crates/voicevox_core/src/infer/session_set.rs | 3 +- crates/voicevox_core/src/lib.rs | 26 ++ crates/voicevox_core/src/status.rs | 36 +- crates/voicevox_core/src/synthesizer.rs | 121 ++++-- crates/voicevox_core/src/tokio.rs | 11 +- crates/voicevox_core_c_api/Cargo.toml | 5 + crates/voicevox_core_c_api/build.rs | 1 + crates/voicevox_core_c_api/cbindgen.toml | 17 +- .../include/voicevox_core.h | 128 +++++- crates/voicevox_core_c_api/src/c_impls.rs | 47 ++- .../src/compatible_engine.rs | 25 +- crates/voicevox_core_c_api/src/helpers.rs | 19 +- crates/voicevox_core_c_api/src/lib.rs | 160 +++++++- crates/voicevox_core_c_api/src/result_code.rs | 5 + .../tests/e2e/assert_cdylib.rs | 12 +- .../voicevox_core_c_api/tests/e2e/log_mask.rs | 7 + crates/voicevox_core_c_api/tests/e2e/main.rs | 1 + .../tests/e2e/snapshots.toml | 32 +- .../tests/e2e/testcases/compatible_engine.rs | 13 +- ...ble_engine_load_model_before_initialize.rs | 1 + .../tests/e2e/testcases/global_info.rs | 22 +- .../tests/e2e/testcases/simple_tts.rs | 11 + .../testcases/synthesizer_new_output_json.rs | 11 + .../e2e/testcases/tts_via_audio_query.rs | 11 + .../tests/e2e/testcases/user_dict_load.rs | 11 + .../e2e/testcases/user_dict_manipulate.rs | 1 + crates/voicevox_core_java_api/Cargo.toml | 3 +- .../jp/hiroshiba/voicevoxcore/GlobalInfo.java | 18 +- .../hiroshiba/voicevoxcore/Onnxruntime.java | 131 ++++++ .../hiroshiba/voicevoxcore/Synthesizer.java | 29 +- .../InitInferenceRuntimeException.java | 14 + .../jp/hiroshiba/voicevoxcore/InfoTest.java | 5 +- .../voicevoxcore/SynthesizerTest.java | 17 +- .../jp/hiroshiba/voicevoxcore/TestUtils.java | 10 + .../hiroshiba/voicevoxcore/UserDictTest.java | 3 +- crates/voicevox_core_java_api/src/common.rs | 9 +- crates/voicevox_core_java_api/src/info.rs | 11 - crates/voicevox_core_java_api/src/lib.rs | 1 + .../voicevox_core_java_api/src/onnxruntime.rs | 56 +++ .../voicevox_core_java_api/src/synthesizer.rs | 10 +- crates/voicevox_core_python_api/Cargo.toml | 3 +- .../python/test/conftest.py | 8 + .../python/test/test_asyncio_metas.py | 7 +- .../test/test_asyncio_user_dict_load.py | 5 +- .../python/test/test_blocking_metas.py | 7 +- .../test/test_blocking_user_dict_load.py | 5 +- ...est_pseudo_raii_for_asyncio_synthesizer.py | 11 +- ...st_pseudo_raii_for_blocking_synthesizer.py | 11 +- .../python/test/test_type_stub_consts.py | 50 +++ .../python/voicevox_core/__init__.py | 4 +- .../python/voicevox_core/_load_dlls.py | 1 + .../python/voicevox_core/_rust/__init__.pyi | 22 +- .../python/voicevox_core/_rust/asyncio.pyi | 72 ++++ .../python/voicevox_core/_rust/blocking.pyi | 72 ++++ .../python/voicevox_core/asyncio.py | 4 +- .../python/voicevox_core/blocking.py | 4 +- .../voicevox_core_python_api/src/convert.rs | 9 +- crates/voicevox_core_python_api/src/lib.rs | 171 +++++++- docs/apis/c_api/doxygen/Doxyfile | 2 +- docs/usage.md | 6 +- example/cpp/unix/simple_tts.cpp | 11 +- example/cpp/windows/simple_tts/simple_tts.cpp | 11 +- example/kotlin/README.md | 8 +- example/kotlin/app/src/main/kotlin/app/App.kt | 9 +- example/python/run-asyncio.py | 31 +- example/python/run.py | 31 +- 83 files changed, 2002 insertions(+), 321 deletions(-) create mode 100644 crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Onnxruntime.java create mode 100644 crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/exceptions/InitInferenceRuntimeException.java create mode 100644 crates/voicevox_core_java_api/src/onnxruntime.rs create mode 100644 crates/voicevox_core_python_api/python/test/test_type_stub_consts.py diff --git a/.github/workflows/build_and_deploy.yml b/.github/workflows/build_and_deploy.yml index 8ac31c592..d336d1003 100644 --- a/.github/workflows/build_and_deploy.yml +++ b/.github/workflows/build_and_deploy.yml @@ -57,6 +57,7 @@ jobs: "features": "", "target": "x86_64-pc-windows-msvc", "artifact_name": "windows-x64-cpu", + "c_release_format": "plain-cdylib", "whl_local_version": "cpu", "can_skip_in_simple_test": true }, @@ -65,6 +66,7 @@ jobs: "features": "directml", "target": "x86_64-pc-windows-msvc", "artifact_name": "windows-x64-directml", + "c_release_format": "plain-cdylib", "whl_local_version": "directml", "can_skip_in_simple_test": false }, @@ -73,6 +75,7 @@ jobs: "features": "cuda", "target": "x86_64-pc-windows-msvc", "artifact_name": "windows-x64-cuda", + "c_release_format": "plain-cdylib", "whl_local_version": "cuda", "can_skip_in_simple_test": true }, @@ -81,6 +84,7 @@ jobs: "features": "", "target": "i686-pc-windows-msvc", "artifact_name": "windows-x86-cpu", + "c_release_format": "plain-cdylib", "whl_local_version": "cpu", "can_skip_in_simple_test": true }, @@ -89,6 +93,7 @@ jobs: "features": "", "target": "x86_64-unknown-linux-gnu", "artifact_name": "linux-x64-cpu", + "c_release_format": "plain-cdylib", "whl_local_version": "cpu", "can_skip_in_simple_test": true }, @@ -97,6 +102,7 @@ jobs: "features": "cuda", "target": "x86_64-unknown-linux-gnu", "artifact_name": "linux-x64-gpu", + "c_release_format": "plain-cdylib", "whl_local_version": "cuda", "can_skip_in_simple_test": false }, @@ -105,6 +111,7 @@ jobs: "features": "", "target": "aarch64-unknown-linux-gnu", "artifact_name": "linux-arm64-cpu", + "c_release_format": "plain-cdylib", "whl_local_version": "cpu", "can_skip_in_simple_test": true }, @@ -113,6 +120,7 @@ jobs: "features": "", "target": "aarch64-linux-android", "artifact_name": "android-arm64-cpu", + "c_release_format": "plain-cdylib", "can_skip_in_simple_test": true }, { @@ -120,6 +128,7 @@ jobs: "features": "", "target": "x86_64-linux-android", "artifact_name": "android-x86_64-cpu", + "c_release_format": "plain-cdylib", "can_skip_in_simple_test": true }, { @@ -127,6 +136,7 @@ jobs: "features": "", "target": "aarch64-apple-darwin", "artifact_name": "osx-arm64-cpu", + "c_release_format": "plain-cdylib", "whl_local_version": "cpu", "can_skip_in_simple_test": false }, @@ -135,6 +145,7 @@ jobs: "features": "", "target": "x86_64-apple-darwin", "artifact_name": "osx-x64-cpu", + "c_release_format": "plain-cdylib", "whl_local_version": "cpu", "can_skip_in_simple_test": true }, @@ -143,6 +154,7 @@ jobs: "features": "", "target": "aarch64-apple-ios", "artifact_name": "ios-arm64-cpu", + "c_release_format": "ios-xcframework", "can_skip_in_simple_test": true }, { @@ -150,6 +162,7 @@ jobs: "features": "", "target": "aarch64-apple-ios-sim", "artifact_name": "ios-arm64-cpu-sim", + "c_release_format": "ios-xcframework", "can_skip_in_simple_test": true }, { @@ -157,6 +170,7 @@ jobs: "features": "", "target": "x86_64-apple-ios", "artifact_name": "ios-x64-cpu", + "c_release_format": "ios-xcframework", "can_skip_in_simple_test": true } ]' @@ -244,8 +258,12 @@ jobs: - name: build voicevox_core_c_api shell: bash run: | + case ${{ matrix.c_release_format }} in + plain-cdylib) linking=onnxruntime-libloading ;; + ios-xcframework) linking=onnxruntime-link-dylib ;; + esac function build() { - cargo build -p voicevox_core_c_api -vv --features ${{ matrix.features }}, --target ${{ matrix.target }} --release + cargo build -p voicevox_core_c_api -vv --features "$linking",${{ matrix.features }} --target ${{ matrix.target }} --release } if ${{ !inputs.is_production }}; then build @@ -285,7 +303,12 @@ jobs: - name: Organize artifact run: | mkdir -p "artifact/${{ env.ASSET_NAME }}" - cp -v crates/voicevox_core_c_api/include/voicevox_core.h "artifact/${{ env.ASSET_NAME }}" + case ${{ matrix.c_release_format }} in + plain-cdylib) feature=VOICEVOX_ONNXRUNTIME_LIBLOADING ;; + ios-xcframework) feature=VOICEVOX_ONNXRUNTIME_LINK_DYLIB ;; + esac + sed 's:^//\(#define VOICEVOX_'"$feature"'\)$:\1:' crates/voicevox_core_c_api/include/voicevox_core.h \ + > "artifact/${{ env.ASSET_NAME }}/voicevox_core.h" cp -v target/${{ matrix.target }}/release/*voicevox_core.{dll,so,dylib} "artifact/${{ env.ASSET_NAME }}" || true cp -v target/${{ matrix.target }}/release/voicevox_core.dll.lib "artifact/${{ env.ASSET_NAME }}/voicevox_core.lib" || true cp -v -n target/${{ matrix.target }}/release/{,lib}onnxruntime*.{dll,so.*,so,dylib} "artifact/${{ env.ASSET_NAME }}" || true @@ -305,7 +328,7 @@ jobs: ESIGNERCKA_PASSWORD: ${{ secrets.ESIGNERCKA_PASSWORD }} ESIGNERCKA_TOTP_SECRET: ${{ secrets.ESIGNERCKA_TOTP_SECRET }} - name: Upload artifact to build XCFramework - if: contains(matrix.target, 'ios') + if: matrix.c_release_format == 'ios-xcframework' uses: actions/upload-artifact@v4 with: name: voicevox_core-${{ matrix.target }} @@ -315,7 +338,7 @@ jobs: cd artifact 7z a "../${{ env.ASSET_NAME }}.zip" "${{ env.ASSET_NAME }}" - name: Upload to Release - if: fromJson(needs.config.outputs.deploy) && !contains(matrix.target, 'ios') + if: fromJson(needs.config.outputs.deploy) && matrix.c_release_format == 'plain-cdylib' uses: softprops/action-gh-release@v2 with: prerelease: true diff --git a/.github/workflows/generate_document.yml b/.github/workflows/generate_document.yml index 531e800c3..e8ed5d8e7 100644 --- a/.github/workflows/generate_document.yml +++ b/.github/workflows/generate_document.yml @@ -46,9 +46,7 @@ jobs: with: working-directory: "docs/apis/c_api/doxygen" - name: Build voicevox_core_python_api - run: | - cargo build -p voicevox_core_c_api -vv - maturin develop --manifest-path ./crates/voicevox_core_python_api/Cargo.toml --locked + run: maturin develop --manifest-path ./crates/voicevox_core_python_api/Cargo.toml --locked # https://github.com/readthedocs/sphinx-autoapi/issues/405 - name: Workaround to make Sphinx recognize `_rust` as a module run: touch ./crates/voicevox_core_python_api/python/voicevox_core/_rust/__init__.py diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml index 14bfbddd0..c87fd727e 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test.yml @@ -72,8 +72,10 @@ jobs: with: python-version: "3.8" - uses: Swatinem/rust-cache@v2 - - run: cargo clippy -vv --all-features --tests -- -D clippy::all -D warnings --no-deps - - run: cargo clippy -vv --all-features -- -D clippy::all -D warnings --no-deps + - run: cargo clippy -vv --features directml,cuda --tests -- -D clippy::all -D warnings --no-deps + - run: cargo clippy -vv --features directml,cuda -- -D clippy::all -D warnings --no-deps + - run: cargo clippy -vv -p voicevox_core -p voicevox_core_c_api --features onnxruntime-link-dylib,directml,cuda --tests -- -D clippy::all -D warnings --no-deps + - run: cargo clippy -vv -p voicevox_core -p voicevox_core_c_api --features onnxruntime-link-dylib,directml,cuda -- -D clippy::all -D warnings --no-deps - run: cargo fmt -- --check rust-unit-test: @@ -193,11 +195,13 @@ jobs: - name: Install cargo-binstall uses: taiki-e/install-action@cargo-binstall - name: build voicevox_core_c_api - run: cargo build -p voicevox_core_c_api -vv + run: cargo build -p voicevox_core_c_api --features onnxruntime-libloading -vv - name: 必要なfileをunix用exampleのディレクトリに移動させる run: | mkdir -p example/cpp/unix/voicevox_core/ - cp -v crates/voicevox_core_c_api/include/voicevox_core.h example/cpp/unix/voicevox_core/ + sed 's:^//\(#define VOICEVOX_ONNXRUNTIME_LIBLOADING\)$:\1:' \ + crates/voicevox_core_c_api/include/voicevox_core.h \ + > example/cpp/unix/voicevox_core/voicevox_core.h cp -v target/debug/libvoicevox_core.{so,dylib} example/cpp/unix/voicevox_core/ || true cp -v target/debug/libonnxruntime.so.* example/cpp/unix/voicevox_core/ || true cp -v target/debug/libonnxruntime.*.dylib example/cpp/unix/voicevox_core/ || true @@ -235,11 +239,14 @@ jobs: - name: Install cargo-binstall uses: taiki-e/install-action@cargo-binstall - name: build voicevox_core_c_api - run: cargo build -p voicevox_core_c_api -vv + run: cargo build -p voicevox_core_c_api --features onnxruntime-libloading -vv - name: 必要なfileをexampleのディレクトリに移動させる + shell: bash run: | mkdir -p example/cpp/windows/simple_tts/lib/x64 - cp -v crates/voicevox_core_c_api/include/voicevox_core.h example/cpp/windows/simple_tts/ + sed 's:^//\(#define VOICEVOX_ONNXRUNTIME_LIBLOADING\)$:\1:' \ + crates/voicevox_core_c_api/include/voicevox_core.h \ + > example/cpp/windows/simple_tts/voicevox_core.h cp target/debug/voicevox_core.dll.lib example/cpp/windows/simple_tts/lib/x64/voicevox_core.lib - name: Add MSBuild to PATH diff --git a/Cargo.lock b/Cargo.lock index d3846b8a0..9875b69db 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -362,6 +362,32 @@ name = "camino" version = "1.1.6" source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "c59e92b5a388f549b863a7bea62612c09f24c8393560709a54558a9abdfb3b9c" +dependencies = [ + "serde", +] + +[[package]] +name = "cargo-platform" +version = "0.1.8" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "24b1f0365a6c6bb4020cd05806fd0d33c44d38046b8bd7f0e40814b9763cabfc" +dependencies = [ + "serde", +] + +[[package]] +name = "cargo_metadata" +version = "0.18.1" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "2d886547e41f740c616ae73108f6eb70afe6d940c7bc697cb30f13daec073037" +dependencies = [ + "camino", + "cargo-platform", + "semver", + "serde", + "serde_json", + "thiserror", +] [[package]] name = "cbindgen" @@ -443,7 +469,7 @@ checksum = "77ed9a53e5d4d9c573ae844bfac6872b159cb1d1585a83b29e7a64b7eef7332a" dependencies = [ "glob", "libc", - "libloading", + "libloading 0.7.3", ] [[package]] @@ -599,6 +625,26 @@ dependencies = [ "windows-sys 0.42.0", ] +[[package]] +name = "const_format" +version = "0.2.32" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "e3a214c7af3d04997541b18d432afaff4c455e79e2029079647e72fc2bd27673" +dependencies = [ + "const_format_proc_macros", +] + +[[package]] +name = "const_format_proc_macros" +version = "0.2.32" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "c7f6ff08fd20f4f299298a28e2dfa8a8ba1036e6cd2460ac1de7b425d76f2500" +dependencies = [ + "proc-macro2", + "quote", + "unicode-xid", +] + [[package]] name = "constant_time_eq" version = "0.1.5" @@ -1100,10 +1146,11 @@ dependencies = [ [[package]] name = "fs-err" -version = "2.9.0" +version = "2.11.0" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "0845fa252299212f0389d64ba26f34fa32cfe41588355f21ed507c59a0f64541" +checksum = "88a41f105fe1d5b6b34b2055e3dc59bb79b46b48b2040b9e6c7b4b5de097aa41" dependencies = [ + "autocfg", "tokio", ] @@ -1675,6 +1722,16 @@ dependencies = [ "winapi", ] +[[package]] +name = "libloading" +version = "0.8.3" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "0c2a198fb6b0eada2a8df47933734e6d35d350665a33a3593d7164fa52c75c19" +dependencies = [ + "cfg-if", + "windows-targets 0.48.0", +] + [[package]] name = "libm" version = "0.2.6" @@ -1990,9 +2047,9 @@ dependencies = [ [[package]] name = "once_cell" -version = "1.18.0" +version = "1.19.0" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "dd8b5dd2ae5ed71462c540258bedcb51965123ad7e7ccf4b9a8cafaa4a63576d" +checksum = "3fdb12b2476b595f9358c5161aa467c2438859caa136dec86c26fdd2efe17b92" [[package]] name = "opaque-debug" @@ -2502,6 +2559,26 @@ dependencies = [ "bitflags 1.3.2", ] +[[package]] +name = "ref-cast" +version = "1.0.23" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "ccf0a6f84d5f1d581da8b41b47ec8600871962f2a528115b542b362d4b744931" +dependencies = [ + "ref-cast-impl", +] + +[[package]] +name = "ref-cast-impl" +version = "1.0.23" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "bcc303e793d3734489387d205e9b186fac9c6cfacedd98cbb2e8a5943595f3e6" +dependencies = [ + "proc-macro2", + "quote", + "syn 2.0.48", +] + [[package]] name = "regex" version = "1.10.0" @@ -2793,21 +2870,24 @@ name = "semver" version = "1.0.14" source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "e25dfac463d778e353db5be2449d1cce89bd6fd23c9f1ea21310ce6e5a1b29c4" +dependencies = [ + "serde", +] [[package]] name = "serde" -version = "1.0.164" +version = "1.0.203" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "9e8c8cf938e98f769bc164923b06dce91cea1751522f46f8466461af04c9027d" +checksum = "7253ab4de971e72fb7be983802300c30b5a7f0c2e56fab8abfc6a214307c0094" dependencies = [ "serde_derive", ] [[package]] name = "serde_derive" -version = "1.0.164" +version = "1.0.203" source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "d9735b638ccc51c28bf6914d90a2e9725b377144fc612c49a611fddd1b631d68" +checksum = "500cbc0ebeb6f46627f50f3f5811ccf6bf00643be300b4c3eabc0ef55dc5b5ba" dependencies = [ "proc-macro2", "quote", @@ -3136,16 +3216,18 @@ dependencies = [ "anyhow", "bindgen 0.69.4", "camino", + "cargo_metadata", "flate2", "fs-err", "indoc", - "libloading", + "libloading 0.7.3", "once_cell", "reqwest", "serde", "serde_json", "tar", "tokio", + "voicevox-ort", "zip", ] @@ -3465,6 +3547,12 @@ version = "0.1.10" source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "c0edd1e5b14653f783770bce4a4dabb4a5108a5370a5f5d8cfe8710c361f6c8b" +[[package]] +name = "unicode-xid" +version = "0.2.4" +source = "registry+https://github.com/rust-lang/crates.io-index" +checksum = "f962df74c8c05a667b5ee8bcf162993134c104e96440b663c8daa176dc772d8c" + [[package]] name = "unindent" version = "0.2.3" @@ -3535,11 +3623,14 @@ checksum = "49874b5167b65d7193b8aba1567f5c7d93d001cafc34600cee003eda787e483f" [[package]] name = "voicevox-ort" version = "2.0.0-rc.2" -source = "git+https://github.com/VOICEVOX/ort.git?rev=a2d6ae22327869e896bf4c16828734d09516d2d9#a2d6ae22327869e896bf4c16828734d09516d2d9" +source = "git+https://github.com/qryxip/ort.git?rev=7642ebc93831e65afa114a19ebd70f9c09dee2f8#7642ebc93831e65afa114a19ebd70f9c09dee2f8" dependencies = [ + "anyhow", "half", "js-sys", + "libloading 0.8.3", "ndarray", + "once_cell", "thiserror", "tracing", "voicevox-ort-sys", @@ -3549,7 +3640,7 @@ dependencies = [ [[package]] name = "voicevox-ort-sys" version = "2.0.0-rc.2" -source = "git+https://github.com/VOICEVOX/ort.git?rev=a2d6ae22327869e896bf4c16828734d09516d2d9#a2d6ae22327869e896bf4c16828734d09516d2d9" +source = "git+https://github.com/qryxip/ort.git?rev=7642ebc93831e65afa114a19ebd70f9c09dee2f8#7642ebc93831e65afa114a19ebd70f9c09dee2f8" dependencies = [ "flate2", "sha2", @@ -3564,6 +3655,7 @@ dependencies = [ "anyhow", "async_zip", "camino", + "const_format", "derive-getters", "derive-new", "derive_more", @@ -3583,7 +3675,9 @@ dependencies = [ "open_jtalk", "ouroboros", "pretty_assertions", + "process_path", "rayon", + "ref-cast", "regex", "rstest", "rstest_reuse", @@ -3616,6 +3710,7 @@ dependencies = [ "chrono", "clap 4.0.10", "colorchoice", + "const_format", "cstr", "derive-getters", "duct", @@ -3624,12 +3719,13 @@ dependencies = [ "inventory", "itertools 0.10.5", "libc", - "libloading", + "libloading 0.7.3", "libtest-mimic", "ndarray", "ndarray-stats", "once_cell", "process_path", + "ref-cast", "regex", "serde", "serde_json", @@ -3642,6 +3738,7 @@ dependencies = [ "tracing-subscriber", "typetag", "uuid", + "voicevox-ort", "voicevox_core", ] @@ -3652,6 +3749,7 @@ dependencies = [ "android_logger", "chrono", "derive_more", + "duplicate", "easy-ext", "jni", "once_cell", @@ -3681,6 +3779,7 @@ dependencies = [ "camino", "easy-ext", "log", + "once_cell", "pyo3", "pyo3-asyncio", "pyo3-log", diff --git a/Cargo.toml b/Cargo.toml index 4ea3d3757..39043e884 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -13,11 +13,13 @@ bindgen = "0.69.4" binstall-tar = "0.4.39" bytes = "1.1.0" camino = "1.1.6" +cargo_metadata = "0.18.1" cbindgen = "0.24.3" chrono = { version = "0.4.26", default-features = false } clap = "4.0.10" color-eyre = "0.6.2" colorchoice = "1.0.0" +const_format = "0.2.32" cstr = "0.2.11" # https://github.com/dtolnay/syn/issues/1502 derive-getters = "0.2.0" derive-new = "0.5.9" @@ -29,7 +31,7 @@ educe = "0.4.23" enum-map = "3.0.0-beta.1" eyre = "0.6.8" flate2 = "1.0.25" -fs-err = "2.9.0" +fs-err = "2.11.0" futures = "0.3.26" futures-core = "0.3.25" futures-util = "0.3.25" @@ -50,7 +52,7 @@ log = "0.4.17" ndarray = "0.15.6" ndarray-stats = "0.5.1" octocrab = { version = "0.19.0", default-features = false } -once_cell = "1.18.0" +once_cell = "1.19.0" ouroboros = "0.18.0" parse-display = "0.8.2" pretty_assertions = "1.3.0" @@ -60,11 +62,12 @@ pyo3-asyncio = "0.20.0" pyo3-log = "0.9.0" quote = "1.0.33" rayon = "1.6.1" +ref-cast = "1.0.23" regex = "1.10.0" reqwest = { version = "0.11.13", default-features = false } rstest = "0.15.0" rstest_reuse = "0.6.0" -serde = "1.0.145" +serde = "1.0.203" serde_json = "1.0.85" serde_with = "3.3.0" smallvec = "1.13.1" @@ -86,8 +89,8 @@ windows = "0.43.0" zip = "0.6.3" [workspace.dependencies.voicevox-ort] -git = "https://github.com/VOICEVOX/ort.git" -rev = "a2d6ae22327869e896bf4c16828734d09516d2d9" +git = "https://github.com/qryxip/ort.git" +rev = "7642ebc93831e65afa114a19ebd70f9c09dee2f8" [workspace.dependencies.open_jtalk] git = "https://github.com/VOICEVOX/open_jtalk-rs.git" diff --git a/README.md b/README.md index 4024e9721..7f5c05c8e 100644 --- a/README.md +++ b/README.md @@ -150,7 +150,7 @@ model フォルダにある onnx モデルはダミーのため、ノイズの ```bash # DLLをビルド -cargo build --release -p voicevox_core_c_api +cargo build --release -p voicevox_core_c_api --features onnxruntime-libloading ``` DLL 用のヘッダファイルは [crates/voicevox_core_c_api/include/voicevox_core.h](https://github.com/VOICEVOX/voicevox_core/tree/main/crates/voicevox_core_c_api/include/voicevox_core.h) にあります。 diff --git a/crates/test_util/Cargo.toml b/crates/test_util/Cargo.toml index d113b57ce..fe0c2f6dd 100644 --- a/crates/test_util/Cargo.toml +++ b/crates/test_util/Cargo.toml @@ -13,6 +13,7 @@ anyhow.workspace = true tokio = { workspace = true, features = ["macros", "rt-multi-thread"] } bindgen.workspace = true camino.workspace = true +cargo_metadata.workspace = true flate2.workspace = true fs-err.workspace = true indoc.workspace = true @@ -20,6 +21,7 @@ serde = { workspace = true, features = ["derive"] } serde_json = { workspace = true, features = ["preserve_order"] } reqwest = { workspace = true, features = ["rustls-tls"] } tar.workspace = true +voicevox-ort.workspace = true zip.workspace = true [lints.rust] diff --git a/crates/test_util/build.rs b/crates/test_util/build.rs index 79e3bc90e..30755e8b9 100644 --- a/crates/test_util/build.rs +++ b/crates/test_util/build.rs @@ -6,6 +6,7 @@ use std::{ use anyhow::{anyhow, ensure}; use camino::{Utf8Path, Utf8PathBuf}; +use cargo_metadata::MetadataCommand; use flate2::read::GzDecoder; use indoc::formatdoc; use tar::Archive; @@ -27,6 +28,8 @@ async fn main() -> anyhow::Result<()> { ensure!(dic_dir.exists(), "`{dic_dir}` does not exist"); } + copy_onnxruntime(out_dir.as_ref(), dist)?; + create_sample_voice_model_file(out_dir, dist)?; generate_example_data_json(dist.as_ref())?; @@ -92,6 +95,35 @@ fn create_sample_voice_model_file(out_dir: &Utf8Path, dist: &Utf8Path) -> anyhow Ok(()) } +fn copy_onnxruntime(out_dir: &Path, dist: &Utf8Path) -> anyhow::Result<()> { + use std::env::consts::{DLL_PREFIX, DLL_SUFFIX}; + + let cargo_metadata::Metadata { + target_directory, .. + } = MetadataCommand::new() + .manifest_path(Path::new(env!("CARGO_MANIFEST_DIR")).join("Cargo.toml")) + .exec()?; + + const VERSION: &str = ort::downloaded_version!(); + let filename = &if cfg!(target_os = "linux") { + format!("libonnxruntime.so.{VERSION}") + } else if cfg!(any(target_os = "macos", target_os = "ios")) { + format!("libonnxruntime.{VERSION}.dylib") + } else { + format!("{DLL_PREFIX}onnxruntime{DLL_SUFFIX}") + }; + let src = &target_directory.join("debug").join(filename); + let dst_dir = &dist.join("lib"); + let dst = &dst_dir.join(filename); + fs_err::create_dir_all(dst_dir)?; + fs_err::copy(src, dst)?; + println!("cargo:rerun-if-changed={src}"); + + fs_err::write(out_dir.join("onnxruntime-dylib-path.txt"), dst.as_str())?; + + Ok(()) +} + /// OpenJTalkの辞書をダウンロードして展開する。 async fn download_open_jtalk_dict(dist: &Path) -> anyhow::Result<()> { let download_url = format!( @@ -187,6 +219,8 @@ fn generate_c_api_rs_bindings(out_dir: &Utf8Path) -> anyhow::Result<()> { bindgen::Builder::default() .header(C_BINDINGS_PATH) .header(ADDITIONAL_C_BINDINGS_PATH) + // we test for `--feature onnxruntime-libloading` + .clang_arg("-DVOICEVOX_ONNXRUNTIME_LIBLOADING=") .parse_callbacks(Box::new(bindgen::CargoCallbacks::new())) .dynamic_library_name("CApi") .generate()? diff --git a/crates/test_util/src/lib.rs b/crates/test_util/src/lib.rs index f234d7f76..6473e438e 100644 --- a/crates/test_util/src/lib.rs +++ b/crates/test_util/src/lib.rs @@ -23,6 +23,9 @@ pub use self::typing::{ DecodeExampleData, DurationExampleData, ExampleData, IntonationExampleData, }; +pub const ONNXRUNTIME_DYLIB_PATH: &str = + include_str!(concat!(env!("OUT_DIR"), "/onnxruntime-dylib-path.txt")); + pub const OPEN_JTALK_DIC_DIR: &str = concat!( env!("CARGO_MANIFEST_DIR"), "/data/open_jtalk_dic_utf_8-1.11" diff --git a/crates/voicevox_core/Cargo.toml b/crates/voicevox_core/Cargo.toml index 9957c1373..00cc1dd63 100644 --- a/crates/voicevox_core/Cargo.toml +++ b/crates/voicevox_core/Cargo.toml @@ -4,8 +4,14 @@ version.workspace = true edition.workspace = true publish.workspace = true +[package.metadata.docs.rs] +features = ["onnxruntime-libloading", "onnxruntime-link-dylib"] +rustdoc-args = ["--cfg", "docsrs"] + [features] default = [] +onnxruntime-libloading = ["voicevox-ort/load-dynamic"] +onnxruntime-link-dylib = [] cuda = ["voicevox-ort/cuda"] directml = ["voicevox-ort/directml"] @@ -13,6 +19,7 @@ directml = ["voicevox-ort/directml"] anyhow.workspace = true async_zip = { workspace = true, features = ["deflate"] } camino.workspace = true +const_format.workspace = true derive-getters.workspace = true derive-new.workspace = true derive_more.workspace = true @@ -29,7 +36,9 @@ ndarray.workspace = true once_cell.workspace = true open_jtalk.workspace = true ouroboros.workspace = true +process_path.workspace = true rayon.workspace = true +ref-cast.workspace = true regex.workspace = true serde = { workspace = true, features = ["derive", "rc"] } serde_json = { workspace = true, features = ["preserve_order"] } @@ -41,8 +50,8 @@ thiserror.workspace = true tokio = { workspace = true, features = ["rt"] } # FIXME: feature-gateする tracing.workspace = true uuid = { workspace = true, features = ["v4", "serde"] } +voicevox-ort = { workspace = true, features = ["download-binaries", "__init-for-voicevox"] } voicevox_core_macros = { path = "../voicevox_core_macros" } -voicevox-ort = { workspace = true, features = ["ndarray", "download-binaries"] } zip.workspace = true [dev-dependencies] diff --git a/crates/voicevox_core/src/__internal/doctest_fixtures.rs b/crates/voicevox_core/src/__internal/doctest_fixtures.rs index f314845fe..1d881856f 100644 --- a/crates/voicevox_core/src/__internal/doctest_fixtures.rs +++ b/crates/voicevox_core/src/__internal/doctest_fixtures.rs @@ -1,4 +1,4 @@ -use std::path::Path; +use std::{ffi::OsString, path::Path}; use camino::Utf8Path; @@ -6,9 +6,19 @@ use crate::{AccelerationMode, InitializeOptions}; pub async fn synthesizer_with_sample_voice_model( voice_model_path: impl AsRef, + #[cfg_attr(feature = "onnxruntime-link-dylib", allow(unused_variables))] onnxruntime_dylib_path: impl Into< + OsString, + >, open_jtalk_dic_dir: impl AsRef, ) -> anyhow::Result> { let syntesizer = crate::tokio::Synthesizer::new( + #[cfg(feature = "onnxruntime-libloading")] + crate::tokio::Onnxruntime::load_once() + .filename(onnxruntime_dylib_path) + .exec() + .await?, + #[cfg(feature = "onnxruntime-link-dylib")] + crate::tokio::Onnxruntime::init_once().await?, crate::tokio::OpenJtalk::new(open_jtalk_dic_dir).await?, &InitializeOptions { acceleration_mode: AccelerationMode::Cpu, diff --git a/crates/voicevox_core/src/blocking.rs b/crates/voicevox_core/src/blocking.rs index aa600c598..a19a88ccc 100644 --- a/crates/voicevox_core/src/blocking.rs +++ b/crates/voicevox_core/src/blocking.rs @@ -1,6 +1,13 @@ //! ブロッキング版API。 pub use crate::{ - engine::open_jtalk::blocking::OpenJtalk, synthesizer::blocking::Synthesizer, - user_dict::dict::blocking::UserDict, voice_model::blocking::VoiceModel, + engine::open_jtalk::blocking::OpenJtalk, infer::runtimes::onnxruntime::blocking::Onnxruntime, + synthesizer::blocking::Synthesizer, user_dict::dict::blocking::UserDict, + voice_model::blocking::VoiceModel, }; + +pub mod onnxruntime { + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub use crate::infer::runtimes::onnxruntime::blocking::LoadOnce; +} diff --git a/crates/voicevox_core/src/devices.rs b/crates/voicevox_core/src/devices.rs index 54b3de2f7..140105962 100644 --- a/crates/voicevox_core/src/devices.rs +++ b/crates/voicevox_core/src/devices.rs @@ -1,8 +1,6 @@ use derive_getters::Getters; use serde::{Deserialize, Serialize}; -use crate::{infer::InferenceRuntime, synthesizer::InferenceRuntimeImpl, Result}; - /// このライブラリで利用可能なデバイスの情報。 /// /// あくまで本ライブラリが対応しているデバイスの情報であることに注意。GPUが使える環境ではなかったと @@ -30,37 +28,7 @@ pub struct SupportedDevices { } impl SupportedDevices { - /// `SupportedDevices`をコンストラクトする。 - /// - /// # Example - /// - #[cfg_attr(windows, doc = "```no_run")] // https://github.com/VOICEVOX/voicevox_core/issues/537 - #[cfg_attr(not(windows), doc = "```")] - /// use voicevox_core::SupportedDevices; - /// - /// let supported_devices = SupportedDevices::create()?; - /// # - /// # Result::<_, anyhow::Error>::Ok(()) - /// ``` - pub fn create() -> Result { - ::supported_devices() - } - pub fn to_json(&self) -> serde_json::Value { serde_json::to_value(self).expect("should not fail") } } - -#[cfg(test)] -mod tests { - use rstest::rstest; - - use super::SupportedDevices; - - #[rstest] - fn supported_devices_create_works() { - let result = SupportedDevices::create(); - // 環境によって結果が変わるので、関数呼び出しが成功するかどうかの確認のみ行う - assert!(result.is_ok(), "{result:?}"); - } -} diff --git a/crates/voicevox_core/src/error.rs b/crates/voicevox_core/src/error.rs index 916964429..d0e7fced0 100644 --- a/crates/voicevox_core/src/error.rs +++ b/crates/voicevox_core/src/error.rs @@ -34,6 +34,7 @@ impl Error { match &self.0 { ErrorRepr::NotLoadedOpenjtalkDict => ErrorKind::NotLoadedOpenjtalkDict, ErrorRepr::GpuSupport => ErrorKind::GpuSupport, + ErrorRepr::InitInferenceRuntime { .. } => ErrorKind::InitInferenceRuntime, ErrorRepr::LoadModel(LoadModelError { context, .. }) => match context { LoadModelErrorKind::OpenZipFile => ErrorKind::OpenZipFile, LoadModelErrorKind::ReadZipEntry { .. } => ErrorKind::ReadZipEntry, @@ -65,6 +66,13 @@ pub(crate) enum ErrorRepr { #[error("GPU機能をサポートすることができません")] GpuSupport, + #[error("{runtime_display_name}のロードまたは初期化ができませんでした")] + InitInferenceRuntime { + runtime_display_name: &'static str, + #[source] + source: anyhow::Error, + }, + #[error(transparent)] LoadModel(#[from] LoadModelError), @@ -119,6 +127,8 @@ pub enum ErrorKind { NotLoadedOpenjtalkDict, /// GPUモードがサポートされていない。 GpuSupport, + /// 推論ライブラリのロードまたは初期化ができなかった。 + InitInferenceRuntime, /// ZIPファイルを開くことに失敗した。 OpenZipFile, /// ZIP内のファイルが読めなかった。 diff --git a/crates/voicevox_core/src/infer.rs b/crates/voicevox_core/src/infer.rs index c2cad1d7d..cffd0d524 100644 --- a/crates/voicevox_core/src/infer.rs +++ b/crates/voicevox_core/src/infer.rs @@ -18,10 +18,12 @@ pub(crate) trait InferenceRuntime: 'static { type Session: Sized + Send + 'static; type RunContext<'a>: From<&'a mut Self::Session> + PushInputTensor; - fn supported_devices() -> crate::Result; + /// このライブラリで利用可能なデバイスの情報を取得する。 + fn supported_devices(&self) -> crate::Result; #[allow(clippy::type_complexity)] fn new_session( + &self, model: impl FnOnce() -> std::result::Result, DecryptModelError>, options: InferenceSessionOptions, ) -> anyhow::Result<( diff --git a/crates/voicevox_core/src/infer/runtimes.rs b/crates/voicevox_core/src/infer/runtimes.rs index 7934027b6..e9d3d31c4 100644 --- a/crates/voicevox_core/src/infer/runtimes.rs +++ b/crates/voicevox_core/src/infer/runtimes.rs @@ -1,3 +1 @@ -mod onnxruntime; - -pub(crate) use self::onnxruntime::Onnxruntime; +pub(crate) mod onnxruntime; diff --git a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs index f8f376837..ed5b03fa3 100644 --- a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs +++ b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs @@ -15,17 +15,14 @@ use super::super::{ OutputScalarKind, OutputTensor, ParamInfo, PushInputTensor, }; -#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Debug)] -pub(crate) enum Onnxruntime {} - -impl InferenceRuntime for Onnxruntime { +// TODO: `trait AsyncRuntime`みたいなものを作って抽象化しながら同期版と非同期版に別個の役割を +// 持たせる +// （なぜそうしたいかの理由の一つとしては） +impl InferenceRuntime for self::blocking::Onnxruntime { type Session = ort::Session; type RunContext<'a> = OnnxruntimeRunContext<'a>; - fn supported_devices() -> crate::Result { - // TODO: `InferenceRuntime::init`と`InitInferenceRuntimeError`を作る - build_ort_env_once().unwrap(); - + fn supported_devices(&self) -> crate::Result { (|| { let cpu = CPUExecutionProvider::default().is_available()?; let cuda = CUDAExecutionProvider::default().is_available()?; @@ -44,6 +41,7 @@ impl InferenceRuntime for Onnxruntime { } fn new_session( + &self, model: impl FnOnce() -> std::result::Result, DecryptModelError>, options: InferenceSessionOptions, ) -> anyhow::Result<( @@ -51,9 +49,6 @@ impl InferenceRuntime for Onnxruntime { Vec>, Vec>, )> { - // TODO: `InferenceRuntime::init`と`InitInferenceRuntimeError`を作る - build_ort_env_once().unwrap(); - let mut builder = ort::Session::builder()? .with_optimization_level(GraphOptimizationLevel::Level1)? .with_intra_threads(options.cpu_num_threads.into())?; @@ -181,12 +176,6 @@ impl InferenceRuntime for Onnxruntime { } } -fn build_ort_env_once() -> ort::Result<()> { - static ONCE: once_cell::sync::OnceCell<()> = once_cell::sync::OnceCell::new(); - ONCE.get_or_try_init(|| ort::init().with_name(env!("CARGO_PKG_NAME")).commit())?; - Ok(()) -} - pub(crate) struct OnnxruntimeRunContext<'sess> { sess: &'sess ort::Session, inputs: Vec>, @@ -225,3 +214,357 @@ impl PushInputTensor for OnnxruntimeRunContext<'_> { self.push_input(tensor) } } + +pub(crate) mod blocking { + use ort::EnvHandle; + use ref_cast::{ref_cast_custom, RefCastCustom}; + + use crate::{error::ErrorRepr, SupportedDevices}; + + use super::super::super::InferenceRuntime; + + /// ONNX Runtime。 + /// + /// シングルトンであり、インスタンスは高々一つ。 + /// + /// # Rust APIにおけるインスタンスの共有 + /// + /// インスタンスは[voicevox-ort]側に作られる。Rustのクレートとしてこのライブラリを利用する場合、 + /// Tokio版APIやvoicevox-ortを利用する他クレートともインスタンスが共有される。 + /// + /// ``` + /// # use voicevox_core as another_lib; + /// # + /// # fn main() -> anyhow::Result<()> { + /// # if cfg!(windows) { + /// # // Windows\System32\onnxruntime.dllを回避 + /// # voicevox_core::blocking::Onnxruntime::load_once() + /// # .filename(test_util::ONNXRUNTIME_DYLIB_PATH) + /// # .exec()?; + /// # } + /// let ort1 = voicevox_core::blocking::Onnxruntime::load_once().exec()?; + /// let ort2 = another_lib::tokio::Onnxruntime::get().expect("`ort1`と同一のはず"); + /// assert_eq!(ptr_addr(ort1), ptr_addr(ort2)); + /// + /// fn ptr_addr(obj: &impl Sized) -> usize { + /// obj as *const _ as _ + /// } + /// # Ok(()) + /// # } + /// ``` + /// + /// [voicevox-ort]: https://github.com/VOICEVOX/ort + #[derive(Debug, RefCastCustom)] + #[repr(transparent)] + pub struct Onnxruntime { + _inner: EnvHandle, + } + + impl Onnxruntime { + /// ONNX Runtimeのライブラリ名。 + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub const LIB_NAME: &'static str = "onnxruntime"; + + /// 推奨されるONNX Runtimeのバージョン。 + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub const LIB_VERSION: &'static str = ort::downloaded_version!(); + + /// [`LIB_NAME`]と[`LIB_VERSION`]からなる動的ライブラリのファイル名。 + /// + /// WindowsとAndroidでは[`LIB_UNVERSIONED_FILENAME`]と同じ。 + /// + /// [`LIB_NAME`]: Self::LIB_NAME + /// [`LIB_VERSION`]: Self::LIB_VERSION + /// [`LIB_UNVERSIONED_FILENAME`]: Self::LIB_UNVERSIONED_FILENAME + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub const LIB_VERSIONED_FILENAME: &'static str = if cfg!(target_os = "linux") { + const_format::concatcp!( + "lib", + Onnxruntime::LIB_NAME, + ".so.", + Onnxruntime::LIB_VERSION, + ) + } else if cfg!(any(target_os = "macos", target_os = "ios")) { + const_format::concatcp!( + "lib", + Onnxruntime::LIB_NAME, + ".", + Onnxruntime::LIB_VERSION, + ".dylib", + ) + } else { + Self::LIB_UNVERSIONED_FILENAME + }; + + /// [`LIB_NAME`]からなる動的ライブラリのファイル名。 + /// + /// [`LIB_NAME`]: Self::LIB_NAME + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub const LIB_UNVERSIONED_FILENAME: &'static str = const_format::concatcp!( + std::env::consts::DLL_PREFIX, + Onnxruntime::LIB_NAME, + std::env::consts::DLL_SUFFIX, + ); + + #[ref_cast_custom] + const fn new(inner: &EnvHandle) -> &Self; + + /// インスタンスが既に作られているならそれを得る。 + /// + /// 作られていなければ`None`を返す。 + pub fn get() -> Option<&'static Self> { + EnvHandle::get().map(Self::new) + } + + fn once( + init: impl FnOnce() -> anyhow::Result<&'static EnvHandle>, + ) -> crate::Result<&'static Self> { + let inner = init().map_err(|source| ErrorRepr::InitInferenceRuntime { + runtime_display_name: "ONNX Runtime", + source, + })?; + Ok(Self::new(inner)) + } + + /// ONNX Runtimeをロードして初期化する。 + /// + /// 一度成功したら、以後は引数を無視して同じ参照を返す。 + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub fn load_once() -> LoadOnce { + LoadOnce::default() + } + + /// ONNX Runtimeを初期化する。 + /// + /// 一度成功したら以後は同じ参照を返す。 + #[cfg(feature = "onnxruntime-link-dylib")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-link-dylib")))] + pub fn init_once() -> crate::Result<&'static Self> { + Self::once(|| ort::try_init(None)) + } + + #[cfg(test)] + pub(crate) fn from_test_util_data() -> anyhow::Result<&'static Self> { + #[cfg(feature = "onnxruntime-libloading")] + { + Self::load_once() + .filename(test_util::ONNXRUNTIME_DYLIB_PATH) + .exec() + .map_err(Into::into) + } + + #[cfg(feature = "onnxruntime-link-dylib")] + { + Self::init_once().map_err(Into::into) + } + } + + /// このライブラリで利用可能なデバイスの情報を取得する。 + pub fn supported_devices(&self) -> crate::Result { + ::supported_devices(self) + } + } + + /// [`Onnxruntime::load_once`]のビルダー。 + #[cfg(feature = "onnxruntime-libloading")] + pub struct LoadOnce { + filename: std::ffi::OsString, + } + + #[cfg(feature = "onnxruntime-libloading")] + impl Default for LoadOnce { + fn default() -> Self { + let filename = Onnxruntime::LIB_VERSIONED_FILENAME.into(); + Self { filename } + } + } + + #[cfg(feature = "onnxruntime-libloading")] + impl LoadOnce { + /// ONNX Runtimeのfilenameを指定する。 + /// + /// デフォルトは[`Onnxruntime::LIB_VERSIONED_FILENAME`]。filenameは + /// `dlopen`/`LoadLibraryExW`の引数に使われる。 + pub fn filename(mut self, filename: impl Into) -> Self { + self.filename = filename.into(); + self + } + + /// 実行する。 + pub fn exec(self) -> crate::Result<&'static Onnxruntime> { + Onnxruntime::once(|| ort::try_init_from(&self.filename, None)) + } + } +} + +pub(crate) mod tokio { + use ref_cast::{ref_cast_custom, RefCastCustom}; + + use crate::SupportedDevices; + + /// ONNX Runtime。 + /// + /// シングルトンであり、インスタンスは高々一つ。 + /// + /// # Rust APIにおけるインスタンスの共有 + /// + /// インスタンスは[voicevox-ort]側に作られる。Rustのクレートとしてこのライブラリを利用する場合、 + /// ブロッキング版APIやvoicevox-ortを利用する他クレートともインスタンスが共有される。 + /// + /// ``` + /// # use voicevox_core as another_lib; + /// # + /// # #[tokio::main] + /// # async fn main() -> anyhow::Result<()> { + /// # if cfg!(windows) { + /// # // Windows\System32\onnxruntime.dllを回避 + /// # voicevox_core::blocking::Onnxruntime::load_once() + /// # .filename(test_util::ONNXRUNTIME_DYLIB_PATH) + /// # .exec()?; + /// # } + /// let ort1 = voicevox_core::tokio::Onnxruntime::load_once().exec().await?; + /// let ort2 = another_lib::blocking::Onnxruntime::get().expect("`ort1`と同一のはず"); + /// assert_eq!(ptr_addr(ort1), ptr_addr(ort2)); + /// + /// fn ptr_addr(obj: &impl Sized) -> usize { + /// obj as *const _ as _ + /// } + /// # Ok(()) + /// # } + /// ``` + /// + /// [voicevox-ort]: https://github.com/VOICEVOX/ort + #[derive(Debug, RefCastCustom)] + #[repr(transparent)] + pub struct Onnxruntime(pub(crate) super::blocking::Onnxruntime); + + impl Onnxruntime { + /// ONNX Runtimeのライブラリ名。 + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + // ブロッキング版と等しいことはテストで担保 + pub const LIB_NAME: &'static str = "onnxruntime"; + + /// 推奨されるONNX Runtimeのバージョン。 + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + // ブロッキング版と等しいことはテストで担保 + pub const LIB_VERSION: &'static str = ort::downloaded_version!(); + + /// [`LIB_NAME`]と[`LIB_VERSION`]からなる動的ライブラリのファイル名。 + /// + /// WindowsとAndroidでは[`LIB_UNVERSIONED_FILENAME`]と同じ。 + /// + /// [`LIB_NAME`]: Self::LIB_NAME + /// [`LIB_VERSION`]: Self::LIB_VERSION + /// [`LIB_UNVERSIONED_FILENAME`]: Self::LIB_UNVERSIONED_FILENAME + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub const LIB_VERSIONED_FILENAME: &'static str = + super::blocking::Onnxruntime::LIB_VERSIONED_FILENAME; + + /// [`LIB_NAME`]からなる動的ライブラリのファイル名。 + /// + /// [`LIB_NAME`]: Self::LIB_NAME + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub const LIB_UNVERSIONED_FILENAME: &'static str = + super::blocking::Onnxruntime::LIB_UNVERSIONED_FILENAME; + + #[ref_cast_custom] + pub(crate) const fn from_blocking(blocking: &super::blocking::Onnxruntime) -> &Self; + + /// インスタンスが既に作られているならそれを得る。 + /// + /// 作られていなければ`None`を返す。 + pub fn get() -> Option<&'static Self> { + super::blocking::Onnxruntime::get().map(Self::from_blocking) + } + + /// ONNX Runtimeをロードして初期化する。 + /// + /// 一度成功したら、以後は引数を無視して同じ参照を返す。 + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub fn load_once() -> LoadOnce { + LoadOnce::default() + } + + /// ONNX Runtimeを初期化する。 + /// + /// 一度成功したら以後は同じ参照を返す。 + #[cfg(feature = "onnxruntime-link-dylib")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-link-dylib")))] + pub async fn init_once() -> crate::Result<&'static Self> { + let inner = crate::task::asyncify(super::blocking::Onnxruntime::init_once).await?; + Ok(Self::from_blocking(inner)) + } + + #[cfg(test)] + pub(crate) async fn from_test_util_data() -> anyhow::Result<&'static Self> { + crate::task::asyncify(super::blocking::Onnxruntime::from_test_util_data) + .await + .map(Self::from_blocking) + } + + /// このライブラリで利用可能なデバイスの情報を取得する。 + pub fn supported_devices(&self) -> crate::Result { + self.0.supported_devices() + } + } + + /// [`Onnxruntime::load_once`]のビルダー。 + #[cfg(feature = "onnxruntime-libloading")] + #[derive(Default)] + pub struct LoadOnce(super::blocking::LoadOnce); + + #[cfg(feature = "onnxruntime-libloading")] + impl LoadOnce { + /// ONNX Runtimeのfilenameを指定する。 + /// + /// デフォルトは[`Onnxruntime::LIB_VERSIONED_FILENAME`]。filenameは + /// `dlopen`/`LoadLibraryExW`の引数に使われる。 + pub fn filename(self, filename: impl Into) -> Self { + Self(self.0.filename(filename)) + } + + /// 実行する。 + pub async fn exec(self) -> crate::Result<&'static Onnxruntime> { + let inner = crate::task::asyncify(|| self.0.exec()).await?; + Ok(Onnxruntime::from_blocking(inner)) + } + } +} + +#[cfg(test)] +mod tests { + use rstest::rstest; + + #[cfg(feature = "onnxruntime-libloading")] + #[test] + fn assert_same_lib_names_and_versions() { + use pretty_assertions::assert_eq; + + assert_eq!( + super::blocking::Onnxruntime::LIB_NAME, + super::tokio::Onnxruntime::LIB_NAME, + ); + assert_eq!( + super::blocking::Onnxruntime::LIB_VERSION, + super::tokio::Onnxruntime::LIB_VERSION, + ); + } + + #[rstest] + fn supported_devices_works() { + let result = super::blocking::Onnxruntime::from_test_util_data() + .and_then(|o| o.supported_devices().map_err(Into::into)); + // 環境によって結果が変わるので、関数呼び出しが成功するかどうかの確認のみ行う + assert!(result.is_ok(), "{result:?}"); + } +} diff --git a/crates/voicevox_core/src/infer/session_set.rs b/crates/voicevox_core/src/infer/session_set.rs index cdd179680..95f081bac 100644 --- a/crates/voicevox_core/src/infer/session_set.rs +++ b/crates/voicevox_core/src/infer/session_set.rs @@ -17,6 +17,7 @@ pub(crate) struct InferenceSessionSet( impl InferenceSessionSet { pub(crate) fn new( + rt: &R, model_bytes: &EnumMap>, options: &EnumMap, ) -> anyhow::Result { @@ -27,7 +28,7 @@ impl InferenceSessionSet { ::PARAM_INFOS[op]; let (sess, actual_input_param_infos, actual_output_param_infos) = - R::new_session(|| model_file::decrypt(model_bytes), options[op])?; + rt.new_session(|| model_file::decrypt(model_bytes), options[op])?; check_param_infos(expected_input_param_infos, &actual_input_param_infos)?; check_param_infos(expected_output_param_infos, &actual_output_param_infos)?; diff --git a/crates/voicevox_core/src/lib.rs b/crates/voicevox_core/src/lib.rs index 0f34c5962..5880f1f37 100644 --- a/crates/voicevox_core/src/lib.rs +++ b/crates/voicevox_core/src/lib.rs @@ -1,5 +1,31 @@ //! 無料で使える中品質なテキスト読み上げソフトウェア、VOICEVOXのコア。 +#![cfg_attr(docsrs, feature(doc_cfg))] + +#[cfg(not(any(feature = "onnxruntime-libloading", feature = "onnxruntime-link-dylib")))] +compile_error!("either `onnxruntime-libloading` or `onnxruntime-link-dylib` must be enabled"); + +#[cfg(not(doc))] +const _: () = { + #[cfg(all(feature = "onnxruntime-libloading", feature = "onnxruntime-link-dylib"))] + compile_error!( + "`onnxruntime-libloading` and `onnxruntime-link-dylib` cannot be enabled at the same time", + ); + + // Rust APIでvoicevox-ortを他のクレートが利用する可能性を考え、voicevox-ort側とfeatureがズレ + // ないようにする + + #[cfg(feature = "onnxruntime-libloading")] + ort::assert_load_dynamic_is_enabled!( + "when `onnxruntime-libloading` is enabled,`voicevox-ort/load-dynamic` must be also enabled", + ); + + #[cfg(feature = "onnxruntime-link-dylib")] + ort::assert_load_dynamic_is_disabled!( + "when `onnxruntime-link-dylib` is enabled,`voicevox-ort/load-dynamic` must be disabled", + ); +}; + mod devices; /// cbindgen:ignore mod engine; diff --git a/crates/voicevox_core/src/status.rs b/crates/voicevox_core/src/status.rs index f590e18f4..475031f1c 100644 --- a/crates/voicevox_core/src/status.rs +++ b/crates/voicevox_core/src/status.rs @@ -21,13 +21,18 @@ use crate::{ }; pub(crate) struct Status { + pub(crate) rt: &'static R, loaded_models: std::sync::Mutex>, session_options: InferenceDomainMap, } impl Status { - pub(crate) fn new(session_options: InferenceDomainMap) -> Self { + pub(crate) fn new( + rt: &'static R, + session_options: InferenceDomainMap, + ) -> Self { Self { + rt, loaded_models: Default::default(), session_options, } @@ -44,7 +49,7 @@ impl Status { .ensure_acceptable(model_header)?; let session_sets_with_inner_ids = model_contents - .create_session_sets(&self.session_options) + .create_session_sets(self.rt, &self.session_options) .map_err(|source| LoadModelError { path: model_header.path.clone(), context: LoadModelErrorKind::InvalidModelData, @@ -310,6 +315,7 @@ impl InferenceDomainMap { fn create_session_sets( &self, + rt: &R, session_options: &InferenceDomainMap, ) -> anyhow::Result>> { duplicate! { @@ -321,7 +327,7 @@ impl InferenceDomainMap { .field .as_ref() .map(|(inner_voice_ids, model_bytes)| { - let session_set = InferenceSessionSet::new(model_bytes, &session_options.field)?; + let session_set = InferenceSessionSet::new(rt, model_bytes, &session_options.field)?; Ok::<_, anyhow::Error>((inner_voice_ids.clone(), session_set)) }) .transpose()?; @@ -348,7 +354,6 @@ mod tests { InferenceSessionOptions, }, macros::tests::assert_debug_fmt_eq, - synthesizer::InferenceRuntimeImpl, }; use super::Status; @@ -371,7 +376,10 @@ mod tests { TalkOperation::Decode => heavy_session_options, }, }; - let status = Status::::new(session_options); + let status = Status::new( + crate::blocking::Onnxruntime::from_test_util_data().unwrap(), + session_options, + ); assert_eq!( light_session_options, @@ -392,9 +400,12 @@ mod tests { #[rstest] #[tokio::test] async fn status_load_model_works() { - let status = Status::::new(InferenceDomainMap { - talk: enum_map!(_ => InferenceSessionOptions::new(0, false)), - }); + let status = Status::new( + crate::blocking::Onnxruntime::from_test_util_data().unwrap(), + InferenceDomainMap { + talk: enum_map!(_ => InferenceSessionOptions::new(0, false)), + }, + ); let model = &crate::tokio::VoiceModel::sample().await.unwrap(); let model_contents = &model.read_inference_models().await.unwrap(); let result = status.insert_model(model.header(), model_contents); @@ -405,9 +416,12 @@ mod tests { #[rstest] #[tokio::test] async fn status_is_model_loaded_works() { - let status = Status::::new(InferenceDomainMap { - talk: enum_map!(_ => InferenceSessionOptions::new(0, false)), - }); + let status = Status::new( + crate::blocking::Onnxruntime::from_test_util_data().unwrap(), + InferenceDomainMap { + talk: enum_map!(_ => InferenceSessionOptions::new(0, false)), + }, + ); let vvm = &crate::tokio::VoiceModel::sample().await.unwrap(); let model_header = vvm.header(); let model_contents = &vvm.read_inference_models().await.unwrap(); diff --git a/crates/voicevox_core/src/synthesizer.rs b/crates/voicevox_core/src/synthesizer.rs index 4b26eb56b..e96ebe9cb 100644 --- a/crates/voicevox_core/src/synthesizer.rs +++ b/crates/voicevox_core/src/synthesizer.rs @@ -1,5 +1,3 @@ -use crate::infer::runtimes::Onnxruntime; - /// [`blocking::Synthesizer::synthesis`]および[`tokio::Synthesizer::synthesis`]のオプション。 /// /// [`blocking::Synthesizer::synthesis`]: blocking::Synthesizer::synthesis @@ -68,8 +66,6 @@ pub struct InitializeOptions { pub cpu_num_threads: u16, } -pub(crate) type InferenceRuntimeImpl = Onnxruntime; - pub(crate) mod blocking { // FIXME: ここのdocのコードブロックはasync版のものなので、`tokio`モジュールの方に移した上で、 // (ブロッキング版をpublic APIにするならの話ではあるが)ブロッキング版はブロッキング版でコード例 @@ -93,16 +89,16 @@ pub(crate) mod blocking { status::Status, text_analyzer::{KanaAnalyzer, OpenJTalkAnalyzer, TextAnalyzer}, AccentPhraseModel, AudioQueryModel, FullcontextExtractor, Result, StyleId, - SupportedDevices, SynthesisOptions, VoiceModelId, VoiceModelMeta, + SynthesisOptions, VoiceModelId, VoiceModelMeta, }; - use super::{AccelerationMode, InferenceRuntimeImpl, InitializeOptions, TtsOptions}; + use super::{AccelerationMode, InitializeOptions, TtsOptions}; const DEFAULT_SAMPLING_RATE: u32 = 24000; /// 音声シンセサイザ。 pub struct Synthesizer { - pub(super) status: Status, + pub(super) status: Status, open_jtalk_analyzer: OpenJTalkAnalyzer, kana_analyzer: KanaAnalyzer, use_gpu: bool, @@ -113,22 +109,28 @@ pub(crate) mod blocking { /// /// # Example /// - #[cfg_attr(windows, doc = "```no_run")] // https://github.com/VOICEVOX/voicevox_core/issues/537 - #[cfg_attr(not(windows), doc = "```")] + /// ``` /// # #[tokio::main] /// # async fn main() -> anyhow::Result<()> { - /// # use test_util::OPEN_JTALK_DIC_DIR; + /// # use test_util::{ONNXRUNTIME_DYLIB_PATH, OPEN_JTALK_DIC_DIR}; /// # /// # const ACCELERATION_MODE: AccelerationMode = AccelerationMode::Cpu; /// # /// use std::sync::Arc; /// /// use voicevox_core::{ - /// tokio::{OpenJtalk, Synthesizer}, + /// tokio::{Onnxruntime, OpenJtalk, Synthesizer}, /// AccelerationMode, InitializeOptions, /// }; /// + /// # if cfg!(windows) { + /// # // Windows\System32\onnxruntime.dllを回避 + /// # voicevox_core::blocking::Onnxruntime::load_once() + /// # .filename(test_util::ONNXRUNTIME_DYLIB_PATH) + /// # .exec()?; + /// # } /// let mut syntesizer = Synthesizer::new( + /// Onnxruntime::load_once().exec().await?, /// Arc::new(OpenJtalk::new(OPEN_JTALK_DIC_DIR).await.unwrap()), /// &InitializeOptions { /// acceleration_mode: ACCELERATION_MODE, @@ -139,13 +141,17 @@ pub(crate) mod blocking { /// # Ok(()) /// # } /// ``` - pub fn new(open_jtalk: O, options: &InitializeOptions) -> Result { + pub fn new( + onnxruntime: &'static crate::blocking::Onnxruntime, + open_jtalk: O, + options: &InitializeOptions, + ) -> Result { #[cfg(windows)] list_windows_video_cards(); let use_gpu = match options.acceleration_mode { AccelerationMode::Auto => { - let supported_devices = SupportedDevices::create()?; + let supported_devices = onnxruntime.supported_devices()?; if cfg!(feature = "directml") { *supported_devices.dml() @@ -157,7 +163,7 @@ pub(crate) mod blocking { AccelerationMode::Gpu => true, }; - if use_gpu && !can_support_gpu_feature()? { + if use_gpu && !can_support_gpu_feature(onnxruntime)? { return Err(ErrorRepr::GpuSupport.into()); } @@ -169,13 +175,16 @@ pub(crate) mod blocking { let heavy_session_options = InferenceSessionOptions::new(options.cpu_num_threads, use_gpu); - let status = Status::new(InferenceDomainMap { - talk: enum_map! { - TalkOperation::PredictDuration - | TalkOperation::PredictIntonation => light_session_options, - TalkOperation::Decode => heavy_session_options, + let status = Status::new( + onnxruntime, + InferenceDomainMap { + talk: enum_map! { + TalkOperation::PredictDuration + | TalkOperation::PredictIntonation => light_session_options, + TalkOperation::Decode => heavy_session_options, + }, }, - }); + ); return Ok(Self { status, @@ -184,8 +193,8 @@ pub(crate) mod blocking { use_gpu, }); - fn can_support_gpu_feature() -> Result { - let supported_devices = SupportedDevices::create()?; + fn can_support_gpu_feature(onnxruntime: &crate::blocking::Onnxruntime) -> Result { + let supported_devices = onnxruntime.supported_devices()?; if cfg!(feature = "directml") { Ok(*supported_devices.dml()) @@ -195,6 +204,10 @@ pub(crate) mod blocking { } } + pub fn onnxruntime(&self) -> &'static crate::blocking::Onnxruntime { + self.status.rt + } + /// ハードウェアアクセラレーションがGPUモードか判定する。 pub fn is_gpu_mode(&self) -> bool { self.use_gpu @@ -437,13 +450,13 @@ pub(crate) mod blocking { /// /// # Example /// - #[cfg_attr(windows, doc = "```no_run")] // https://github.com/VOICEVOX/voicevox_core/issues/537 - #[cfg_attr(not(windows), doc = "```")] + /// ``` /// # #[tokio::main] /// # async fn main() -> anyhow::Result<()> { /// # let synthesizer = /// # voicevox_core::__internal::doctest_fixtures::synthesizer_with_sample_voice_model( /// # test_util::SAMPLE_VOICE_MODEL_FILE_PATH, + /// # test_util::ONNXRUNTIME_DYLIB_PATH, /// # test_util::OPEN_JTALK_DIC_DIR, /// # ) /// # .await?; @@ -677,13 +690,13 @@ pub(crate) mod blocking { /// /// # Example /// - #[cfg_attr(windows, doc = "```no_run")] // https://github.com/VOICEVOX/voicevox_core/issues/537 - #[cfg_attr(not(windows), doc = "```")] + /// ``` /// # #[tokio::main] /// # async fn main() -> anyhow::Result<()> { /// # let synthesizer = /// # voicevox_core::__internal::doctest_fixtures::synthesizer_with_sample_voice_model( /// # test_util::SAMPLE_VOICE_MODEL_FILE_PATH, + /// # test_util::ONNXRUNTIME_DYLIB_PATH, /// # test_util::OPEN_JTALK_DIC_DIR, /// # ) /// # .await?; @@ -726,13 +739,13 @@ pub(crate) mod blocking { /// /// # Example /// - #[cfg_attr(windows, doc = "```no_run")] // https://github.com/VOICEVOX/voicevox_core/issues/537 - #[cfg_attr(not(windows), doc = "```")] + /// ``` /// # #[tokio::main] /// # async fn main() -> anyhow::Result<()> { /// # let synthesizer = /// # voicevox_core::__internal::doctest_fixtures::synthesizer_with_sample_voice_model( /// # test_util::SAMPLE_VOICE_MODEL_FILE_PATH, + /// # test_util::ONNXRUNTIME_DYLIB_PATH, /// # test_util::OPEN_JTALK_DIC_DIR, /// # ) /// # .await?; @@ -759,13 +772,13 @@ pub(crate) mod blocking { /// /// # Examples /// - #[cfg_attr(windows, doc = "```no_run")] // https://github.com/VOICEVOX/voicevox_core/issues/537 - #[cfg_attr(not(windows), doc = "```")] + /// ``` /// # #[tokio::main] /// # async fn main() -> anyhow::Result<()> { /// # let synthesizer = /// # voicevox_core::__internal::doctest_fixtures::synthesizer_with_sample_voice_model( /// # test_util::SAMPLE_VOICE_MODEL_FILE_PATH, + /// # test_util::ONNXRUNTIME_DYLIB_PATH, /// # test_util::OPEN_JTALK_DIC_DIR, /// # ) /// # .await?; @@ -1135,12 +1148,20 @@ pub(crate) mod tokio { // FIXME: docを書く impl self::Synthesizer { - pub fn new(open_jtalk: O, options: &InitializeOptions) -> Result { - super::blocking::Synthesizer::new(open_jtalk, options) + pub fn new( + onnxruntime: &'static crate::tokio::Onnxruntime, + open_jtalk: O, + options: &InitializeOptions, + ) -> Result { + super::blocking::Synthesizer::new(&onnxruntime.0, open_jtalk, options) .map(Into::into) .map(Self) } + pub fn onnxruntime(&self) -> &'static crate::tokio::Onnxruntime { + crate::tokio::Onnxruntime::from_blocking(self.0.onnxruntime()) + } + pub fn is_gpu_mode(&self) -> bool { self.0.is_gpu_mode() } @@ -1305,6 +1326,9 @@ mod tests { #[tokio::test] async fn load_model_works(#[case] expected_result_at_initialized: Result<()>) { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), (), &InitializeOptions { acceleration_mode: AccelerationMode::Cpu, @@ -1328,6 +1352,9 @@ mod tests { #[tokio::test] async fn is_use_gpu_works() { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), (), &InitializeOptions { acceleration_mode: AccelerationMode::Cpu, @@ -1344,6 +1371,9 @@ mod tests { async fn is_loaded_model_by_style_id_works(#[case] style_id: u32, #[case] expected: bool) { let style_id = StyleId::new(style_id); let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), (), &InitializeOptions { acceleration_mode: AccelerationMode::Cpu, @@ -1372,6 +1402,9 @@ mod tests { #[tokio::test] async fn predict_duration_works() { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), (), &InitializeOptions { acceleration_mode: AccelerationMode::Cpu, @@ -1403,6 +1436,9 @@ mod tests { #[tokio::test] async fn predict_intonation_works() { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), (), &InitializeOptions { acceleration_mode: AccelerationMode::Cpu, @@ -1442,6 +1478,9 @@ mod tests { #[tokio::test] async fn decode_works() { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), (), &InitializeOptions { acceleration_mode: AccelerationMode::Cpu, @@ -1534,6 +1573,9 @@ mod tests { #[case] expected_kana_text: &str, ) { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), crate::tokio::OpenJtalk::new(OPEN_JTALK_DIC_DIR) .await .unwrap(), @@ -1604,6 +1646,9 @@ mod tests { #[case] expected_text_consonant_vowel_data: &TextConsonantVowelData, ) { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), crate::tokio::OpenJtalk::new(OPEN_JTALK_DIC_DIR) .await .unwrap(), @@ -1671,6 +1716,9 @@ mod tests { #[tokio::test] async fn create_accent_phrases_works_for_japanese_commas_and_periods() { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), crate::tokio::OpenJtalk::new(OPEN_JTALK_DIC_DIR) .await .unwrap(), @@ -1732,6 +1780,9 @@ mod tests { #[tokio::test] async fn mora_length_works() { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), crate::tokio::OpenJtalk::new(OPEN_JTALK_DIC_DIR) .await .unwrap(), @@ -1770,6 +1821,9 @@ mod tests { #[tokio::test] async fn mora_pitch_works() { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), crate::tokio::OpenJtalk::new(OPEN_JTALK_DIC_DIR) .await .unwrap(), @@ -1804,6 +1858,9 @@ mod tests { #[tokio::test] async fn mora_data_works() { let syntesizer = super::tokio::Synthesizer::new( + crate::tokio::Onnxruntime::from_test_util_data() + .await + .unwrap(), crate::tokio::OpenJtalk::new(OPEN_JTALK_DIC_DIR) .await .unwrap(), diff --git a/crates/voicevox_core/src/tokio.rs b/crates/voicevox_core/src/tokio.rs index 49451a310..0fdccc3c3 100644 --- a/crates/voicevox_core/src/tokio.rs +++ b/crates/voicevox_core/src/tokio.rs @@ -1,6 +1,13 @@ //! Tokio版API。 pub use crate::{ - engine::open_jtalk::tokio::OpenJtalk, synthesizer::tokio::Synthesizer, - user_dict::dict::tokio::UserDict, voice_model::tokio::VoiceModel, + engine::open_jtalk::tokio::OpenJtalk, infer::runtimes::onnxruntime::tokio::Onnxruntime, + synthesizer::tokio::Synthesizer, user_dict::dict::tokio::UserDict, + voice_model::tokio::VoiceModel, }; + +pub mod onnxruntime { + #[cfg(feature = "onnxruntime-libloading")] + #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + pub use crate::infer::runtimes::onnxruntime::tokio::LoadOnce; +} diff --git a/crates/voicevox_core_c_api/Cargo.toml b/crates/voicevox_core_c_api/Cargo.toml index f10e2a6f4..4a9110849 100644 --- a/crates/voicevox_core_c_api/Cargo.toml +++ b/crates/voicevox_core_c_api/Cargo.toml @@ -13,6 +13,8 @@ harness = false name = "e2e" [features] +onnxruntime-libloading = ["voicevox_core/onnxruntime-libloading"] +onnxruntime-link-dylib = ["voicevox_core/onnxruntime-link-dylib"] cuda = ["voicevox_core/cuda"] directml = ["voicevox_core/directml"] @@ -22,6 +24,7 @@ anstyle-query.workspace = true camino.workspace = true chrono = { workspace = true, default-features = false, features = ["clock"] } colorchoice.workspace = true +const_format.workspace = true cstr.workspace = true derive-getters.workspace = true easy-ext.workspace = true @@ -30,6 +33,7 @@ itertools.workspace = true libc.workspace = true once_cell.workspace = true process_path.workspace = true +ref-cast.workspace = true serde_json = { workspace = true, features = ["preserve_order"] } thiserror.workspace = true tracing.workspace = true @@ -55,6 +59,7 @@ tempfile.workspace = true test_util.workspace = true toml.workspace = true typetag.workspace = true +voicevox-ort.workspace = true [lints.rust] unsafe_code = "allow" # C APIのための操作 diff --git a/crates/voicevox_core_c_api/build.rs b/crates/voicevox_core_c_api/build.rs index 535e73676..6b0934882 100644 --- a/crates/voicevox_core_c_api/build.rs +++ b/crates/voicevox_core_c_api/build.rs @@ -1,3 +1,4 @@ +// TODO: voicevox_onnxruntimeになったらやめる fn main() { #[cfg(target_os = "linux")] println!("cargo:rustc-link-arg=-Wl,-rpath,$ORIGIN"); diff --git a/crates/voicevox_core_c_api/cbindgen.toml b/crates/voicevox_core_c_api/cbindgen.toml index 7615280f6..550be7ca1 100644 --- a/crates/voicevox_core_c_api/cbindgen.toml +++ b/crates/voicevox_core_c_api/cbindgen.toml @@ -55,7 +55,18 @@ after_includes = """ #else // __cplusplus #include #include -#endif // __cplusplus""" +#endif // __cplusplus + +//#define VOICEVOX_ONNXRUNTIME_LIBLOADING +//#define VOICEVOX_ONNXRUNTIME_LINK_DYLIB + +#if !(defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) || defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB)) +#error "either `VOICEVOX_ONNXRUNTIME_LIBLOADING` or `VOICEVOX_ONNXRUNTIME_LINK_DYLIB` must be enabled" +#endif + +#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) && defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB) +#error "`VOICEVOX_ONNXRUNTIME_LIBLOADING` or `VOICEVOX_ONNXRUNTIME_LINK_DYLIB` cannot be enabled at the same time" +#endif""" # Code Style Options @@ -78,3 +89,7 @@ rename_variants = "ScreamingSnakeCase" [parse] parse_deps = true include = ["voicevox_core"] + +[defines] +"feature = onnxruntime-libloading" = "VOICEVOX_ONNXRUNTIME_LIBLOADING" +"feature = onnxruntime-link-dylib" = "VOICEVOX_ONNXRUNTIME_LINK_DYLIB" diff --git a/crates/voicevox_core_c_api/include/voicevox_core.h b/crates/voicevox_core_c_api/include/voicevox_core.h index da921c869..aa4f049a8 100644 --- a/crates/voicevox_core_c_api/include/voicevox_core.h +++ b/crates/voicevox_core_c_api/include/voicevox_core.h @@ -53,6 +53,17 @@ #include #endif // __cplusplus +//#define VOICEVOX_ONNXRUNTIME_LIBLOADING +//#define VOICEVOX_ONNXRUNTIME_LINK_DYLIB + +#if !(defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) || defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB)) +#error "either `VOICEVOX_ONNXRUNTIME_LIBLOADING` or `VOICEVOX_ONNXRUNTIME_LINK_DYLIB` must be enabled" +#endif + +#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) && defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB) +#error "`VOICEVOX_ONNXRUNTIME_LIBLOADING` or `VOICEVOX_ONNXRUNTIME_LINK_DYLIB` cannot be enabled at the same time" +#endif + /** * ハードウェアアクセラレーションモードを設定する設定値。 */ @@ -102,6 +113,10 @@ enum VoicevoxResultCode * GPUモードがサポートされていない */ VOICEVOX_RESULT_GPU_SUPPORT_ERROR = 4, + /** + * 推論ライブラリのロードまたは初期化ができなかった + */ + VOICEVOX_RESULT_INIT_INFERENCE_RUNTIME_ERROR = 29, /** * スタイルIDに対するスタイルが見つからなかった */ @@ -239,6 +254,21 @@ typedef int32_t VoicevoxUserDictWordType; */ typedef struct OpenJtalkRc OpenJtalkRc; +/** + * ONNX Runtime。 + * + * シングルトンであり、インスタンスは高々一つ。 + * + * ```c + * const VoicevoxOnnxruntime *ort1; + * voicevox_onnxruntime_load_once(voicevox_make_default_load_onnxruntime_options, + * &ort1); + * const VoicevoxOnnxruntime *ort2 = voicevox_onnxruntime_get(); + * assert(ort1 == ort2); + * ``` + */ +typedef struct VoicevoxOnnxruntime VoicevoxOnnxruntime; + /** * 音声シンセサイザ。 * @@ -259,6 +289,12 @@ typedef struct VoicevoxUserDict VoicevoxUserDict; */ typedef struct VoicevoxVoiceModel VoicevoxVoiceModel; +#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) +typedef struct VoicevoxLoadOnnxruntimeOptions { + const char *filename; +} VoicevoxLoadOnnxruntimeOptions; +#endif + /** * ::voicevox_synthesizer_new のオプション。 */ @@ -336,6 +372,68 @@ typedef struct VoicevoxUserDictWord { extern "C" { #endif // __cplusplus +#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) +#ifdef _WIN32 +__declspec(dllimport) +#endif +struct VoicevoxLoadOnnxruntimeOptions voicevox_make_default_load_onnxruntime_options(void); +#endif + +/** + * ONNX Runtimeを初期化済みであれば、 ::VoicevoxOnnxruntime のインスタンスを得る。 + * + * 初期化していなければ`NULL`を返す。 + * + * @returns ::VoicevoxOnnxruntime のインスタンス + */ +#ifdef _WIN32 +__declspec(dllimport) +#endif +const struct VoicevoxOnnxruntime *voicevox_onnxruntime_get(void); + +#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) +/** + * ONNX Runtimeをロードして初期化する。 + * + * 二度目以降はパラメータに関わらず既存のインスタンスを返す。 + * + * @param [in] options オプション + * @param [out] out_onnxruntime ::VoicevoxOnnxruntime のインスタンス + * + * @returns 結果コード + * + * \safety{ + * - `options.filename`はヌル終端文字列を指し、かつ読み込みについて有効でなければならない。 + * - `out_onnxruntime`は書き込みについて有効でなければならない。 + * } + */ +#ifdef _WIN32 +__declspec(dllimport) +#endif +VoicevoxResultCode voicevox_onnxruntime_load_once(struct VoicevoxLoadOnnxruntimeOptions options, + const struct VoicevoxOnnxruntime **out_onnxruntime); +#endif + +#if defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB) +/** + * ONNX Runtimeを初期化する。 + * + * 二度目以降は既存のインスタンスを返す。 + * + * @param [out] out_onnxruntime ::VoicevoxOnnxruntime のインスタンス + * + * @returns 結果コード + * + * \safety{ + * - `out_onnxruntime`は書き込みについて有効でなければならない。 + * } + */ +#ifdef _WIN32 +__declspec(dllimport) +#endif +VoicevoxResultCode voicevox_onnxruntime_init_once(const struct VoicevoxOnnxruntime **out_onnxruntime); +#endif + /** * ::OpenJtalkRc を構築(_construct_)する。 * @@ -492,6 +590,7 @@ void voicevox_voice_model_delete(struct VoicevoxVoiceModel *model); /** * ::VoicevoxSynthesizer を構築(_construct_)する。 * + * @param [in] onnxruntime * @param [in] open_jtalk Open JTalkのオブジェクト * @param [in] options オプション * @param [out] out_synthesizer 構築先 @@ -499,6 +598,7 @@ void voicevox_voice_model_delete(struct VoicevoxVoiceModel *model); * @returns 結果コード * * \safety{ + * - `onnxruntime`は ::voicevox_onnxruntime_load_once または ::voicevox_onnxruntime_init_once で得たものでなければならない。 * - `open_jtalk`は ::voicevox_voice_model_new_from_path で得たものでなければならず、また ::voicevox_open_jtalk_rc_new で解放されていてはいけない。 * - `out_synthesizer`は書き込みについて有効でなければならない。 * } @@ -506,7 +606,8 @@ void voicevox_voice_model_delete(struct VoicevoxVoiceModel *model); #ifdef _WIN32 __declspec(dllimport) #endif -VoicevoxResultCode voicevox_synthesizer_new(const struct OpenJtalkRc *open_jtalk, +VoicevoxResultCode voicevox_synthesizer_new(const struct VoicevoxOnnxruntime *onnxruntime, + const struct OpenJtalkRc *open_jtalk, struct VoicevoxInitializeOptions options, struct VoicevoxSynthesizer **out_synthesizer); @@ -563,6 +664,22 @@ __declspec(dllimport) VoicevoxResultCode voicevox_synthesizer_unload_voice_model(const struct VoicevoxSynthesizer *synthesizer, VoicevoxVoiceModelId model_id); +/** + * ::VoicevoxOnnxruntime のインスタンスを得る。 + * + * @param [in] synthesizer 音声シンセサイザ + * + * @returns ::VoicevoxOnnxruntime のインスタンス + * + * \safety{ + * - `synthesizer`は ::voicevox_synthesizer_new で得たものでなければならず、また ::voicevox_synthesizer_delete で解放されていてはいけない。 + * } + */ +#ifdef _WIN32 +__declspec(dllimport) +#endif +const struct VoicevoxOnnxruntime *voicevox_synthesizer_get_onnxruntime(const struct VoicevoxSynthesizer *synthesizer); + /** * ハードウェアアクセラレーションがGPUモードか判定する。 * @@ -623,6 +740,7 @@ char *voicevox_synthesizer_create_metas_json(const struct VoicevoxSynthesizer *s * * あくまで本ライブラリが対応しているデバイスの情報であることに注意。GPUが使える環境ではなかったとしても`cuda`や`dml`は`true`を示しうる。 * + * @param [in] onnxruntime * @param [out] output_supported_devices_json サポートデバイス情報のJSON文字列 * * @returns 結果コード @@ -630,18 +748,20 @@ char *voicevox_synthesizer_create_metas_json(const struct VoicevoxSynthesizer *s * \example{ * ```c * char *supported_devices; - * VoicevoxResultCode result = voicevox_create_supported_devices_json(&supported_devices); + * VoicevoxResultCode result = voicevox_onnxruntime_create_supported_devices_json(onnxruntime, &supported_devices); * ``` * } * * \safety{ + * - `onnxruntime`は ::voicevox_onnxruntime_load_once または ::voicevox_onnxruntime_init_once で得たものでなければならない。 * - `output_supported_devices_json`は書き込みについて有効でなければならない。 * } */ #ifdef _WIN32 __declspec(dllimport) #endif -VoicevoxResultCode voicevox_create_supported_devices_json(char **output_supported_devices_json); +VoicevoxResultCode voicevox_onnxruntime_create_supported_devices_json(const struct VoicevoxOnnxruntime *onnxruntime, + char **output_supported_devices_json); /** * AquesTalk風記法から、AudioQueryをJSONとして生成する。 @@ -980,7 +1100,7 @@ VoicevoxResultCode voicevox_synthesizer_tts(const struct VoicevoxSynthesizer *sy * * \safety{ * - `json`は以下のAPIで得られたポインタでなくてはいけない。 - * - ::voicevox_create_supported_devices_json + * - ::voicevox_onnxruntime_create_supported_devices_json * - ::voicevox_synthesizer_create_metas_json * - ::voicevox_synthesizer_create_audio_query * - ::voicevox_synthesizer_create_accent_phrases diff --git a/crates/voicevox_core_c_api/src/c_impls.rs b/crates/voicevox_core_c_api/src/c_impls.rs index fe4afcf65..1b471a6fe 100644 --- a/crates/voicevox_core_c_api/src/c_impls.rs +++ b/crates/voicevox_core_c_api/src/c_impls.rs @@ -1,13 +1,41 @@ use std::{ffi::CString, path::Path}; use camino::Utf8Path; +use ref_cast::ref_cast_custom; use voicevox_core::{InitializeOptions, Result, VoiceModelId}; -use crate::{helpers::CApiResult, OpenJtalkRc, VoicevoxSynthesizer, VoicevoxVoiceModel}; +use crate::{ + helpers::CApiResult, OpenJtalkRc, VoicevoxOnnxruntime, VoicevoxSynthesizer, VoicevoxVoiceModel, +}; // FIXME: 中身(Rust API)を直接操作するかラッパーメソッド越しにするのかが混在していて、一貫性を // 欠いている +impl VoicevoxOnnxruntime { + #[ref_cast_custom] + fn new(rust: &voicevox_core::blocking::Onnxruntime) -> &Self; + + pub(crate) fn get() -> Option<&'static Self> { + voicevox_core::blocking::Onnxruntime::get().map(Self::new) + } + + #[cfg(feature = "onnxruntime-libloading")] + pub(crate) fn load_once(filename: &std::ffi::CStr) -> CApiResult<&'static Self> { + use crate::helpers::ensure_utf8; + + let inner = voicevox_core::blocking::Onnxruntime::load_once() + .filename(ensure_utf8(filename)?) + .exec()?; + Ok(Self::new(inner)) + } + + #[cfg(feature = "onnxruntime-link-dylib")] + pub(crate) fn init_once() -> CApiResult<&'static Self> { + let inner = voicevox_core::blocking::Onnxruntime::init_once()?; + Ok(Self::new(inner)) + } +} + impl OpenJtalkRc { pub(crate) fn new(open_jtalk_dic_dir: impl AsRef) -> Result { Ok(Self { @@ -17,12 +45,23 @@ impl OpenJtalkRc { } impl VoicevoxSynthesizer { - pub(crate) fn new(open_jtalk: &OpenJtalkRc, options: &InitializeOptions) -> Result { - let synthesizer = - voicevox_core::blocking::Synthesizer::new(open_jtalk.open_jtalk.clone(), options)?; + pub(crate) fn new( + onnxruntime: &'static VoicevoxOnnxruntime, + open_jtalk: &OpenJtalkRc, + options: &InitializeOptions, + ) -> Result { + let synthesizer = voicevox_core::blocking::Synthesizer::new( + &onnxruntime.0, + open_jtalk.open_jtalk.clone(), + options, + )?; Ok(Self { synthesizer }) } + pub(crate) fn onnxruntime(&self) -> &'static VoicevoxOnnxruntime { + VoicevoxOnnxruntime::new(self.synthesizer.onnxruntime()) + } + pub(crate) fn load_voice_model( &self, model: &voicevox_core::blocking::VoiceModel, diff --git a/crates/voicevox_core_c_api/src/compatible_engine.rs b/crates/voicevox_core_c_api/src/compatible_engine.rs index ae4d21a93..bd66210e2 100644 --- a/crates/voicevox_core_c_api/src/compatible_engine.rs +++ b/crates/voicevox_core_c_api/src/compatible_engine.rs @@ -8,11 +8,9 @@ use std::{ use libc::c_int; use once_cell::sync::Lazy; -use voicevox_core::{ - StyleId, SupportedDevices, VoiceModelId, __internal::interop::PerformInference as _, -}; +use voicevox_core::{StyleId, VoiceModelId, __internal::interop::PerformInference as _}; -use crate::init_logger_once; +use crate::{helpers::display_error, init_logger_once}; macro_rules! ensure_initialized { ($synthesizer:expr $(,)?) => { @@ -28,6 +26,15 @@ macro_rules! ensure_initialized { static ERROR_MESSAGE: Lazy> = Lazy::new(|| Mutex::new(String::new())); +static ONNXRUNTIME: Lazy<&'static voicevox_core::blocking::Onnxruntime> = Lazy::new(|| { + voicevox_core::blocking::Onnxruntime::load_once() + .exec() + .unwrap_or_else(|err| { + display_error(&err); + panic!("ONNX Runtimeをロードもしくは初期化ができなかったため、クラッシュします"); + }) +}); + struct VoiceModelSet { all_vvms: Vec, all_metas_json: CString, @@ -111,6 +118,7 @@ pub extern "C" fn initialize(use_gpu: bool, cpu_num_threads: c_int, load_all_mod init_logger_once(); let result = (|| { let synthesizer = voicevox_core::blocking::Synthesizer::new( + *ONNXRUNTIME, (), &voicevox_core::InitializeOptions { acceleration_mode: if use_gpu { @@ -196,7 +204,14 @@ pub extern "C" fn supported_devices() -> *const c_char { return SUPPORTED_DEVICES.as_ptr(); static SUPPORTED_DEVICES: Lazy = Lazy::new(|| { - CString::new(SupportedDevices::create().unwrap().to_json().to_string()).unwrap() + CString::new( + ONNXRUNTIME + .supported_devices() + .unwrap() + .to_json() + .to_string(), + ) + .unwrap() }); } diff --git a/crates/voicevox_core_c_api/src/helpers.rs b/crates/voicevox_core_c_api/src/helpers.rs index ac0cab286..5066b643a 100644 --- a/crates/voicevox_core_c_api/src/helpers.rs +++ b/crates/voicevox_core_c_api/src/helpers.rs @@ -1,5 +1,5 @@ use easy_ext::ext; -use std::{error::Error as _, ffi::CStr, fmt::Debug, iter}; +use std::{ffi::CStr, fmt::Debug, iter}; use uuid::Uuid; use voicevox_core::{AudioQueryModel, UserDictWord, VoiceModelId}; @@ -19,14 +19,6 @@ pub(crate) fn into_result_code_with_error(result: CApiResult<()>) -> VoicevoxRes } return into_result_code(result); - fn display_error(err: &CApiError) { - itertools::chain( - [err.to_string()], - iter::successors(err.source(), |&e| e.source()).map(|e| format!("Caused by: {e}")), - ) - .for_each(|msg| error!("{msg}")); - } - fn into_result_code(result: CApiResult<()>) -> VoicevoxResultCode { use voicevox_core::ErrorKind::*; use CApiError::*; @@ -37,6 +29,7 @@ pub(crate) fn into_result_code_with_error(result: CApiResult<()>) -> VoicevoxRes Err(RustApi(err)) => match err.kind() { NotLoadedOpenjtalkDict => VOICEVOX_RESULT_NOT_LOADED_OPENJTALK_DICT_ERROR, GpuSupport => VOICEVOX_RESULT_GPU_SUPPORT_ERROR, + InitInferenceRuntime => VOICEVOX_RESULT_INIT_INFERENCE_RUNTIME_ERROR, OpenZipFile => VOICEVOX_RESULT_OPEN_ZIP_FILE_ERROR, ReadZipEntry => VOICEVOX_RESULT_READ_ZIP_ENTRY_ERROR, InvalidModelFormat => VOICEVOX_RESULT_INVALID_MODEL_HEADER_ERROR, @@ -63,6 +56,14 @@ pub(crate) fn into_result_code_with_error(result: CApiResult<()>) -> VoicevoxRes } } +pub(crate) fn display_error(err: &impl std::error::Error) { + itertools::chain( + [err.to_string()], + iter::successors(err.source(), |&e| e.source()).map(|e| format!("Caused by: {e}")), + ) + .for_each(|msg| error!("{msg}")); +} + pub(crate) type CApiResult = std::result::Result; #[derive(Error, Debug)] diff --git a/crates/voicevox_core_c_api/src/lib.rs b/crates/voicevox_core_c_api/src/lib.rs index eb2e4698e..dce26dbb8 100644 --- a/crates/voicevox_core_c_api/src/lib.rs +++ b/crates/voicevox_core_c_api/src/lib.rs @@ -4,6 +4,7 @@ mod c_impls; /// cbindgen:ignore +#[cfg(feature = "onnxruntime-libloading")] mod compatible_engine; mod drop_check; mod helpers; @@ -20,6 +21,7 @@ use anstream::{AutoStream, RawStream}; use chrono::SecondsFormat; use colorchoice::ColorChoice; use derive_getters::Getters; +use ref_cast::RefCastCustom; use std::env; use std::ffi::{CStr, CString}; use std::fmt; @@ -32,7 +34,7 @@ use tracing_subscriber::EnvFilter; use uuid::Uuid; use voicevox_core::__internal::interop::IdRef as _; use voicevox_core::{AccentPhraseModel, AudioQueryModel, TtsOptions, UserDictWord}; -use voicevox_core::{StyleId, SupportedDevices, SynthesisOptions}; +use voicevox_core::{StyleId, SynthesisOptions}; fn init_logger_once() { static ONCE: Once = Once::new(); @@ -60,13 +62,7 @@ fn init_logger_once() { .with_env_filter(if env::var_os(EnvFilter::DEFAULT_ENV).is_some() { EnvFilter::from_default_env() } else { - pub const ORT_LOGGING_LEVEL: &str = if cfg!(debug_assertions) { - "info" - } else { - "warn" - }; - format!("error,voicevox_core=info,voicevox_core_c_api=info,ort={ORT_LOGGING_LEVEL}") - .into() + "error,voicevox_core=info,voicevox_core_c_api=info,ort=info".into() }) .with_timer(local_time as fn(&mut Writer<'_>) -> _) .with_ansi(ansi) @@ -92,6 +88,121 @@ fn init_logger_once() { * voicevox_core/publish.rsにある対応する関数とはこのファイルに定義してある公開関数からvoicevoxプレフィックスを取り除いた名前の関数である */ +#[cfg(feature = "onnxruntime-libloading")] +#[repr(C)] +pub struct VoicevoxLoadOnnxruntimeOptions { + filename: *const c_char, +} + +#[cfg(feature = "onnxruntime-libloading")] +#[no_mangle] +pub extern "C" fn voicevox_make_default_load_onnxruntime_options() -> VoicevoxLoadOnnxruntimeOptions +{ + use const_format::concatcp; + + const RUST_FILENAME: &str = voicevox_core::blocking::Onnxruntime::LIB_VERSIONED_FILENAME; + static C_FILENAME: &[u8] = concatcp!(RUST_FILENAME, "\0").as_bytes(); + + init_logger_once(); + + let filename = CStr::from_bytes_with_nul(C_FILENAME) + .unwrap_or_else(|e| panic!("{RUST_FILENAME:?} should not contain `\\0`: {e}")); + let filename = C_STRING_DROP_CHECKER.blacklist(filename).as_ptr(); + VoicevoxLoadOnnxruntimeOptions { filename } +} + +// https://github.com/mozilla/cbindgen/issues/967 +// FIXME: このコードブロックのコードが動くかどうか未確認 +/// ONNX Runtime。 +/// +/// シングルトンであり、インスタンスは高々一つ。 +/// +/// ```c +/// const VoicevoxOnnxruntime *ort1; +/// voicevox_onnxruntime_load_once(voicevox_make_default_load_onnxruntime_options, +/// &ort1); +/// const VoicevoxOnnxruntime *ort2 = voicevox_onnxruntime_get(); +/// assert(ort1 == ort2); +/// ``` +#[cfg(any())] +pub struct VoicevoxOnnxruntime(!); + +/// cbindgen:ignore +#[derive(RefCastCustom)] +#[repr(transparent)] +pub struct VoicevoxOnnxruntime(voicevox_core::blocking::Onnxruntime); + +/// ONNX Runtimeを初期化済みであれば、 ::VoicevoxOnnxruntime のインスタンスを得る。 +/// +/// 初期化していなければ`NULL`を返す。 +/// +/// @returns ::VoicevoxOnnxruntime のインスタンス +#[no_mangle] +pub extern "C" fn voicevox_onnxruntime_get() -> Option<&'static VoicevoxOnnxruntime> { + VoicevoxOnnxruntime::get() +} + +/// ONNX Runtimeをロードして初期化する。 +/// +/// 二度目以降はパラメータに関わらず既存のインスタンスを返す。 +/// +/// @param [in] options オプション +/// @param [out] out_onnxruntime ::VoicevoxOnnxruntime のインスタンス +/// +/// @returns 結果コード +/// +/// \safety{ +/// - `options.filename`はヌル終端文字列を指し、かつ読み込みについて有効でなければならない。 +/// - `out_onnxruntime`は書き込みについて有効でなければならない。 +/// } +#[cfg(feature = "onnxruntime-libloading")] +#[no_mangle] +pub unsafe extern "C" fn voicevox_onnxruntime_load_once( + options: VoicevoxLoadOnnxruntimeOptions, + out_onnxruntime: NonNull<&'static VoicevoxOnnxruntime>, +) -> VoicevoxResultCode { + init_logger_once(); + let filename = unsafe { + // SAFETY: ユーザーに要求している条件で十分 + CStr::from_ptr(options.filename) + }; + into_result_code_with_error((|| { + let instance = VoicevoxOnnxruntime::load_once(filename)?; + unsafe { + // SAFETY: ユーザーに要求している条件で十分 + out_onnxruntime.as_ptr().write_unaligned(instance); + } + Ok(()) + })()) +} + +/// ONNX Runtimeを初期化する。 +/// +/// 二度目以降は既存のインスタンスを返す。 +/// +/// @param [out] out_onnxruntime ::VoicevoxOnnxruntime のインスタンス +/// +/// @returns 結果コード +/// +/// \safety{ +/// - `out_onnxruntime`は書き込みについて有効でなければならない。 +/// } +#[cfg(feature = "onnxruntime-link-dylib")] +#[no_mangle] +pub unsafe extern "C" fn voicevox_onnxruntime_init_once( + out_onnxruntime: NonNull<&'static VoicevoxOnnxruntime>, +) -> VoicevoxResultCode { + init_logger_once(); + into_result_code_with_error((|| { + let instance = VoicevoxOnnxruntime::init_once()?; + unsafe { + // SAFETY: ユーザーに要求している条件で十分 + out_onnxruntime.as_ptr().write_unaligned(instance); + } + Ok(()) + })()) +} + /// テキスト解析器としてのOpen JTalk。 /// /// 構築(_construction_)は ::voicevox_open_jtalk_rc_new で行い、破棄(_destruction_)は ::voicevox_open_jtalk_rc_delete で行う。 @@ -226,6 +337,7 @@ pub extern "C" fn voicevox_get_version() -> *const c_char { init_logger_once(); return C_STRING_DROP_CHECKER.blacklist(VERSION).as_ptr(); + // FIXME: 実行時チェックにすることでこの`unsafe`は削れるはず const VERSION: &CStr = unsafe { // SAFETY: The package version is a SemVer, so it should not contain '\0' CStr::from_bytes_with_nul_unchecked(concat!(env!("CARGO_PKG_VERSION"), '\0').as_bytes()) @@ -330,6 +442,7 @@ pub struct VoicevoxSynthesizer { /// ::VoicevoxSynthesizer を構築(_construct_)する。 /// +/// @param [in] onnxruntime /// @param [in] open_jtalk Open JTalkのオブジェクト /// @param [in] options オプション /// @param [out] out_synthesizer 構築先 @@ -337,11 +450,13 @@ pub struct VoicevoxSynthesizer { /// @returns 結果コード /// /// \safety{ +/// - `onnxruntime`は ::voicevox_onnxruntime_load_once または ::voicevox_onnxruntime_init_once で得たものでなければならない。 /// - `open_jtalk`は ::voicevox_voice_model_new_from_path で得たものでなければならず、また ::voicevox_open_jtalk_rc_new で解放されていてはいけない。 /// - `out_synthesizer`は書き込みについて有効でなければならない。 /// } #[no_mangle] pub unsafe extern "C" fn voicevox_synthesizer_new( + onnxruntime: &'static VoicevoxOnnxruntime, open_jtalk: &OpenJtalkRc, options: VoicevoxInitializeOptions, out_synthesizer: NonNull>, @@ -350,7 +465,7 @@ pub unsafe extern "C" fn voicevox_synthesizer_new( into_result_code_with_error((|| { let options = options.into(); - let synthesizer = VoicevoxSynthesizer::new(open_jtalk, &options)?.into(); + let synthesizer = VoicevoxSynthesizer::new(onnxruntime, open_jtalk, &options)?.into(); out_synthesizer.as_ptr().write_unaligned(synthesizer); Ok(()) })()) @@ -411,6 +526,22 @@ pub extern "C" fn voicevox_synthesizer_unload_voice_model( into_result_code_with_error(synthesizer.unload_voice_model(model_id).map_err(Into::into)) } +/// ::VoicevoxOnnxruntime のインスタンスを得る。 +/// +/// @param [in] synthesizer 音声シンセサイザ +/// +/// @returns ::VoicevoxOnnxruntime のインスタンス +/// +/// \safety{ +/// - `synthesizer`は ::voicevox_synthesizer_new で得たものでなければならず、また ::voicevox_synthesizer_delete で解放されていてはいけない。 +/// } +#[no_mangle] +pub extern "C" fn voicevox_synthesizer_get_onnxruntime( + synthesizer: &VoicevoxSynthesizer, +) -> &'static VoicevoxOnnxruntime { + synthesizer.onnxruntime() +} + /// ハードウェアアクセラレーションがGPUモードか判定する。 /// /// @param [in] synthesizer 音声シンセサイザ @@ -473,6 +604,7 @@ pub extern "C" fn voicevox_synthesizer_create_metas_json( /// /// あくまで本ライブラリが対応しているデバイスの情報であることに注意。GPUが使える環境ではなかったとしても`cuda`や`dml`は`true`を示しうる。 /// +/// @param [in] onnxruntime /// @param [out] output_supported_devices_json サポートデバイス情報のJSON文字列 /// /// @returns 結果コード @@ -480,21 +612,23 @@ pub extern "C" fn voicevox_synthesizer_create_metas_json( /// \example{ /// ```c /// char *supported_devices; -/// VoicevoxResultCode result = voicevox_create_supported_devices_json(&supported_devices); +/// VoicevoxResultCode result = voicevox_onnxruntime_create_supported_devices_json(onnxruntime, &supported_devices); /// ``` /// } /// /// \safety{ +/// - `onnxruntime`は ::voicevox_onnxruntime_load_once または ::voicevox_onnxruntime_init_once で得たものでなければならない。 /// - `output_supported_devices_json`は書き込みについて有効でなければならない。 /// } #[no_mangle] -pub unsafe extern "C" fn voicevox_create_supported_devices_json( +pub unsafe extern "C" fn voicevox_onnxruntime_create_supported_devices_json( + onnxruntime: &'static VoicevoxOnnxruntime, output_supported_devices_json: NonNull<*mut c_char>, ) -> VoicevoxResultCode { init_logger_once(); into_result_code_with_error((|| { let supported_devices = - CString::new(SupportedDevices::create()?.to_json().to_string()).unwrap(); + CString::new(onnxruntime.0.supported_devices()?.to_json().to_string()).unwrap(); output_supported_devices_json.as_ptr().write_unaligned( C_STRING_DROP_CHECKER .whitelist(supported_devices) @@ -980,7 +1114,7 @@ pub unsafe extern "C" fn voicevox_synthesizer_tts( /// /// \safety{ /// - `json`は以下のAPIで得られたポインタでなくてはいけない。 -/// - ::voicevox_create_supported_devices_json +/// - ::voicevox_onnxruntime_create_supported_devices_json /// - ::voicevox_synthesizer_create_metas_json /// - ::voicevox_synthesizer_create_audio_query /// - ::voicevox_synthesizer_create_accent_phrases diff --git a/crates/voicevox_core_c_api/src/result_code.rs b/crates/voicevox_core_c_api/src/result_code.rs index 0897dfa87..085f2b8c6 100644 --- a/crates/voicevox_core_c_api/src/result_code.rs +++ b/crates/voicevox_core_c_api/src/result_code.rs @@ -17,6 +17,8 @@ pub enum VoicevoxResultCode { VOICEVOX_RESULT_GET_SUPPORTED_DEVICES_ERROR = 3, /// GPUモードがサポートされていない VOICEVOX_RESULT_GPU_SUPPORT_ERROR = 4, + /// 推論ライブラリのロードまたは初期化ができなかった + VOICEVOX_RESULT_INIT_INFERENCE_RUNTIME_ERROR = 29, /// スタイルIDに対するスタイルが見つからなかった VOICEVOX_RESULT_STYLE_NOT_FOUND_ERROR = 6, /// 音声モデルIDに対する音声モデルが見つからなかった @@ -69,6 +71,9 @@ pub(crate) const fn error_result_to_message(result_code: VoicevoxResultCode) -> VOICEVOX_RESULT_GET_SUPPORTED_DEVICES_ERROR => { cstr!("サポートされているデバイス情報取得中にエラーが発生しました") } + VOICEVOX_RESULT_INIT_INFERENCE_RUNTIME_ERROR => { + cstr!("推論ライブラリのロードまたは初期化ができませんでした") + } VOICEVOX_RESULT_OK => cstr!("エラーが発生しませんでした"), VOICEVOX_RESULT_STYLE_NOT_FOUND_ERROR => cstr!( "指定されたIDに対するスタイルが見つかりませんでした。音声モデルが読み込まれていないか\ diff --git a/crates/voicevox_core_c_api/tests/e2e/assert_cdylib.rs b/crates/voicevox_core_c_api/tests/e2e/assert_cdylib.rs index cfbec5c31..e779fdc23 100644 --- a/crates/voicevox_core_c_api/tests/e2e/assert_cdylib.rs +++ b/crates/voicevox_core_c_api/tests/e2e/assert_cdylib.rs @@ -8,6 +8,7 @@ use assert_cmd::assert::{Assert, AssertResult, OutputAssertExt as _}; use clap::Parser as _; use duct::cmd; use easy_ext::ext; +use itertools::Itertools as _; use libloading::Library; use libtest_mimic::{Failed, Trial}; @@ -46,7 +47,15 @@ pub(crate) fn exec() -> anyhow::Result<()> { // テスト対象が無いときに`cargo build`をスキップしたいが、判定部分がプライベート。 // そのためスキップするのはCLIオプションに`--ignored`か`--include-ignored`が無いときのみ if args.ignored || args.include_ignored { - cmd!(env!("CARGO"), "build", "--release", "--lib").run()?; + cmd!( + env!("CARGO"), + "build", + "--release", + "--lib", + "--features", + &format!(",{}", C::FEATURES.iter().format(",")), + ) + .run()?; ensure!( C::cdylib_path().exists(), @@ -96,6 +105,7 @@ pub(crate) fn exec() -> anyhow::Result<()> { } pub(crate) trait TestContext { + const FEATURES: &'static [&'static str]; const TARGET_DIR: &'static str; const CDYLIB_NAME: &'static str; const RUNTIME_ENVS: &'static [(&'static str, &'static str)]; diff --git a/crates/voicevox_core_c_api/tests/e2e/log_mask.rs b/crates/voicevox_core_c_api/tests/e2e/log_mask.rs index b28442bdb..9b08c9af7 100644 --- a/crates/voicevox_core_c_api/tests/e2e/log_mask.rs +++ b/crates/voicevox_core_c_api/tests/e2e/log_mask.rs @@ -20,6 +20,13 @@ impl Utf8Output { ) } + pub(crate) fn mask_onnxruntime_version(self) -> Self { + self.mask_stderr( + static_regex!(regex::escape(ort::downloaded_version!())), + "{onnxruntime_version}", + ) + } + pub(crate) fn mask_windows_video_cards(self) -> Self { self.mask_stderr( static_regex!( diff --git a/crates/voicevox_core_c_api/tests/e2e/main.rs b/crates/voicevox_core_c_api/tests/e2e/main.rs index 43dc3a95e..a826c02e9 100644 --- a/crates/voicevox_core_c_api/tests/e2e/main.rs +++ b/crates/voicevox_core_c_api/tests/e2e/main.rs @@ -22,6 +22,7 @@ fn main() -> anyhow::Result<()> { enum TestContext {} impl assert_cdylib::TestContext for TestContext { + const FEATURES: &'static [&'static str] = &["onnxruntime-libloading"]; const TARGET_DIR: &'static str = "../../target"; const CDYLIB_NAME: &'static str = "voicevox_core"; const RUNTIME_ENVS: &'static [(&'static str, &'static str)] = diff --git a/crates/voicevox_core_c_api/tests/e2e/snapshots.toml b/crates/voicevox_core_c_api/tests/e2e/snapshots.toml index 151074cb3..b623119dd 100644 --- a/crates/voicevox_core_c_api/tests/e2e/snapshots.toml +++ b/crates/voicevox_core_c_api/tests/e2e/snapshots.toml @@ -51,9 +51,12 @@ metas = ''' } ]''' stderr.windows = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' {windows-video-cards} ''' -stderr.unix = "" +stderr.unix = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' +''' [compatible_engine_load_model_before_initialize] last_error_message = "Statusが初期化されていません" @@ -83,14 +86,21 @@ result_messages.22 = "ユーザー辞書に単語が見つかりませんでし result_messages.23 = "OpenJTalkのユーザー辞書の設定に失敗しました" result_messages.24 = "ユーザー辞書の単語のバリデーションに失敗しました" result_messages.25 = "UUIDの変換に失敗しました" -stderr = "" +# FIXME: 26, 27, 28が抜けている +result_messages.29 = "推論ライブラリのロードまたは初期化ができませんでした" +stderr = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' +''' [simple_tts] output."こんにちは、音声合成の世界へようこそ".wav_length = 176172 stderr.windows = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' {windows-video-cards} ''' -stderr.unix = "" +stderr.unix = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' +''' [synthesizer_new_output_json] metas = ''' @@ -145,22 +155,32 @@ metas = ''' } ]''' stderr.windows = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' {windows-video-cards} ''' -stderr.unix = "" +stderr.unix = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' +''' [tts_via_audio_query] output."こんにちは、音声合成の世界へようこそ".wav_length = 176172 stderr.windows = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' {windows-video-cards} ''' -stderr.unix = "" +stderr.unix = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' +''' +# FIXME: "user_dict_load"のはず [user_dict] stderr.windows = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' {windows-video-cards} ''' -stderr.unix = "" +stderr.unix = ''' +{timestamp} INFO ort: Loaded ONNX Runtime dylib with version '{onnxruntime_version}' +''' [user_dict_manipulate] stderr = "" diff --git a/crates/voicevox_core_c_api/tests/e2e/testcases/compatible_engine.rs b/crates/voicevox_core_c_api/tests/e2e/testcases/compatible_engine.rs index c2fc211c2..79b1d5dc8 100644 --- a/crates/voicevox_core_c_api/tests/e2e/testcases/compatible_engine.rs +++ b/crates/voicevox_core_c_api/tests/e2e/testcases/compatible_engine.rs @@ -31,12 +31,10 @@ impl assert_cdylib::TestCase for TestCase { serde_json::to_string_pretty(&metas_json.parse::()?).unwrap() }; - let supported_devices = { + { let supported_devices = lib.supported_devices(); - CStr::from_ptr(supported_devices) - .to_str()? - .parse::()? - }; + serde_json::from_str::(CStr::from_ptr(supported_devices).to_str()?)?; + } assert!(lib.initialize(false, 0, false)); @@ -86,10 +84,6 @@ impl assert_cdylib::TestCase for TestCase { }; std::assert_eq!(SNAPSHOTS.metas, metas_json); - std::assert_eq!( - SupportedDevices::create().unwrap().to_json(), - supported_devices, - ); float_assert::close_l1(&phoneme_length, &EXAMPLE_DATA.duration.result, 0.01); float_assert::close_l1(&intonation_list, &EXAMPLE_DATA.intonation.result, 0.01); @@ -103,6 +97,7 @@ impl assert_cdylib::TestCase for TestCase { fn assert_output(&self, output: Utf8Output) -> AssertResult { output .mask_timestamps() + .mask_onnxruntime_version() .mask_windows_video_cards() .assert() .try_success()? diff --git a/crates/voicevox_core_c_api/tests/e2e/testcases/compatible_engine_load_model_before_initialize.rs b/crates/voicevox_core_c_api/tests/e2e/testcases/compatible_engine_load_model_before_initialize.rs index 173e32f8c..18482f0de 100644 --- a/crates/voicevox_core_c_api/tests/e2e/testcases/compatible_engine_load_model_before_initialize.rs +++ b/crates/voicevox_core_c_api/tests/e2e/testcases/compatible_engine_load_model_before_initialize.rs @@ -34,6 +34,7 @@ impl assert_cdylib::TestCase for TestCase { fn assert_output(&self, output: Utf8Output) -> AssertResult { output .mask_timestamps() + .mask_onnxruntime_version() .mask_windows_video_cards() .assert() .try_success()? diff --git a/crates/voicevox_core_c_api/tests/e2e/testcases/global_info.rs b/crates/voicevox_core_c_api/tests/e2e/testcases/global_info.rs index c6ea390ed..d2a8fc7e7 100644 --- a/crates/voicevox_core_c_api/tests/e2e/testcases/global_info.rs +++ b/crates/voicevox_core_c_api/tests/e2e/testcases/global_info.rs @@ -28,16 +28,23 @@ impl assert_cdylib::TestCase for TestCase { CStr::from_ptr(lib.voicevox_get_version()).to_str()?, ); + let onnxruntime = { + let mut onnxruntime = MaybeUninit::uninit(); + assert_ok(lib.voicevox_onnxruntime_load_once( + lib.voicevox_make_default_load_onnxruntime_options(), + onnxruntime.as_mut_ptr(), + )); + onnxruntime.assume_init() + }; + { let mut supported_devices = MaybeUninit::uninit(); - assert_ok(lib.voicevox_create_supported_devices_json(supported_devices.as_mut_ptr())); + assert_ok(lib.voicevox_onnxruntime_create_supported_devices_json( + onnxruntime, + supported_devices.as_mut_ptr(), + )); let supported_devices = supported_devices.assume_init(); - std::assert_eq!( - SupportedDevices::create()?.to_json(), - CStr::from_ptr(supported_devices) - .to_str()? - .parse::()?, - ); + serde_json::from_str::(CStr::from_ptr(supported_devices).to_str()?)?; lib.voicevox_json_free(supported_devices); } @@ -83,6 +90,7 @@ impl assert_cdylib::TestCase for TestCase { fn assert_output(&self, output: Utf8Output) -> AssertResult { output .mask_timestamps() + .mask_onnxruntime_version() .mask_windows_video_cards() .assert() .try_success()? diff --git a/crates/voicevox_core_c_api/tests/e2e/testcases/simple_tts.rs b/crates/voicevox_core_c_api/tests/e2e/testcases/simple_tts.rs index 4ac4030e1..4635b0271 100644 --- a/crates/voicevox_core_c_api/tests/e2e/testcases/simple_tts.rs +++ b/crates/voicevox_core_c_api/tests/e2e/testcases/simple_tts.rs @@ -37,6 +37,15 @@ impl assert_cdylib::TestCase for TestCase { model.assume_init() }; + let onnxruntime = { + let mut onnxruntime = MaybeUninit::uninit(); + assert_ok(lib.voicevox_onnxruntime_load_once( + lib.voicevox_make_default_load_onnxruntime_options(), + onnxruntime.as_mut_ptr(), + )); + onnxruntime.assume_init() + }; + let openjtalk = { let mut openjtalk = MaybeUninit::uninit(); let open_jtalk_dic_dir = CString::new(OPEN_JTALK_DIC_DIR).unwrap(); @@ -49,6 +58,7 @@ impl assert_cdylib::TestCase for TestCase { let synthesizer = { let mut synthesizer = MaybeUninit::uninit(); assert_ok(lib.voicevox_synthesizer_new( + onnxruntime, openjtalk, VoicevoxInitializeOptions { acceleration_mode: @@ -96,6 +106,7 @@ impl assert_cdylib::TestCase for TestCase { fn assert_output(&self, output: Utf8Output) -> AssertResult { output .mask_timestamps() + .mask_onnxruntime_version() .mask_windows_video_cards() .assert() .try_success()? diff --git a/crates/voicevox_core_c_api/tests/e2e/testcases/synthesizer_new_output_json.rs b/crates/voicevox_core_c_api/tests/e2e/testcases/synthesizer_new_output_json.rs index c27bd4703..8a2fd1951 100644 --- a/crates/voicevox_core_c_api/tests/e2e/testcases/synthesizer_new_output_json.rs +++ b/crates/voicevox_core_c_api/tests/e2e/testcases/synthesizer_new_output_json.rs @@ -28,6 +28,15 @@ impl assert_cdylib::TestCase for TestCase { unsafe fn exec(&self, lib: Library) -> anyhow::Result<()> { let lib = CApi::from_library(lib)?; + let onnxruntime = { + let mut onnxruntime = MaybeUninit::uninit(); + assert_ok(lib.voicevox_onnxruntime_load_once( + lib.voicevox_make_default_load_onnxruntime_options(), + onnxruntime.as_mut_ptr(), + )); + onnxruntime.assume_init() + }; + let openjtalk = { let mut openjtalk = MaybeUninit::uninit(); let open_jtalk_dic_dir = CString::new(OPEN_JTALK_DIC_DIR).unwrap(); @@ -40,6 +49,7 @@ impl assert_cdylib::TestCase for TestCase { let synthesizer = { let mut synthesizer = MaybeUninit::uninit(); assert_ok(lib.voicevox_synthesizer_new( + onnxruntime, openjtalk, VoicevoxInitializeOptions { acceleration_mode: @@ -85,6 +95,7 @@ impl assert_cdylib::TestCase for TestCase { fn assert_output(&self, output: Utf8Output) -> AssertResult { output .mask_timestamps() + .mask_onnxruntime_version() .mask_windows_video_cards() .assert() .try_success()? diff --git a/crates/voicevox_core_c_api/tests/e2e/testcases/tts_via_audio_query.rs b/crates/voicevox_core_c_api/tests/e2e/testcases/tts_via_audio_query.rs index d380b71b2..728f21339 100644 --- a/crates/voicevox_core_c_api/tests/e2e/testcases/tts_via_audio_query.rs +++ b/crates/voicevox_core_c_api/tests/e2e/testcases/tts_via_audio_query.rs @@ -37,6 +37,15 @@ impl assert_cdylib::TestCase for TestCase { model.assume_init() }; + let onnxruntime = { + let mut onnxruntime = MaybeUninit::uninit(); + assert_ok(lib.voicevox_onnxruntime_load_once( + lib.voicevox_make_default_load_onnxruntime_options(), + onnxruntime.as_mut_ptr(), + )); + onnxruntime.assume_init() + }; + let openjtalk = { let mut openjtalk = MaybeUninit::uninit(); let open_jtalk_dic_dir = CString::new(OPEN_JTALK_DIC_DIR).unwrap(); @@ -49,6 +58,7 @@ impl assert_cdylib::TestCase for TestCase { let synthesizer = { let mut synthesizer = MaybeUninit::uninit(); assert_ok(lib.voicevox_synthesizer_new( + onnxruntime, openjtalk, VoicevoxInitializeOptions { acceleration_mode: @@ -108,6 +118,7 @@ impl assert_cdylib::TestCase for TestCase { fn assert_output(&self, output: Utf8Output) -> AssertResult { output .mask_timestamps() + .mask_onnxruntime_version() .mask_windows_video_cards() .assert() .try_success()? diff --git a/crates/voicevox_core_c_api/tests/e2e/testcases/user_dict_load.rs b/crates/voicevox_core_c_api/tests/e2e/testcases/user_dict_load.rs index 2e6875e97..646c5647b 100644 --- a/crates/voicevox_core_c_api/tests/e2e/testcases/user_dict_load.rs +++ b/crates/voicevox_core_c_api/tests/e2e/testcases/user_dict_load.rs @@ -54,6 +54,15 @@ impl assert_cdylib::TestCase for TestCase { model.assume_init() }; + let onnxruntime = { + let mut onnxruntime = MaybeUninit::uninit(); + assert_ok(lib.voicevox_onnxruntime_load_once( + lib.voicevox_make_default_load_onnxruntime_options(), + onnxruntime.as_mut_ptr(), + )); + onnxruntime.assume_init() + }; + let openjtalk = { let mut openjtalk = MaybeUninit::uninit(); let open_jtalk_dic_dir = CString::new(OPEN_JTALK_DIC_DIR).unwrap(); @@ -66,6 +75,7 @@ impl assert_cdylib::TestCase for TestCase { let synthesizer = { let mut synthesizer = MaybeUninit::uninit(); assert_ok(lib.voicevox_synthesizer_new( + onnxruntime, openjtalk, VoicevoxInitializeOptions { acceleration_mode: @@ -125,6 +135,7 @@ impl assert_cdylib::TestCase for TestCase { fn assert_output(&self, output: Utf8Output) -> AssertResult { output .mask_timestamps() + .mask_onnxruntime_version() .mask_windows_video_cards() .assert() .try_success()? diff --git a/crates/voicevox_core_c_api/tests/e2e/testcases/user_dict_manipulate.rs b/crates/voicevox_core_c_api/tests/e2e/testcases/user_dict_manipulate.rs index fd3d575e3..bec0d61b5 100644 --- a/crates/voicevox_core_c_api/tests/e2e/testcases/user_dict_manipulate.rs +++ b/crates/voicevox_core_c_api/tests/e2e/testcases/user_dict_manipulate.rs @@ -140,6 +140,7 @@ impl assert_cdylib::TestCase for TestCase { fn assert_output(&self, output: Utf8Output) -> AssertResult { output .mask_timestamps() + .mask_onnxruntime_version() .mask_windows_video_cards() .assert() .try_success()? diff --git a/crates/voicevox_core_java_api/Cargo.toml b/crates/voicevox_core_java_api/Cargo.toml index b39a98b2a..fd785e17b 100644 --- a/crates/voicevox_core_java_api/Cargo.toml +++ b/crates/voicevox_core_java_api/Cargo.toml @@ -15,6 +15,7 @@ directml = ["voicevox_core/directml"] android_logger.workspace = true chrono = { workspace = true, default-features = false, features = ["clock"] } derive_more.workspace = true +duplicate.workspace = true easy-ext.workspace = true jni.workspace = true once_cell.workspace = true @@ -24,7 +25,7 @@ serde_json = { workspace = true, features = ["preserve_order"] } tracing = { workspace = true, features = ["log"] } tracing-subscriber = { workspace = true, features = ["env-filter"] } uuid.workspace = true -voicevox_core.workspace = true +voicevox_core = { workspace = true, features = ["onnxruntime-libloading"] } [lints.rust] unsafe_code = "allow" # jni-rsが要求 diff --git a/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/GlobalInfo.java b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/GlobalInfo.java index 26f9ccddd..010e69073 100644 --- a/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/GlobalInfo.java +++ b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/GlobalInfo.java @@ -1,6 +1,5 @@ package jp.hiroshiba.voicevoxcore; -import com.google.gson.Gson; import com.google.gson.annotations.Expose; import com.google.gson.annotations.SerializedName; import jakarta.annotation.Nonnull; @@ -17,28 +16,13 @@ public static String getVersion() { return rsGetVersion(); } - /** - * このライブラリで利用可能なデバイスの情報を取得する。 - * - * @return {@link SupportedDevices}。 - */ - @Nonnull - public static SupportedDevices getSupportedDevices() { - Gson gson = new Gson(); - String supportedDevicesJson = rsGetSupportedDevicesJson(); - SupportedDevices supportedDevices = gson.fromJson(supportedDevicesJson, SupportedDevices.class); - if (supportedDevices == null) { - throw new NullPointerException("supported_devices"); - } - return supportedDevices; - } - @Nonnull private static native String rsGetVersion(); @Nonnull private static native String rsGetSupportedDevicesJson(); + // FIXME: `Onnxruntime`に移すか、独立させる /** * このライブラリで利用可能なデバイスの情報。 * diff --git a/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Onnxruntime.java b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Onnxruntime.java new file mode 100644 index 000000000..d0c45e6a4 --- /dev/null +++ b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Onnxruntime.java @@ -0,0 +1,131 @@ +package jp.hiroshiba.voicevoxcore; + +import static jp.hiroshiba.voicevoxcore.GlobalInfo.SupportedDevices; + +import com.google.gson.Gson; +import jakarta.annotation.Nonnull; +import jakarta.annotation.Nullable; +import java.util.Optional; + +/** + * ONNX Runtime。 + * + *

シングルトンであり、インスタンスは高々一つ。 + * + * 

+ * Onnxruntime ort1 = Onnxruntime.loadOnce().exec();
+ * Onnxruntime ort2 = Onnxruntime.get().get();
+ * assert ort1 == ort2;
+ *
+ */ +public class Onnxruntime extends Dll { + /** ONNX Runtimeのライブラリ名。 */ + public static final String LIB_NAME = "onnxruntime"; + + /** 推奨されるONNX Runtimeのバージョン。 */ + public static final String LIB_VERSION = "1.17.3"; + + /** + * {@link LIB_NAME}と{@link LIB_VERSION}からなる動的ライブラリのファイル名。 + * + *

WindowsとAndroidでは{@link LIB_UNVERSIONED_FILENAME}と同じ。 + */ + public static final String LIB_VERSIONED_FILENAME = rsLibVersionedFilename(); + + /** {@link LIB_NAME}からなる動的ライブラリのファイル名。 */ + public static final String LIB_UNVERSIONED_FILENAME = rsLibUnversionedFilename(); + + @Nullable private static Onnxruntime instance = null; + + /** + * インスタンスが既に作られているならそれを得る。 + * + * @return インスタンスがあるなら{@code Optional.of(…)}、そうでなければ{@code Optional.empty()}。 + */ + public static Optional get() { + synchronized (Onnxruntime.class) { + return Optional.ofNullable(instance); + } + } + + /** + * ONNX Runtimeをロードして初期化する。 + * + *

一度成功したら、以後は引数を無視して同じインスタンスを返す。 + * + * @return {@link LoadOnce}。 + */ + public static LoadOnce loadOnce() { + return new LoadOnce(); + } + + private static native String rsLibName(); + + private static native String rsLibVersion(); + + private static native String rsLibVersionedFilename(); + + private static native String rsLibUnversionedFilename(); + + static { + assert LIB_NAME.equals(rsLibName()) && LIB_VERSION.equals(rsLibVersion()); + } + + /** {@link #loadOnce}のビルダー。 */ + public static class LoadOnce { + /** + * ONNX Runtimeのfilenameを指定する。 + * + * @param filename {@code dlopen}/{@code LoadLibraryExW}の引数に使われるfilename。デフォルトは{@link + * LIB_VERSIONED_FILENAME}。 + * @return このオブジェクト。 + */ + public LoadOnce filename(@Nonnull String filename) { + this.filename = filename; + return this; + } + + /** + * 実行する。 + * + * @return {@link Onnxruntime}。 + */ + public Onnxruntime exec() { + synchronized (Onnxruntime.class) { + if (instance == null) { + instance = new Onnxruntime(filename); + } + } + return instance; + } + + private LoadOnce() {} + + @Nonnull private String filename = LIB_VERSIONED_FILENAME; + } + + private long handle; + + private Onnxruntime(@Nullable String filename) { + rsNew(filename); + } + + /** + * このライブラリで利用可能なデバイスの情報を取得する。 + * + * @return {@link SupportedDevices}。 + */ + public SupportedDevices supportedDevices() { + Gson gson = new Gson(); + String supportedDevicesJson = rsSupportedDevices(); + SupportedDevices supportedDevices = gson.fromJson(supportedDevicesJson, SupportedDevices.class); + if (supportedDevices == null) { + throw new NullPointerException("supported_devices"); + } + return supportedDevices; + } + + private native void rsNew(@Nullable String filename); + + private native String rsSupportedDevices(); +} diff --git a/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Synthesizer.java b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Synthesizer.java index 6ec6d9108..8d5694267 100644 --- a/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Synthesizer.java +++ b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Synthesizer.java @@ -5,6 +5,7 @@ import java.util.ArrayList; import java.util.Arrays; import java.util.List; +import java.util.Optional; import java.util.UUID; import jp.hiroshiba.voicevoxcore.exceptions.InferenceFailedException; import jp.hiroshiba.voicevoxcore.exceptions.InvalidModelDataException; @@ -17,8 +18,8 @@ public class Synthesizer extends Dll { private long handle; - private Synthesizer(OpenJtalk openJtalk, Builder builder) { - rsNew(openJtalk, builder); + private Synthesizer(Onnxruntime onnxruntime, OpenJtalk openJtalk, Builder builder) { + rsNew(onnxruntime, openJtalk, builder); } protected void finalize() throws Throwable { @@ -26,6 +27,18 @@ protected void finalize() throws Throwable { super.finalize(); } + /** + * ONNX Runtime。 + * + * @return {@link Onnxruntime}。 + */ + @Nonnull + public Onnxruntime getOnnxruntime() { + Optional onnxruntime = Onnxruntime.get(); + assert onnxruntime.isPresent() : "`Synthesizer`のコンストラクタで要求しているはず"; + return onnxruntime.get(); + } + /** * ハードウェアアクセラレーションがGPUモードかどうかを返す。 * @@ -266,7 +279,7 @@ public TtsConfigurator tts(String text, int styleId) { return new TtsConfigurator(this, text, styleId); } - private native void rsNew(OpenJtalk openJtalk, Builder builder); + private native void rsNew(Onnxruntime onnxruntime, OpenJtalk openJtalk, Builder builder); private native boolean rsIsGpuMode(); @@ -320,8 +333,8 @@ private native byte[] rsTts(String text, int styleId, boolean enableInterrogativ private native void rsDrop(); - public static Builder builder(OpenJtalk openJtalk) { - return new Builder(openJtalk); + public static Builder builder(Onnxruntime onnxruntime, OpenJtalk openJtalk) { + return new Builder(onnxruntime, openJtalk); } /** @@ -330,6 +343,7 @@ public static Builder builder(OpenJtalk openJtalk) { * @see Synthesizer#builder */ public static class Builder { + private Onnxruntime onnxruntime; private OpenJtalk openJtalk; @SuppressWarnings("unused") @@ -338,7 +352,8 @@ public static class Builder { @SuppressWarnings("unused") private int cpuNumThreads; - public Builder(OpenJtalk openJtalk) { + public Builder(Onnxruntime onnxruntime, OpenJtalk openJtalk) { + this.onnxruntime = onnxruntime; this.openJtalk = openJtalk; } @@ -373,7 +388,7 @@ public Builder cpuNumThreads(int cpuNumThreads) { * @return {@link Synthesizer}。 */ public Synthesizer build() { - Synthesizer synthesizer = new Synthesizer(openJtalk, this); + Synthesizer synthesizer = new Synthesizer(onnxruntime, openJtalk, this); return synthesizer; } } diff --git a/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/exceptions/InitInferenceRuntimeException.java b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/exceptions/InitInferenceRuntimeException.java new file mode 100644 index 000000000..c981ea034 --- /dev/null +++ b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/exceptions/InitInferenceRuntimeException.java @@ -0,0 +1,14 @@ +package jp.hiroshiba.voicevoxcore.exceptions; + +import java.io.IOException; + +/** 推論ライブラリのロードまたは初期化ができなかった。 */ +public class InitInferenceRuntimeException extends IOException { + public InitInferenceRuntimeException(String message) { + super(message); + } + + public InitInferenceRuntimeException(String message, Throwable cause) { + super(message, cause); + } +} diff --git a/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/InfoTest.java b/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/InfoTest.java index 52915abad..c9e71ed1c 100644 --- a/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/InfoTest.java +++ b/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/InfoTest.java @@ -8,15 +8,16 @@ import org.junit.jupiter.api.Test; -class InfoTest { +class InfoTest extends TestUtils { @Test void checkVersion() { assertNotNull(GlobalInfo.getVersion()); } + // TODO: 別の場所に移す @Test void checkSupportedDevices() { - GlobalInfo.SupportedDevices supportedDevices = GlobalInfo.getSupportedDevices(); + GlobalInfo.SupportedDevices supportedDevices = loadOnnxruntime().supportedDevices(); assertNotNull(supportedDevices); assertTrue(supportedDevices.cpu); diff --git a/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/SynthesizerTest.java b/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/SynthesizerTest.java index 1eb8fe057..4de2f617b 100644 --- a/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/SynthesizerTest.java +++ b/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/SynthesizerTest.java @@ -20,9 +20,12 @@ interface MoraCheckCallback { @Test void checkIsGpuMode() { + Onnxruntime onnxruntime = loadOnnxruntime(); OpenJtalk openJtalk = loadOpenJtalk(); Synthesizer synthesizer = - Synthesizer.builder(openJtalk).accelerationMode(Synthesizer.AccelerationMode.CPU).build(); + Synthesizer.builder(onnxruntime, openJtalk) + .accelerationMode(Synthesizer.AccelerationMode.CPU) + .build(); assertFalse(synthesizer.isGpuMode()); } @@ -45,9 +48,10 @@ boolean checkAllMoras( @Test void checkModel() throws InvalidModelDataException { + Onnxruntime onnxruntime = loadOnnxruntime(); VoiceModel model = loadModel(); OpenJtalk openJtalk = loadOpenJtalk(); - Synthesizer synthesizer = Synthesizer.builder(openJtalk).build(); + Synthesizer synthesizer = Synthesizer.builder(onnxruntime, openJtalk).build(); assertTrue(synthesizer.metas().length == 0); @@ -65,8 +69,9 @@ void checkModel() throws InvalidModelDataException { @Test void checkAudioQuery() throws InferenceFailedException, InvalidModelDataException { VoiceModel model = loadModel(); + Onnxruntime onnxruntime = loadOnnxruntime(); OpenJtalk openJtalk = loadOpenJtalk(); - Synthesizer synthesizer = Synthesizer.builder(openJtalk).build(); + Synthesizer synthesizer = Synthesizer.builder(onnxruntime, openJtalk).build(); synthesizer.loadVoiceModel(model); AudioQuery query = synthesizer.createAudioQuery("こんにちは", model.metas[0].styles[0].id); @@ -77,7 +82,8 @@ void checkAudioQuery() throws InferenceFailedException, InvalidModelDataExceptio void checkAccentPhrases() throws InferenceFailedException, InvalidModelDataException { VoiceModel model = loadModel(); OpenJtalk openJtalk = loadOpenJtalk(); - Synthesizer synthesizer = Synthesizer.builder(openJtalk).build(); + Onnxruntime onnxruntime = loadOnnxruntime(); + Synthesizer synthesizer = Synthesizer.builder(onnxruntime, openJtalk).build(); synthesizer.loadVoiceModel(model); List accentPhrases = synthesizer.createAccentPhrases("こんにちは", model.metas[0].styles[0].id); @@ -106,8 +112,9 @@ void checkAccentPhrases() throws InferenceFailedException, InvalidModelDataExcep @Test void checkTts() throws InferenceFailedException, InvalidModelDataException { VoiceModel model = loadModel(); + Onnxruntime onnxruntime = loadOnnxruntime(); OpenJtalk openJtalk = loadOpenJtalk(); - Synthesizer synthesizer = Synthesizer.builder(openJtalk).build(); + Synthesizer synthesizer = Synthesizer.builder(onnxruntime, openJtalk).build(); synthesizer.loadVoiceModel(model); synthesizer.tts("こんにちは", model.metas[0].styles[0].id); } diff --git a/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/TestUtils.java b/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/TestUtils.java index 032c38a3d..9ab731cd9 100644 --- a/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/TestUtils.java +++ b/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/TestUtils.java @@ -15,6 +15,16 @@ VoiceModel loadModel() { } } + Onnxruntime loadOnnxruntime() { + final String FILENAME = "../../test_util/data/lib/" + Onnxruntime.LIB_VERSIONED_FILENAME; + + try { + return Onnxruntime.loadOnce().filename(FILENAME).exec(); + } catch (Exception e) { + throw new RuntimeException(e); + } + } + OpenJtalk loadOpenJtalk() { String cwd = System.getProperty("user.dir"); File path = new File(cwd + "/../../test_util/data/open_jtalk_dic_utf_8-1.11"); diff --git a/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/UserDictTest.java b/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/UserDictTest.java index ce9b7631a..5300b81f3 100644 --- a/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/UserDictTest.java +++ b/crates/voicevox_core_java_api/lib/src/test/java/jp/hiroshiba/voicevoxcore/UserDictTest.java @@ -17,8 +17,9 @@ class UserDictTest extends TestUtils { void checkLoad() throws InferenceFailedException, InvalidModelDataException, LoadUserDictException { VoiceModel model = loadModel(); + Onnxruntime onnxruntime = loadOnnxruntime(); OpenJtalk openJtalk = loadOpenJtalk(); - Synthesizer synthesizer = Synthesizer.builder(openJtalk).build(); + Synthesizer synthesizer = Synthesizer.builder(onnxruntime, openJtalk).build(); UserDict userDict = new UserDict(); synthesizer.loadVoiceModel(model); AudioQuery query1 = diff --git a/crates/voicevox_core_java_api/src/common.rs b/crates/voicevox_core_java_api/src/common.rs index 6e13cee89..b38fdf648 100644 --- a/crates/voicevox_core_java_api/src/common.rs +++ b/crates/voicevox_core_java_api/src/common.rs @@ -1,4 +1,4 @@ -use std::{error::Error as _, iter}; +use std::{error::Error as _, iter, str::Utf8Error}; use derive_more::From; use easy_ext::ext; @@ -72,6 +72,7 @@ where let class = class!( NotLoadedOpenjtalkDict, GpuSupport, + InitInferenceRuntime, OpenZipFile, ReadZipEntry, InvalidModelFormat, @@ -148,6 +149,9 @@ where env.throw_new("java/lang/IllegalArgumentException", error.to_string()) ) } + JavaApiError::Utf8(error) => or_panic!( + env.throw_new("java/lang/IllegalArgumentException", error.to_string()) + ), JavaApiError::DeJson(error) => { or_panic!( env.throw_new("java/lang/IllegalArgumentException", error.to_string()) @@ -171,6 +175,9 @@ pub(crate) enum JavaApiError { #[from] Uuid(uuid::Error), + #[from] + Utf8(Utf8Error), + DeJson(serde_json::Error), } diff --git a/crates/voicevox_core_java_api/src/info.rs b/crates/voicevox_core_java_api/src/info.rs index 71b8db228..7328d4de8 100644 --- a/crates/voicevox_core_java_api/src/info.rs +++ b/crates/voicevox_core_java_api/src/info.rs @@ -9,14 +9,3 @@ extern "system" fn Java_jp_hiroshiba_voicevoxcore_GlobalInfo_rsGetVersion( Ok(version.into_raw()) }) } -#[no_mangle] -extern "system" fn Java_jp_hiroshiba_voicevoxcore_GlobalInfo_rsGetSupportedDevicesJson( - env: JNIEnv<'_>, -) -> jobject { - throw_if_err(env, std::ptr::null_mut(), |env| { - let supported_devices = voicevox_core::SupportedDevices::create()?; - let json = serde_json::to_string(&supported_devices).expect("Should not fail"); - let json = env.new_string(json)?; - Ok(json.into_raw()) - }) -} diff --git a/crates/voicevox_core_java_api/src/lib.rs b/crates/voicevox_core_java_api/src/lib.rs index 9615f0a94..4d61414e8 100644 --- a/crates/voicevox_core_java_api/src/lib.rs +++ b/crates/voicevox_core_java_api/src/lib.rs @@ -1,6 +1,7 @@ mod common; mod info; mod logger; +mod onnxruntime; mod open_jtalk; mod synthesizer; mod user_dict; diff --git a/crates/voicevox_core_java_api/src/onnxruntime.rs b/crates/voicevox_core_java_api/src/onnxruntime.rs new file mode 100644 index 000000000..0af7dfb60 --- /dev/null +++ b/crates/voicevox_core_java_api/src/onnxruntime.rs @@ -0,0 +1,56 @@ +use std::ptr; + +use duplicate::duplicate_item; +use jni::{ + objects::{JObject, JString}, + sys::jobject, + JNIEnv, +}; + +use crate::common::throw_if_err; + +#[duplicate_item( + f CONST; + [ Java_jp_hiroshiba_voicevoxcore_Onnxruntime_rsLibName ] [ LIB_NAME ]; + [ Java_jp_hiroshiba_voicevoxcore_Onnxruntime_rsLibVersion ] [ LIB_VERSION ]; + [ Java_jp_hiroshiba_voicevoxcore_Onnxruntime_rsLibVersionedFilename ] [ LIB_VERSIONED_FILENAME ]; + [ Java_jp_hiroshiba_voicevoxcore_Onnxruntime_rsLibUnversionedFilename ] [ LIB_UNVERSIONED_FILENAME ]; +)] +#[no_mangle] +extern "system" fn f(env: JNIEnv<'_>) -> jobject { + throw_if_err(env, ptr::null_mut(), |env| { + let s = env.new_string(voicevox_core::blocking::Onnxruntime::CONST)?; + Ok(s.into_raw()) + }) +} + +#[no_mangle] +unsafe extern "system" fn Java_jp_hiroshiba_voicevoxcore_Onnxruntime_rsNew<'local>( + env: JNIEnv<'local>, + this: JObject<'local>, + filename: JString<'local>, +) { + throw_if_err(env, (), |env| { + let filename = env.get_string(&filename)?.to_str()?.to_owned(); + let internal = voicevox_core::blocking::Onnxruntime::load_once() + .filename(filename) + .exec()?; + env.set_rust_field(&this, "handle", internal)?; + Ok(()) + }) +} + +#[no_mangle] +unsafe extern "system" fn Java_jp_hiroshiba_voicevoxcore_Onnxruntime_rsSupportedDevices<'local>( + env: JNIEnv<'local>, + this: JObject<'local>, +) -> jobject { + throw_if_err(env, ptr::null_mut(), |env| { + let this = *env.get_rust_field::<_, _, &'static voicevox_core::blocking::Onnxruntime>( + &this, "handle", + )?; + let json = this.supported_devices()?.to_json().to_string(); + let json = env.new_string(json)?; + Ok(json.into_raw()) + }) +} diff --git a/crates/voicevox_core_java_api/src/synthesizer.rs b/crates/voicevox_core_java_api/src/synthesizer.rs index dc5dc971d..b2605e8ae 100644 --- a/crates/voicevox_core_java_api/src/synthesizer.rs +++ b/crates/voicevox_core_java_api/src/synthesizer.rs @@ -14,6 +14,7 @@ use std::sync::Arc; unsafe extern "system" fn Java_jp_hiroshiba_voicevoxcore_Synthesizer_rsNew<'local>( env: JNIEnv<'local>, this: JObject<'local>, + onnxruntime: JObject<'local>, open_jtalk: JObject<'local>, builder: JObject<'local>, ) { @@ -45,11 +46,18 @@ unsafe extern "system" fn Java_jp_hiroshiba_voicevoxcore_Synthesizer_rsNew<'loca let cpu_num_threads = env.get_field(&builder, "cpuNumThreads", "I")?; options.cpu_num_threads = cpu_num_threads.i().expect("cpuNumThreads is not integer") as u16; + let onnxruntime = *env + .get_rust_field::<_, _, &'static voicevox_core::blocking::Onnxruntime>( + &onnxruntime, + "handle", + )?; let open_jtalk = env .get_rust_field::<_, _, voicevox_core::blocking::OpenJtalk>(&open_jtalk, "handle")? .clone(); let internal = Arc::new(voicevox_core::blocking::Synthesizer::new( - open_jtalk, &options, + onnxruntime, + open_jtalk, + &options, )?); env.set_rust_field(&this, "handle", internal)?; Ok(()) diff --git a/crates/voicevox_core_python_api/Cargo.toml b/crates/voicevox_core_python_api/Cargo.toml index 5ccd1dc41..cc83d583e 100644 --- a/crates/voicevox_core_python_api/Cargo.toml +++ b/crates/voicevox_core_python_api/Cargo.toml @@ -15,6 +15,7 @@ directml = ["voicevox_core/directml"] camino.workspace = true easy-ext.workspace = true log.workspace = true +once_cell.workspace = true pyo3 = { workspace = true, features = ["abi3-py38", "extension-module"] } pyo3-asyncio = { workspace = true, features = ["tokio-runtime"] } pyo3-log.workspace = true @@ -22,7 +23,7 @@ serde = { workspace = true, features = ["derive"] } serde_json.workspace = true tracing = { workspace = true, features = ["log"] } uuid.workspace = true -voicevox_core.workspace = true +voicevox_core = { workspace = true, features = ["onnxruntime-libloading"] } [lints.rust] unsafe_code = "forbid" diff --git a/crates/voicevox_core_python_api/python/test/conftest.py b/crates/voicevox_core_python_api/python/test/conftest.py index eec642cb9..430e415c1 100644 --- a/crates/voicevox_core_python_api/python/test/conftest.py +++ b/crates/voicevox_core_python_api/python/test/conftest.py @@ -4,9 +4,17 @@ from typing import List, TypedDict import pytest +import voicevox_core root_dir = Path(os.path.dirname(os.path.abspath(__file__))) +onnxruntime_filename = str( + root_dir.parent.parent.parent + / "test_util" + / "data" + / "lib" + / voicevox_core.blocking.Onnxruntime.LIB_VERSIONED_FILENAME +) open_jtalk_dic_dir = ( root_dir.parent.parent.parent / "test_util" / "data" / "open_jtalk_dic_utf_8-1.11" ) diff --git a/crates/voicevox_core_python_api/python/test/test_asyncio_metas.py b/crates/voicevox_core_python_api/python/test/test_asyncio_metas.py index ec69032b1..aea4af999 100644 --- a/crates/voicevox_core_python_api/python/test/test_asyncio_metas.py +++ b/crates/voicevox_core_python_api/python/test/test_asyncio_metas.py @@ -7,7 +7,7 @@ import conftest import pytest import pytest_asyncio -from voicevox_core.asyncio import OpenJtalk, Synthesizer, VoiceModel +from voicevox_core.asyncio import Onnxruntime, OpenJtalk, Synthesizer, VoiceModel def test_voice_model_metas_works(voice_model: VoiceModel) -> None: @@ -16,7 +16,10 @@ def test_voice_model_metas_works(voice_model: VoiceModel) -> None: @pytest.mark.asyncio async def test_synthesizer_metas_works(voice_model: VoiceModel) -> None: - synthesizer = Synthesizer(await OpenJtalk.new(conftest.open_jtalk_dic_dir)) + synthesizer = Synthesizer( + await Onnxruntime.load_once(filename=conftest.onnxruntime_filename), + await OpenJtalk.new(conftest.open_jtalk_dic_dir), + ) await synthesizer.load_voice_model(voice_model) _ = synthesizer.metas diff --git a/crates/voicevox_core_python_api/python/test/test_asyncio_user_dict_load.py b/crates/voicevox_core_python_api/python/test/test_asyncio_user_dict_load.py index c509b8c2d..d6906a6ac 100644 --- a/crates/voicevox_core_python_api/python/test/test_asyncio_user_dict_load.py +++ b/crates/voicevox_core_python_api/python/test/test_asyncio_user_dict_load.py @@ -15,9 +15,12 @@ @pytest.mark.asyncio async def test_user_dict_load() -> None: + onnxruntime = await voicevox_core.asyncio.Onnxruntime.load_once( + filename=conftest.onnxruntime_filename + ) open_jtalk = await voicevox_core.asyncio.OpenJtalk.new(conftest.open_jtalk_dic_dir) model = await voicevox_core.asyncio.VoiceModel.from_path(conftest.model_dir) - synthesizer = voicevox_core.asyncio.Synthesizer(open_jtalk) + synthesizer = voicevox_core.asyncio.Synthesizer(onnxruntime, open_jtalk) await synthesizer.load_voice_model(model) diff --git a/crates/voicevox_core_python_api/python/test/test_blocking_metas.py b/crates/voicevox_core_python_api/python/test/test_blocking_metas.py index c305e2cdb..00eade04b 100644 --- a/crates/voicevox_core_python_api/python/test/test_blocking_metas.py +++ b/crates/voicevox_core_python_api/python/test/test_blocking_metas.py @@ -6,7 +6,7 @@ import conftest import pytest -from voicevox_core.blocking import OpenJtalk, Synthesizer, VoiceModel +from voicevox_core.blocking import Onnxruntime, OpenJtalk, Synthesizer, VoiceModel def test_voice_model_metas_works(voice_model: VoiceModel) -> None: @@ -14,7 +14,10 @@ def test_voice_model_metas_works(voice_model: VoiceModel) -> None: def test_synthesizer_metas_works(voice_model: VoiceModel) -> None: - synthesizer = Synthesizer(OpenJtalk(conftest.open_jtalk_dic_dir)) + synthesizer = Synthesizer( + Onnxruntime.load_once(filename=conftest.onnxruntime_filename), + OpenJtalk(conftest.open_jtalk_dic_dir), + ) synthesizer.load_voice_model(voice_model) _ = synthesizer.metas diff --git a/crates/voicevox_core_python_api/python/test/test_blocking_user_dict_load.py b/crates/voicevox_core_python_api/python/test/test_blocking_user_dict_load.py index ef94d9742..198becbe2 100644 --- a/crates/voicevox_core_python_api/python/test/test_blocking_user_dict_load.py +++ b/crates/voicevox_core_python_api/python/test/test_blocking_user_dict_load.py @@ -13,9 +13,12 @@ def test_user_dict_load() -> None: + onnxruntime = voicevox_core.blocking.Onnxruntime.load_once( + filename=conftest.onnxruntime_filename + ) open_jtalk = voicevox_core.blocking.OpenJtalk(conftest.open_jtalk_dic_dir) model = voicevox_core.blocking.VoiceModel.from_path(conftest.model_dir) - synthesizer = voicevox_core.blocking.Synthesizer(open_jtalk) + synthesizer = voicevox_core.blocking.Synthesizer(onnxruntime, open_jtalk) synthesizer.load_voice_model(model) diff --git a/crates/voicevox_core_python_api/python/test/test_pseudo_raii_for_asyncio_synthesizer.py b/crates/voicevox_core_python_api/python/test/test_pseudo_raii_for_asyncio_synthesizer.py index 93d92ad28..26d389477 100644 --- a/crates/voicevox_core_python_api/python/test/test_pseudo_raii_for_asyncio_synthesizer.py +++ b/crates/voicevox_core_python_api/python/test/test_pseudo_raii_for_asyncio_synthesizer.py @@ -7,7 +7,7 @@ import conftest import pytest import pytest_asyncio -from voicevox_core.asyncio import OpenJtalk, Synthesizer +from voicevox_core.asyncio import Onnxruntime, OpenJtalk, Synthesizer def test_enter_returns_workable_self(synthesizer: Synthesizer) -> None: @@ -38,8 +38,13 @@ def test_access_after_exit_denied(synthesizer: Synthesizer) -> None: @pytest_asyncio.fixture -async def synthesizer(open_jtalk: OpenJtalk) -> Synthesizer: - return Synthesizer(open_jtalk) +async def synthesizer(onnxruntime: Onnxruntime, open_jtalk: OpenJtalk) -> Synthesizer: + return Synthesizer(onnxruntime, open_jtalk) + + +@pytest_asyncio.fixture(scope="function") +async def onnxruntime() -> Onnxruntime: + return await Onnxruntime.load_once(filename=conftest.onnxruntime_filename) @pytest_asyncio.fixture(scope="function") diff --git a/crates/voicevox_core_python_api/python/test/test_pseudo_raii_for_blocking_synthesizer.py b/crates/voicevox_core_python_api/python/test/test_pseudo_raii_for_blocking_synthesizer.py index 3e3f5f823..dc55eafc4 100644 --- a/crates/voicevox_core_python_api/python/test/test_pseudo_raii_for_blocking_synthesizer.py +++ b/crates/voicevox_core_python_api/python/test/test_pseudo_raii_for_blocking_synthesizer.py @@ -6,7 +6,7 @@ import conftest import pytest -from voicevox_core.blocking import OpenJtalk, Synthesizer +from voicevox_core.blocking import Onnxruntime, OpenJtalk, Synthesizer def test_enter_returns_workable_self(synthesizer: Synthesizer) -> None: @@ -37,8 +37,13 @@ def test_access_after_exit_denied(synthesizer: Synthesizer) -> None: @pytest.fixture -def synthesizer(open_jtalk: OpenJtalk) -> Synthesizer: - return Synthesizer(open_jtalk) +def synthesizer(onnxruntime: Onnxruntime, open_jtalk: OpenJtalk) -> Synthesizer: + return Synthesizer(onnxruntime, open_jtalk) + + +@pytest.fixture(scope="session") +def onnxruntime() -> Onnxruntime: + return Onnxruntime.load_once(filename=conftest.onnxruntime_filename) @pytest.fixture(scope="session") diff --git a/crates/voicevox_core_python_api/python/test/test_type_stub_consts.py b/crates/voicevox_core_python_api/python/test/test_type_stub_consts.py new file mode 100644 index 000000000..6a44d2771 --- /dev/null +++ b/crates/voicevox_core_python_api/python/test/test_type_stub_consts.py @@ -0,0 +1,50 @@ +"""pyiに書かれている定数の値が、本物と合致しているかをテストする。""" + +import ast +from ast import AnnAssign, ClassDef, Constant, Name +from pathlib import Path +from typing import Tuple + +import voicevox_core + + +def test() -> None: + REAL_BLOCKING = ( + voicevox_core.blocking.Onnxruntime.LIB_NAME, + voicevox_core.blocking.Onnxruntime.LIB_VERSION, + ) + REAL_ASYNCIO = ( + voicevox_core.asyncio.Onnxruntime.LIB_NAME, + voicevox_core.asyncio.Onnxruntime.LIB_VERSION, + ) + stub_blocking = extract(Path("./python/voicevox_core/_rust/blocking.pyi")) + stub_asyncio = extract(Path("./python/voicevox_core/_rust/asyncio.pyi")) + assert len({REAL_BLOCKING, REAL_ASYNCIO, stub_blocking, stub_asyncio}) == 1 + + +def extract(pyi: Path) -> Tuple[str, str]: + module = ast.parse(pyi.read_text(encoding="utf-8")) + class_def = next( + stmt + for stmt in module.body + if isinstance(stmt, ClassDef) and stmt.name == "Onnxruntime" + ) + lib_name_value = next( + stmt.value.value + for stmt in class_def.body + if isinstance(stmt, AnnAssign) + and isinstance(stmt.target, Name) + and stmt.target.id == "LIB_NAME" + and isinstance(stmt.value, Constant) + and isinstance(stmt.value.value, str) + ) + lib_version_value = next( + stmt.value.value + for stmt in class_def.body + if isinstance(stmt, AnnAssign) + and isinstance(stmt.target, Name) + and stmt.target.id == "LIB_VERSION" + and isinstance(stmt.value, Constant) + and isinstance(stmt.value.value, str) + ) + return (lib_name_value, lib_version_value) diff --git a/crates/voicevox_core_python_api/python/voicevox_core/__init__.py b/crates/voicevox_core_python_api/python/voicevox_core/__init__.py index 4ccbad3fe..e9a0a03c0 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/__init__.py +++ b/crates/voicevox_core_python_api/python/voicevox_core/__init__.py @@ -19,6 +19,7 @@ GetSupportedDevicesError, GpuSupportError, InferenceFailedError, + InitInferenceRuntimeError, InvalidModelDataError, InvalidWordError, LoadUserDictError, @@ -34,7 +35,6 @@ UseUserDictError, WordNotFoundError, __version__, - supported_devices, ) from . import asyncio, blocking # noqa: F401 isort: skip @@ -50,6 +50,7 @@ "GetSupportedDevicesError", "GpuSupportError", "InferenceFailedError", + "InitInferenceRuntimeError", "InvalidModelDataError", "InvalidWordError", "LoadUserDictError", @@ -67,7 +68,6 @@ "StyleNotFoundError", "StyleVersion", "SupportedDevices", - "supported_devices", "UseUserDictError", "UserDictWord", "UserDictWordType", diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_load_dlls.py b/crates/voicevox_core_python_api/python/voicevox_core/_load_dlls.py index db8b4cdc5..0f9e3d034 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/_load_dlls.py +++ b/crates/voicevox_core_python_api/python/voicevox_core/_load_dlls.py @@ -1,3 +1,4 @@ +# TODO: voicevox_onnxruntimeになったらやめる import glob import platform from ctypes import CDLL diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_rust/__init__.pyi b/crates/voicevox_core_python_api/python/voicevox_core/_rust/__init__.pyi index 89a50d230..81168843f 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/_rust/__init__.pyi +++ b/crates/voicevox_core_python_api/python/voicevox_core/_rust/__init__.pyi @@ -1,22 +1,5 @@ -from typing import TYPE_CHECKING - -if TYPE_CHECKING: - from voicevox_core import SupportedDevices - __version__: str -def supported_devices() -> SupportedDevices: - """ - このライブラリで利用可能なデバイスの情報を取得する。 - - .. code-block:: - - import voicevox_core - - supported_devices = voicevox_core.supported_devices() - """ - ... - class NotLoadedOpenjtalkDictError(Exception): """open_jtalk辞書ファイルが読み込まれていない。""" @@ -27,6 +10,11 @@ class GpuSupportError(Exception): ... +class InitInferenceRuntimeError(Exception): + """推論ライブラリのロードまたは初期化ができなかった。""" + + ... + class OpenZipFileError(Exception): """ZIPファイルを開くことに失敗した。""" diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi b/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi index 468d885ee..6066a35aa 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi +++ b/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi @@ -9,6 +9,7 @@ if TYPE_CHECKING: AudioQuery, SpeakerMeta, StyleId, + SupportedDevices, UserDictWord, VoiceModelId, ) @@ -37,6 +38,70 @@ class VoiceModel: """メタ情報。""" ... +class Onnxruntime: + """ + ONNX Runtime。 + + シングルトンであり、インスタンスは高々一つ。 + + .. code-block:: + + ort1 = await Onnxruntime.load_once() + ort2 = Onnxruntime.get() + assert ort2 + assert ort2 is ort1 + + .. code-block:: + + ort = await voicevox_core.asyncio.Onnxruntime.load_once() + assert voicevox_core.blocking.Onnxruntime.get() + """ + + # ここの定数値が本物と合致するかどうかは、test_type_stub_consts.pyで担保する。 + + LIB_NAME: str = "onnxruntime" + """ONNX Runtimeのライブラリ名。""" + + LIB_VERSION: str = "1.17.3" + """推奨されるONNX Runtimeのバージョン。""" + + LIB_VERSIONED_FILENAME: str + """ + :attr:`LIB_NAME` と :attr:`LIB_VERSION` からなる動的ライブラリのファイル名。 + + WindowsとAndroidでは :attr:`LIB_UNVERSIONED_FILENAME` と同じ。 + """ + + LIB_UNVERSIONED_FILENAME: str + """:attr:`LIB_NAME` からなる動的ライブラリのファイル名。""" + + @staticmethod + def get() -> Union["Onnxruntime", None]: + """ + インスタンスが既に作られているならそれを得る。 + + 作られていなければ `None` を返す。 + """ + ... + @staticmethod + async def load_once(*, filename: str = LIB_VERSIONED_FILENAME) -> "Onnxruntime": + """ + ONNX Runtimeをロードして初期化する。 + + 一度成功したら、以後は引数を無視して同じインスタンスを返す。 + + Parameters + ---------- + filename + ``dlopen``/``LoadLibraryExW`` の引数に使われるfilename。 + """ + ... + def supported_devices(self) -> SupportedDevices: + """ + このライブラリで利用可能なデバイスの情報を取得する。 + """ + ... + class OpenJtalk: """ テキスト解析器としてのOpen JTalk。 @@ -72,6 +137,8 @@ class Synthesizer: Parameters ---------- + onnxruntime + ONNX Runtime。 open_jtalk Open JTalk。 acceleration_mode @@ -82,6 +149,7 @@ class Synthesizer: def __init__( self, + onnxruntime: Onnxruntime, open_jtalk: OpenJtalk, acceleration_mode: Union[ AccelerationMode, Literal["AUTO", "CPU", "GPU"] @@ -92,6 +160,10 @@ class Synthesizer: def __enter__(self) -> "Synthesizer": ... def __exit__(self, exc_type, exc_value, traceback) -> None: ... @property + def onnxruntime(self) -> Onnxruntime: + """ONNX Runtime。""" + ... + @property def is_gpu_mode(self) -> bool: """ハードウェアアクセラレーションがGPUモードかどうか。""" ... diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi b/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi index fd09eb8cd..5f96c38bb 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi +++ b/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi @@ -9,6 +9,7 @@ if TYPE_CHECKING: AudioQuery, SpeakerMeta, StyleId, + SupportedDevices, UserDictWord, VoiceModelId, ) @@ -37,6 +38,70 @@ class VoiceModel: """メタ情報。""" ... +class Onnxruntime: + """ + ONNX Runtime。 + + シングルトンであり、インスタンスは高々一つ。 + + .. code-block:: + + ort1 = Onnxruntime.load_once() + ort2 = Onnxruntime.get() + assert ort2 + assert ort2 is ort1 + + .. code-block:: + + ort = voicevox_core.blocking.Onnxruntime.load_once() + assert voicevox_core.asyncio.Onnxruntime.get() + """ + + # ここの定数値が本物と合致するかどうかは、test_type_stub_consts.pyで担保する。 + + LIB_NAME: str = "onnxruntime" + """ONNX Runtimeのライブラリ名。""" + + LIB_VERSION: str = "1.17.3" + """推奨されるONNX Runtimeのバージョン。""" + + LIB_VERSIONED_FILENAME: str + """ + :attr:`LIB_NAME` と :attr:`LIB_VERSION` からなる動的ライブラリのファイル名。 + + WindowsとAndroidでは :attr:`LIB_UNVERSIONED_FILENAME` と同じ。 + """ + + LIB_UNVERSIONED_FILENAME: str + """:attr:`LIB_NAME` からなる動的ライブラリのファイル名。""" + + @staticmethod + def get() -> Union["Onnxruntime", None]: + """ + インスタンスが既に作られているならそれを得る。 + + 作られていなければ `None` を返す。 + """ + ... + @staticmethod + def load_once(*, filename: str = LIB_VERSIONED_FILENAME) -> "Onnxruntime": + """ + ONNX Runtimeをロードして初期化する。 + + 一度成功したら、以後は引数を無視して同じインスタンスを返す。 + + Parameters + ---------- + filename + ``dlopen``/``LoadLibraryExW`` の引数に使われるfilename。 + """ + ... + def supported_devices(self) -> SupportedDevices: + """ + このライブラリで利用可能なデバイスの情報を取得する。 + """ + ... + class OpenJtalk: """ テキスト解析器としてのOpen JTalk。 @@ -67,6 +132,8 @@ class Synthesizer: Parameters ---------- + onnxruntime + ONNX Runtime。 open_jtalk Open JTalk。 acceleration_mode @@ -77,6 +144,7 @@ class Synthesizer: def __init__( self, + onnxruntime: Onnxruntime, open_jtalk: OpenJtalk, acceleration_mode: Union[ AccelerationMode, Literal["AUTO", "CPU", "GPU"] @@ -87,6 +155,10 @@ class Synthesizer: def __enter__(self) -> "Synthesizer": ... def __exit__(self, exc_type, exc_value, traceback) -> None: ... @property + def onnxruntime(self) -> Onnxruntime: + """ONNX Runtime。""" + ... + @property def is_gpu_mode(self) -> bool: """ハードウェアアクセラレーションがGPUモードかどうか。""" ... diff --git a/crates/voicevox_core_python_api/python/voicevox_core/asyncio.py b/crates/voicevox_core_python_api/python/voicevox_core/asyncio.py index 75b160814..2cff19cdf 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/asyncio.py +++ b/crates/voicevox_core_python_api/python/voicevox_core/asyncio.py @@ -1,4 +1,4 @@ # pyright: reportMissingModuleSource=false -from ._rust.asyncio import OpenJtalk, Synthesizer, UserDict, VoiceModel +from ._rust.asyncio import Onnxruntime, OpenJtalk, Synthesizer, UserDict, VoiceModel -__all__ = ["OpenJtalk", "Synthesizer", "UserDict", "VoiceModel"] +__all__ = ["Onnxruntime", "OpenJtalk", "Synthesizer", "UserDict", "VoiceModel"] diff --git a/crates/voicevox_core_python_api/python/voicevox_core/blocking.py b/crates/voicevox_core_python_api/python/voicevox_core/blocking.py index 80f61fdcb..7fed5fac7 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/blocking.py +++ b/crates/voicevox_core_python_api/python/voicevox_core/blocking.py @@ -1,4 +1,4 @@ # pyright: reportMissingModuleSource=false -from ._rust.blocking import OpenJtalk, Synthesizer, UserDict, VoiceModel +from ._rust.blocking import Onnxruntime, OpenJtalk, Synthesizer, UserDict, VoiceModel -__all__ = ["OpenJtalk", "Synthesizer", "UserDict", "VoiceModel"] +__all__ = ["Onnxruntime", "OpenJtalk", "Synthesizer", "UserDict", "VoiceModel"] diff --git a/crates/voicevox_core_python_api/src/convert.rs b/crates/voicevox_core_python_api/src/convert.rs index f40b2d449..0eee14a5b 100644 --- a/crates/voicevox_core_python_api/src/convert.rs +++ b/crates/voicevox_core_python_api/src/convert.rs @@ -16,10 +16,10 @@ use voicevox_core::{ use crate::{ ExtractFullContextLabelError, GetSupportedDevicesError, GpuSupportError, InferenceFailedError, - InvalidModelDataError, InvalidModelFormatError, InvalidWordError, LoadUserDictError, - ModelAlreadyLoadedError, ModelNotFoundError, NotLoadedOpenjtalkDictError, OpenZipFileError, - ParseKanaError, ReadZipEntryError, SaveUserDictError, StyleAlreadyLoadedError, - StyleNotFoundError, UseUserDictError, WordNotFoundError, + InitInferenceRuntimeError, InvalidModelDataError, InvalidModelFormatError, InvalidWordError, + LoadUserDictError, ModelAlreadyLoadedError, ModelNotFoundError, NotLoadedOpenjtalkDictError, + OpenZipFileError, ParseKanaError, ReadZipEntryError, SaveUserDictError, + StyleAlreadyLoadedError, StyleNotFoundError, UseUserDictError, WordNotFoundError, }; pub(crate) fn from_acceleration_mode(ob: &PyAny) -> PyResult { @@ -191,6 +191,7 @@ pub(crate) impl voicevox_core::Result { let top = match err.kind() { ErrorKind::NotLoadedOpenjtalkDict => NotLoadedOpenjtalkDictError::new_err(msg), ErrorKind::GpuSupport => GpuSupportError::new_err(msg), + ErrorKind::InitInferenceRuntime => InitInferenceRuntimeError::new_err(msg), ErrorKind::OpenZipFile => OpenZipFileError::new_err(msg), ErrorKind::ReadZipEntry => ReadZipEntryError::new_err(msg), ErrorKind::ModelAlreadyLoaded => ModelAlreadyLoadedError::new_err(msg), diff --git a/crates/voicevox_core_python_api/src/lib.rs b/crates/voicevox_core_python_api/src/lib.rs index e43fa4476..0da82f9ca 100644 --- a/crates/voicevox_core_python_api/src/lib.rs +++ b/crates/voicevox_core_python_api/src/lib.rs @@ -1,7 +1,7 @@ use std::marker::PhantomData; mod convert; -use self::convert::{from_utf8_path, to_pydantic_dataclass, VoicevoxCoreResultExt as _}; +use self::convert::{from_utf8_path, VoicevoxCoreResultExt as _}; use easy_ext::ext; use log::debug; use pyo3::{ @@ -9,7 +9,7 @@ use pyo3::{ exceptions::{PyException, PyKeyError, PyValueError}, pyfunction, pymodule, types::PyModule, - wrap_pyfunction, PyAny, PyResult, PyTypeInfo, Python, + wrap_pyfunction, PyResult, PyTypeInfo, Python, }; #[pymodule] @@ -18,7 +18,6 @@ fn rust(py: Python<'_>, module: &PyModule) -> PyResult<()> { pyo3_log::init(); module.add("__version__", env!("CARGO_PKG_VERSION"))?; - module.add_wrapped(wrap_pyfunction!(supported_devices))?; module.add_wrapped(wrap_pyfunction!(_validate_pronunciation))?; module.add_wrapped(wrap_pyfunction!(_to_zenkaku))?; @@ -26,6 +25,7 @@ fn rust(py: Python<'_>, module: &PyModule) -> PyResult<()> { let blocking_module = PyModule::new(py, "voicevox_core._rust.blocking")?; blocking_module.add_class::()?; + blocking_module.add_class::()?; blocking_module.add_class::()?; blocking_module.add_class::()?; blocking_module.add_class::()?; @@ -33,6 +33,7 @@ fn rust(py: Python<'_>, module: &PyModule) -> PyResult<()> { let asyncio_module = PyModule::new(py, "voicevox_core._rust.asyncio")?; asyncio_module.add_class::()?; + asyncio_module.add_class::()?; asyncio_module.add_class::()?; asyncio_module.add_class::()?; asyncio_module.add_class::()?; @@ -67,6 +68,7 @@ macro_rules! exceptions { exceptions! { NotLoadedOpenjtalkDictError: PyException; GpuSupportError: PyException; + InitInferenceRuntimeError: PyException; OpenZipFileError: PyException; ReadZipEntryError: PyException; ModelAlreadyLoadedError: PyException; @@ -86,16 +88,6 @@ exceptions! { InvalidWordError: PyValueError; } -#[pyfunction] -fn supported_devices(py: Python<'_>) -> PyResult<&PyAny> { - let class = py - .import("voicevox_core")? - .getattr("SupportedDevices")? - .downcast()?; - let s = voicevox_core::SupportedDevices::create().into_py_result(py)?; - to_pydantic_dataclass(s, class) -} - struct Closable { content: MaybeClosed, marker: PhantomData, @@ -149,13 +141,13 @@ fn _to_zenkaku(text: &str) -> PyResult { } mod blocking { - use std::{path::PathBuf, sync::Arc}; + use std::{ffi::OsString, path::PathBuf, sync::Arc}; use camino::Utf8PathBuf; use pyo3::{ pyclass, pymethods, types::{IntoPyDict as _, PyBytes, PyDict, PyList}, - PyAny, PyObject, PyRef, PyResult, Python, + Py, PyAny, PyObject, PyRef, PyResult, Python, }; use uuid::Uuid; use voicevox_core::{ @@ -191,6 +183,70 @@ mod blocking { } } + static ONNXRUNTIME: once_cell::sync::OnceCell> = + once_cell::sync::OnceCell::new(); + + #[pyclass] + #[derive(Clone)] + pub(crate) struct Onnxruntime(&'static voicevox_core::blocking::Onnxruntime); + + #[pymethods] + impl Onnxruntime { + #[classattr] + const LIB_NAME: &'static str = voicevox_core::blocking::Onnxruntime::LIB_NAME; + + #[classattr] + const LIB_VERSION: &'static str = voicevox_core::blocking::Onnxruntime::LIB_VERSION; + + #[classattr] + const LIB_VERSIONED_FILENAME: &'static str = + voicevox_core::blocking::Onnxruntime::LIB_VERSIONED_FILENAME; + + #[classattr] + const LIB_UNVERSIONED_FILENAME: &'static str = + voicevox_core::blocking::Onnxruntime::LIB_UNVERSIONED_FILENAME; + + #[staticmethod] + fn get(py: Python<'_>) -> PyResult>> { + let result = ONNXRUNTIME.get_or_try_init(|| { + match voicevox_core::blocking::Onnxruntime::get().map(|o| Py::new(py, Self(o))) { + Some(Ok(this)) => Ok(this), + Some(Err(err)) => Err(Some(err)), + None => Err(None), + } + }); + + match result { + Ok(this) => Ok(Some(this.clone())), + Err(Some(err)) => Err(err), + Err(None) => Ok(None), + } + } + + #[staticmethod] + #[pyo3(signature = (*, filename = Self::LIB_VERSIONED_FILENAME.into()))] + fn load_once(filename: OsString, py: Python<'_>) -> PyResult> { + ONNXRUNTIME + .get_or_try_init(|| { + let inner = voicevox_core::blocking::Onnxruntime::load_once() + .filename(filename) + .exec() + .into_py_result(py)?; + Py::new(py, Self(inner)) + }) + .cloned() + } + + fn supported_devices<'py>(&self, py: Python<'py>) -> PyResult<&'py PyAny> { + let class = py + .import("voicevox_core")? + .getattr("SupportedDevices")? + .downcast()?; + let s = self.0.supported_devices().into_py_result(py)?; + crate::convert::to_pydantic_dataclass(s, class) + } + } + #[pyclass] #[derive(Clone)] pub(crate) struct OpenJtalk { @@ -228,11 +284,13 @@ mod blocking { impl Synthesizer { #[new] #[pyo3(signature =( + onnxruntime, open_jtalk, acceleration_mode = InitializeOptions::default().acceleration_mode, cpu_num_threads = InitializeOptions::default().cpu_num_threads, ))] fn new( + onnxruntime: Onnxruntime, open_jtalk: OpenJtalk, #[pyo3(from_py_with = "crate::convert::from_acceleration_mode")] acceleration_mode: AccelerationMode, @@ -240,6 +298,7 @@ mod blocking { py: Python<'_>, ) -> PyResult { let inner = voicevox_core::blocking::Synthesizer::new( + onnxruntime.0, open_jtalk.open_jtalk.clone(), &InitializeOptions { acceleration_mode, @@ -270,6 +329,11 @@ mod blocking { self.close(); } + #[getter] + fn onnxruntime(&self) -> Py { + ONNXRUNTIME.get().expect("should be initialized").clone() + } + #[getter] fn is_gpu_mode(&self) -> PyResult { let synthesizer = self.synthesizer.get()?; @@ -577,13 +641,13 @@ mod blocking { } mod asyncio { - use std::{path::PathBuf, sync::Arc}; + use std::{ffi::OsString, path::PathBuf, sync::Arc}; use camino::Utf8PathBuf; use pyo3::{ pyclass, pymethods, types::{IntoPyDict as _, PyBytes, PyDict, PyList}, - PyAny, PyObject, PyRef, PyResult, Python, ToPyObject as _, + Py, PyAny, PyObject, PyRef, PyResult, Python, ToPyObject as _, }; use uuid::Uuid; use voicevox_core::{ @@ -622,6 +686,71 @@ mod asyncio { } } + static ONNXRUNTIME: once_cell::sync::OnceCell> = + once_cell::sync::OnceCell::new(); + + #[pyclass] + #[derive(Clone)] + pub(crate) struct Onnxruntime(&'static voicevox_core::tokio::Onnxruntime); + + #[pymethods] + impl Onnxruntime { + #[classattr] + const LIB_NAME: &'static str = voicevox_core::tokio::Onnxruntime::LIB_NAME; + + #[classattr] + const LIB_VERSION: &'static str = voicevox_core::tokio::Onnxruntime::LIB_VERSION; + + #[classattr] + const LIB_VERSIONED_FILENAME: &'static str = + voicevox_core::tokio::Onnxruntime::LIB_VERSIONED_FILENAME; + + #[classattr] + const LIB_UNVERSIONED_FILENAME: &'static str = + voicevox_core::tokio::Onnxruntime::LIB_UNVERSIONED_FILENAME; + + #[staticmethod] + fn get(py: Python<'_>) -> PyResult>> { + let result = ONNXRUNTIME.get_or_try_init(|| { + match voicevox_core::tokio::Onnxruntime::get().map(|o| Py::new(py, Self(o))) { + Some(Ok(this)) => Ok(this), + Some(Err(err)) => Err(Some(err)), + None => Err(None), + } + }); + + match result { + Ok(this) => Ok(Some(this.clone())), + Err(Some(err)) => Err(err), + Err(None) => Ok(None), + } + } + + #[staticmethod] + #[pyo3(signature = (*, filename = Self::LIB_VERSIONED_FILENAME.into()))] + fn load_once(filename: OsString, py: Python<'_>) -> PyResult<&PyAny> { + pyo3_asyncio::tokio::future_into_py(py, async move { + let inner = voicevox_core::tokio::Onnxruntime::load_once() + .filename(filename) + .exec() + .await; + + ONNXRUNTIME.get_or_try_init(|| { + Python::with_gil(|py| Py::new(py, Self(inner.into_py_result(py)?))) + }) + }) + } + + fn supported_devices<'py>(&self, py: Python<'py>) -> PyResult<&'py PyAny> { + let class = py + .import("voicevox_core")? + .getattr("SupportedDevices")? + .downcast()?; + let s = self.0.supported_devices().into_py_result(py)?; + crate::convert::to_pydantic_dataclass(s, class) + } + } + #[pyclass] #[derive(Clone)] pub(crate) struct OpenJtalk { @@ -664,17 +793,20 @@ mod asyncio { impl Synthesizer { #[new] #[pyo3(signature =( + onnxruntime, open_jtalk, acceleration_mode = InitializeOptions::default().acceleration_mode, cpu_num_threads = InitializeOptions::default().cpu_num_threads, ))] fn new( + onnxruntime: Onnxruntime, open_jtalk: OpenJtalk, #[pyo3(from_py_with = "crate::convert::from_acceleration_mode")] acceleration_mode: AccelerationMode, cpu_num_threads: u16, ) -> PyResult { let synthesizer = voicevox_core::tokio::Synthesizer::new( + onnxruntime.0, open_jtalk.open_jtalk.clone(), &InitializeOptions { acceleration_mode, @@ -704,6 +836,11 @@ mod asyncio { self.close(); } + #[getter] + fn onnxruntime(&self) -> Py { + ONNXRUNTIME.get().expect("should be initialized").clone() + } + #[getter] fn is_gpu_mode(&self) -> PyResult { let synthesizer = self.synthesizer.get()?; diff --git a/docs/apis/c_api/doxygen/Doxyfile b/docs/apis/c_api/doxygen/Doxyfile index c42bd60e0..cd3313463 100644 --- a/docs/apis/c_api/doxygen/Doxyfile +++ b/docs/apis/c_api/doxygen/Doxyfile @@ -2257,7 +2257,7 @@ INCLUDE_FILE_PATTERNS = # recursively expanded use the := operator instead of the = operator. # This tag requires that the tag ENABLE_PREPROCESSING is set to YES. -PREDEFINED = +PREDEFINED = VOICEVOX_ONNXRUNTIME_LIBLOADING= VOICEVOX_ONNXRUNTIME_LINK_DYLIB= # If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then this # tag can be used to specify a list of macro names that should be expanded. The diff --git a/docs/usage.md b/docs/usage.md index 91da1c62f..426f0bc86 100644 --- a/docs/usage.md +++ b/docs/usage.md @@ -62,11 +62,11 @@ VOICEVOX コアでは`Synthesizer`に音声モデルを読み込むことでテ ```python from pprint import pprint -from voicevox_core.blocking import OpenJtalk, Synthesizer, VoiceModel +from voicevox_core.blocking import Onnxruntime, OpenJtalk, Synthesizer, VoiceModel # 1. Synthesizerの初期化 open_jtalk_dict_dir = "open_jtalk_dic_utf_8-1.11" -synthesizer = Synthesizer(OpenJtalk(open_jtalk_dict_dir)) +synthesizer = Synthesizer(Onnxruntime.load_once(), OpenJtalk(open_jtalk_dict_dir)) # 2. 音声モデルの読み込み model = VoiceModel.from_path("model/0.vvm") @@ -82,7 +82,7 @@ with open("output.wav", "wb") as f: ### 1. Synthesizer の初期化 -辞書などを取り扱う`OpenJtalk`のインスタンスを引数に渡して`Synthesizer`を初期化します。`Synthesizer`は音声合成だけでなく、音声モデルを複数読み込んだり、イントネーションのみを生成することもできます。 +ONNX Runtimeを表す`Onnxruntime`のインスタンスと、辞書などを取り扱う`OpenJtalk`のインスタンスを引数に渡して`Synthesizer`を初期化します。`Synthesizer`は音声合成だけでなく、音声モデルを複数読み込んだり、イントネーションのみを生成することもできます。 ### 2. 音声モデルの読み込み diff --git a/example/cpp/unix/simple_tts.cpp b/example/cpp/unix/simple_tts.cpp index c7683d49a..5db24b12e 100644 --- a/example/cpp/unix/simple_tts.cpp +++ b/example/cpp/unix/simple_tts.cpp @@ -20,14 +20,21 @@ int main(int argc, char *argv[]) { std::cout << "coreの初期化中..." << std::endl; auto initialize_options = voicevox_make_default_initialize_options(); + const VoicevoxOnnxruntime* onnxruntime; + auto load_ort_options = voicevox_make_default_load_onnxruntime_options(); + auto result = voicevox_onnxruntime_load_once(load_ort_options, &onnxruntime); + if (result != VOICEVOX_RESULT_OK){ + std::cerr << voicevox_error_result_to_message(result) << std::endl; + return 1; + } OpenJtalkRc* open_jtalk; - auto result = voicevox_open_jtalk_rc_new(open_jtalk_dict_path.c_str(),&open_jtalk); + result = voicevox_open_jtalk_rc_new(open_jtalk_dict_path.c_str(),&open_jtalk); if (result != VOICEVOX_RESULT_OK){ std::cerr << voicevox_error_result_to_message(result) << std::endl; return 1; } VoicevoxSynthesizer* synthesizer; - result = voicevox_synthesizer_new(open_jtalk,initialize_options,&synthesizer); + result = voicevox_synthesizer_new(onnxruntime,open_jtalk,initialize_options,&synthesizer); if (result != VOICEVOX_RESULT_OK) { std::cerr << voicevox_error_result_to_message(result) << std::endl; return 1; diff --git a/example/cpp/windows/simple_tts/simple_tts.cpp b/example/cpp/windows/simple_tts/simple_tts.cpp index bd070505f..946ef9679 100644 --- a/example/cpp/windows/simple_tts/simple_tts.cpp +++ b/example/cpp/windows/simple_tts/simple_tts.cpp @@ -33,14 +33,21 @@ int main() { VoicevoxInitializeOptions initializeOptions = voicevox_make_default_initialize_options(); std::string dict = GetOpenJTalkDict(); + const VoicevoxOnnxruntime* onnxruntime; + auto load_ort_options = voicevox_make_default_load_onnxruntime_options(); + auto result = voicevox_onnxruntime_load_once(load_ort_options, &onnxruntime); + if (result != VoicevoxResultCode::VOICEVOX_RESULT_OK) { + OutErrorMessage(result); + return 0; + } OpenJtalkRc* open_jtalk; - auto result = voicevox_open_jtalk_rc_new(dict.c_str(),&open_jtalk); + result = voicevox_open_jtalk_rc_new(dict.c_str(),&open_jtalk); if (result != VoicevoxResultCode::VOICEVOX_RESULT_OK) { OutErrorMessage(result); return 0; } VoicevoxSynthesizer* synthesizer; - result = voicevox_synthesizer_new(open_jtalk,initializeOptions,&synthesizer); + result = voicevox_synthesizer_new(onnxruntime,open_jtalk,initializeOptions,&synthesizer); if (result != VoicevoxResultCode::VOICEVOX_RESULT_OK) { OutErrorMessage(result); return 0; diff --git a/example/kotlin/README.md b/example/kotlin/README.md index 3bcdafc94..f6f8170f4 100644 --- a/example/kotlin/README.md +++ b/example/kotlin/README.md @@ -47,6 +47,7 @@ Usage: voicevoxcoreexample options_list Options: --mode [AUTO] -> モード { Value should be one of [auto, cpu, gpu] } --vvm -> vvmファイルへのパス (always required) { String } + --onnxruntime [libonnxruntime.so.1.17.3] -> ONNX Runtimeのファイルネーム { String } --dictDir [./open_jtalk_dic_utf_8-1.11] -> Open JTalkの辞書ディレクトリ { String } --text [この音声は、ボイスボックスを使用して、出力されています。] -> 読み上げさせたい文章 { String } --out [./output.wav] -> 出力wavファイルのパス { String } @@ -56,9 +57,12 @@ Options: ## 実行例 + + ```console -❯ ./gradlew run --args="--vvm ../../crates/test_util/data/model/sample.vvm" -Inititalizing: AUTO, ./open_jtalk_dic_utf_8-1.11 +❯ # Linuxの場合 +❯ ./gradlew run --args="--vvm ../../crates/test_util/data/model/sample.vvm --onnxruntime ../../crates/test_util/data/lib/libonnxruntime.so.1.17.3" +Inititalizing: AUTO, ../../crates/test_util/data/lib/libonnxruntime.so.1.17.3, ./open_jtalk_dic_utf_8-1.11 Loading: ../../crates/test_util/data/model/sample.vvm Creating an AudioQuery from the text: この音声は、ボイスボックスを使用して、出力されています。 Synthesizing... diff --git a/example/kotlin/app/src/main/kotlin/app/App.kt b/example/kotlin/app/src/main/kotlin/app/App.kt index 3e76043f4..1c517c582 100644 --- a/example/kotlin/app/src/main/kotlin/app/App.kt +++ b/example/kotlin/app/src/main/kotlin/app/App.kt @@ -15,6 +15,10 @@ fun main(args: Array) { val mode by parser.option(ArgType.Choice(), description = "モード").default(Mode.AUTO) val vvmPath by parser.option(ArgType.String, fullName = "vvm", description = "vvmファイルへのパス").required() + val onnxruntime by + parser + .option(ArgType.String, description = "ONNX Runtimeのファイルネーム") + .default(Onnxruntime.LIB_VERSIONED_FILENAME) val dictDir by parser .option(ArgType.String, description = "Open JTalkの辞書ディレクトリ") @@ -28,10 +32,11 @@ fun main(args: Array) { parser.parse(args) - println("Inititalizing: ${mode}, ${dictDir}") + println("Inititalizing: ${mode}, ${onnxruntime}, ${dictDir}") + val ort = Onnxruntime.loadOnce().filename(onnxruntime).exec() val openJtalk = OpenJtalk(dictDir) val synthesizer = - Synthesizer.builder(openJtalk) + Synthesizer.builder(ort, openJtalk) .accelerationMode( when (mode) { Mode.AUTO -> Synthesizer.AccelerationMode.AUTO diff --git a/example/python/run-asyncio.py b/example/python/run-asyncio.py index 70d204a92..b75509183 100644 --- a/example/python/run-asyncio.py +++ b/example/python/run-asyncio.py @@ -8,9 +8,8 @@ from pathlib import Path from typing import Tuple -import voicevox_core from voicevox_core import AccelerationMode, AudioQuery -from voicevox_core.asyncio import OpenJtalk, Synthesizer, VoiceModel +from voicevox_core.asyncio import Onnxruntime, OpenJtalk, Synthesizer, VoiceModel async def main() -> None: @@ -23,17 +22,23 @@ async def main() -> None: ( acceleration_mode, vvm_path, + onnxruntime_filename, open_jtalk_dict_dir, text, out, style_id, ) = parse_args() - logger.debug("%s", f"{voicevox_core.supported_devices()=}") + logger.info("%s", f"Loading ONNX Runtime ({onnxruntime_filename=})") + onnxruntime = await Onnxruntime.load_once(filename=onnxruntime_filename) + + logger.debug("%s", f"{onnxruntime.supported_devices()=}") logger.info("%s", f"Initializing ({acceleration_mode=}, {open_jtalk_dict_dir=})") synthesizer = Synthesizer( - await OpenJtalk.new(open_jtalk_dict_dir), acceleration_mode=acceleration_mode + onnxruntime, + await OpenJtalk.new(open_jtalk_dict_dir), + acceleration_mode=acceleration_mode, ) logger.debug("%s", f"{synthesizer.metas=}") @@ -53,7 +58,7 @@ async def main() -> None: logger.info("%s", f"Wrote `{out}`") -def parse_args() -> Tuple[AccelerationMode, Path, Path, str, Path, int]: +def parse_args() -> Tuple[AccelerationMode, Path, str, Path, str, Path, int]: argparser = ArgumentParser() argparser.add_argument( "--mode", @@ -66,6 +71,11 @@ def parse_args() -> Tuple[AccelerationMode, Path, Path, str, Path, int]: type=Path, help="vvmファイルへのパス", ) + argparser.add_argument( + "--onnxruntime", + default=Onnxruntime.LIB_VERSIONED_FILENAME, + help="ONNX Runtimeのライブラリのfilename", + ) argparser.add_argument( "--dict-dir", default="./open_jtalk_dic_utf_8-1.11", @@ -90,7 +100,16 @@ def parse_args() -> Tuple[AccelerationMode, Path, Path, str, Path, int]: help="話者IDを指定", ) args = argparser.parse_args() - return (args.mode, args.vvm, args.dict_dir, args.text, args.out, args.style_id) + # FIXME: 流石に多くなってきたので、`dataclass`化する + return ( + args.mode, + args.vvm, + args.onnxruntime, + args.dict_dir, + args.text, + args.out, + args.style_id, + ) def display_as_json(audio_query: AudioQuery) -> str: diff --git a/example/python/run.py b/example/python/run.py index a57139b1c..3a9fdd9e7 100644 --- a/example/python/run.py +++ b/example/python/run.py @@ -5,9 +5,8 @@ from pathlib import Path from typing import Tuple -import voicevox_core from voicevox_core import AccelerationMode, AudioQuery -from voicevox_core.blocking import OpenJtalk, Synthesizer, VoiceModel +from voicevox_core.blocking import Onnxruntime, OpenJtalk, Synthesizer, VoiceModel def main() -> None: @@ -20,17 +19,23 @@ def main() -> None: ( acceleration_mode, vvm_path, + onnxruntime_filename, open_jtalk_dict_dir, text, out, style_id, ) = parse_args() - logger.debug("%s", f"{voicevox_core.supported_devices()=}") + logger.info("%s", f"Loading ONNX Runtime ({onnxruntime_filename=})") + onnxruntime = Onnxruntime.load_once(filename=onnxruntime_filename) + + logger.debug("%s", f"{onnxruntime.supported_devices()=}") logger.info("%s", f"Initializing ({acceleration_mode=}, {open_jtalk_dict_dir=})") synthesizer = Synthesizer( - OpenJtalk(open_jtalk_dict_dir), acceleration_mode=acceleration_mode + onnxruntime, + OpenJtalk(open_jtalk_dict_dir), + acceleration_mode=acceleration_mode, ) logger.debug("%s", f"{synthesizer.metas=}") @@ -50,7 +55,7 @@ def main() -> None: logger.info("%s", f"Wrote `{out}`") -def parse_args() -> Tuple[AccelerationMode, Path, Path, str, Path, int]: +def parse_args() -> Tuple[AccelerationMode, Path, str, Path, str, Path, int]: argparser = ArgumentParser() argparser.add_argument( "--mode", @@ -63,6 +68,11 @@ def parse_args() -> Tuple[AccelerationMode, Path, Path, str, Path, int]: type=Path, help="vvmファイルへのパス", ) + argparser.add_argument( + "--onnxruntime", + default=Onnxruntime.LIB_VERSIONED_FILENAME, + help="ONNX Runtimeのライブラリのfilename", + ) argparser.add_argument( "--dict-dir", default="./open_jtalk_dic_utf_8-1.11", @@ -87,7 +97,16 @@ def parse_args() -> Tuple[AccelerationMode, Path, Path, str, Path, int]: help="話者IDを指定", ) args = argparser.parse_args() - return (args.mode, args.vvm, args.dict_dir, args.text, args.out, args.style_id) + # FIXME: 流石に多くなってきたので、`dataclass`化する + return ( + args.mode, + args.vvm, + args.onnxruntime, + args.dict_dir, + args.text, + args.out, + args.style_id, + ) def display_as_json(audio_query: AudioQuery) -> str: From 0979c202fd48017b39841f2034a66905f4e14ddd Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Fri, 28 Jun 2024 23:08:31 +0900 Subject: [PATCH 02/26] =?UTF-8?q?Kotlin=E3=81=AEexample=E3=81=ABFIXME?= =?UTF-8?q?=E3=82=92=E8=BF=BD=E5=8A=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- example/kotlin/README.md | 1 + 1 file changed, 1 insertion(+) diff --git a/example/kotlin/README.md b/example/kotlin/README.md index f6f8170f4..9a244dd2c 100644 --- a/example/kotlin/README.md +++ b/example/kotlin/README.md @@ -58,6 +58,7 @@ Options: ## 実行例 + ```console ❯ # Linuxの場合 From ba8139316713f526a85a18b2b14ac175eaca39d3 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Fri, 28 Jun 2024 23:10:13 +0900 Subject: [PATCH 03/26] =?UTF-8?q?"ONNX=20Runtime=E3=82=92=E8=A1=A8?= =?UTF-8?q?=E3=81=99"=20=E2=86=92=20"AI=E3=82=A8=E3=83=B3=E3=82=B8?= =?UTF-8?q?=E3=83=B3=E3=81=AE"?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- docs/usage.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/usage.md b/docs/usage.md index 426f0bc86..e828ae220 100644 --- a/docs/usage.md +++ b/docs/usage.md @@ -82,7 +82,7 @@ with open("output.wav", "wb") as f: ### 1. Synthesizer の初期化 -ONNX Runtimeを表す`Onnxruntime`のインスタンスと、辞書などを取り扱う`OpenJtalk`のインスタンスを引数に渡して`Synthesizer`を初期化します。`Synthesizer`は音声合成だけでなく、音声モデルを複数読み込んだり、イントネーションのみを生成することもできます。 +AIエンジンの`Onnxruntime`のインスタンスと、辞書などを取り扱う`OpenJtalk`のインスタンスを引数に渡して`Synthesizer`を初期化します。`Synthesizer`は音声合成だけでなく、音声モデルを複数読み込んだり、イントネーションのみを生成することもできます。 ### 2. 音声モデルの読み込み From 379cc44efa72dc990a7d3cd60a79dc86422b9bd9 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sat, 29 Jun 2024 01:33:43 +0900 Subject: [PATCH 04/26] =?UTF-8?q?=E3=83=89=E3=82=AD=E3=83=A5=E3=83=A1?= =?UTF-8?q?=E3=83=B3=E3=83=88=E3=81=A8=E3=82=B3=E3=83=A1=E3=83=B3=E3=83=88?= =?UTF-8?q?=E3=82=92=E8=BF=BD=E5=8A=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .github/workflows/build_and_deploy.yml | 9 ++++++++- crates/voicevox_core/Cargo.toml | 8 ++++++++ docs/feature-options.md | 20 ++++++++++++++++++++ 3 files changed, 36 insertions(+), 1 deletion(-) create mode 100644 docs/feature-options.md diff --git a/.github/workflows/build_and_deploy.yml b/.github/workflows/build_and_deploy.yml index d336d1003..fce217748 100644 --- a/.github/workflows/build_and_deploy.yml +++ b/.github/workflows/build_and_deploy.yml @@ -41,7 +41,14 @@ defaults: shell: bash jobs: - config: # 全 jobs で利用する定数の定義。実行対象の条件をフィルタリングする。 + # 全 jobs で利用する定数の定義。実行対象の条件をフィルタリングする。 + # + # c_release_format = plain-cdylib | ios-xcframework + # + # `plain-cdylib`の場合、動的ライブラリとその付属物をZIPに固めたものをC APIとしてリリースする。 + # `ios-xcframework`の場合はiOS用のXCFrameworkをC APIとしてリリースする。また、ONNX Runtimeの + # リンク方法に関わるCargoフィーチャも`c_release_format`によって選択される。 + config: runs-on: ubuntu-latest outputs: includes: ${{ steps.strategy_matrix.outputs.includes }} diff --git a/crates/voicevox_core/Cargo.toml b/crates/voicevox_core/Cargo.toml index 00cc1dd63..222fe1ada 100644 --- a/crates/voicevox_core/Cargo.toml +++ b/crates/voicevox_core/Cargo.toml @@ -10,8 +10,16 @@ rustdoc-args = ["--cfg", "docsrs"] [features] default = [] + +# ONNX Runtimeのリンク方法を決めるフィーチャ。どちらか片方を有効化しなければならず、両方の有効化は +# 許容されない。 +# +# `onnxruntime-libloading`の場合、ONNX Runtimeを`dlopen`/`LoadLibraryExW`で開く。 +# `onnxruntime-link-dylib`の場合はONNX Runtimeをロード時動的リンクする。またそれに伴い、 +# `struct Onnxruntime`の初期化方法も異なる。 onnxruntime-libloading = ["voicevox-ort/load-dynamic"] onnxruntime-link-dylib = [] + cuda = ["voicevox-ort/cuda"] directml = ["voicevox-ort/directml"] diff --git a/docs/feature-options.md b/docs/feature-options.md new file mode 100644 index 000000000..cae419ac5 --- /dev/null +++ b/docs/feature-options.md @@ -0,0 +1,20 @@ +## ONNX Runtimeのリンク方法のオプション + +Rust API（`voicevox_core`）およびC API（`voicevox_core_c_api`）においては、ビルド時に +次のCargoフィーチャのうちどちらかを選択しなければなりません。 +詳しくは[voicevox_core/Cargo.toml](../crates/voicevox_core/Cargo.toml)のコメントを参照して +下さい。Python APIやJava APIでは`onnxruntime-libloading`のみに限定しています。 + +- `onnxruntime-libloading` +- `onnxruntime-link-dylib` + +```console +❯ cargo build --release -p voicevox_core_c_api --features onnxruntime-libloading +``` + +```console +❯ cargo build --release -p voicevox_core_c_api --features onnxruntime-link-dylib +``` + +C APIのリリースでは`dlopen`の利用が厳しいiOSでのみ`onnxruntime-link-dylib`で、その他 +は`onnxruntime-libloading`でビルドしています。 From b1dca9c42f155a312e6373c3a00274892320acac Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sat, 29 Jun 2024 23:31:31 +0900 Subject: [PATCH 05/26] =?UTF-8?q?=E3=83=95=E3=82=A3=E3=83=BC=E3=83=81?= =?UTF-8?q?=E3=83=A3=E3=82=92=E3=83=AA=E3=83=8D=E3=83=BC=E3=83=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- README.md | 2 +- crates/test_util/build.rs | 4 +- crates/voicevox_core/Cargo.toml | 10 +-- .../src/__internal/doctest_fixtures.rs | 6 +- crates/voicevox_core/src/blocking.rs | 4 +- .../src/infer/runtimes/onnxruntime.rs | 64 +++++++++---------- crates/voicevox_core/src/lib.rs | 18 +++--- crates/voicevox_core/src/tokio.rs | 4 +- crates/voicevox_core_c_api/Cargo.toml | 4 +- crates/voicevox_core_c_api/cbindgen.toml | 16 ++--- .../include/voicevox_core.h | 20 +++--- crates/voicevox_core_c_api/src/c_impls.rs | 4 +- crates/voicevox_core_c_api/src/lib.rs | 10 +-- crates/voicevox_core_c_api/tests/e2e/main.rs | 2 +- crates/voicevox_core_java_api/Cargo.toml | 2 +- crates/voicevox_core_python_api/Cargo.toml | 2 +- docs/apis/c_api/doxygen/Doxyfile | 2 +- docs/feature-options.md | 14 ++-- 18 files changed, 93 insertions(+), 95 deletions(-) diff --git a/README.md b/README.md index 7f5c05c8e..67ef5d6b2 100644 --- a/README.md +++ b/README.md @@ -150,7 +150,7 @@ model フォルダにある onnx モデルはダミーのため、ノイズの ```bash # DLLをビルド -cargo build --release -p voicevox_core_c_api --features onnxruntime-libloading +cargo build --release -p voicevox_core_c_api --features load-onnxruntime ``` DLL 用のヘッダファイルは [crates/voicevox_core_c_api/include/voicevox_core.h](https://github.com/VOICEVOX/voicevox_core/tree/main/crates/voicevox_core_c_api/include/voicevox_core.h) にあります。 diff --git a/crates/test_util/build.rs b/crates/test_util/build.rs index 30755e8b9..5df791c58 100644 --- a/crates/test_util/build.rs +++ b/crates/test_util/build.rs @@ -219,8 +219,8 @@ fn generate_c_api_rs_bindings(out_dir: &Utf8Path) -> anyhow::Result<()> { bindgen::Builder::default() .header(C_BINDINGS_PATH) .header(ADDITIONAL_C_BINDINGS_PATH) - // we test for `--feature onnxruntime-libloading` - .clang_arg("-DVOICEVOX_ONNXRUNTIME_LIBLOADING=") + // we test for `--feature load-onnxruntime` + .clang_arg("-DVOICEVOX_LINK_ONNXRUNTIME=") .parse_callbacks(Box::new(bindgen::CargoCallbacks::new())) .dynamic_library_name("CApi") .generate()? diff --git a/crates/voicevox_core/Cargo.toml b/crates/voicevox_core/Cargo.toml index 222fe1ada..782bfbb65 100644 --- a/crates/voicevox_core/Cargo.toml +++ b/crates/voicevox_core/Cargo.toml @@ -5,7 +5,7 @@ edition.workspace = true publish.workspace = true [package.metadata.docs.rs] -features = ["onnxruntime-libloading", "onnxruntime-link-dylib"] +features = ["load-onnxruntime", "link-onnxruntime"] rustdoc-args = ["--cfg", "docsrs"] [features] @@ -14,11 +14,11 @@ default = [] # ONNX Runtimeのリンク方法を決めるフィーチャ。どちらか片方を有効化しなければならず、両方の有効化は # 許容されない。 # -# `onnxruntime-libloading`の場合、ONNX Runtimeを`dlopen`/`LoadLibraryExW`で開く。 -# `onnxruntime-link-dylib`の場合はONNX Runtimeをロード時動的リンクする。またそれに伴い、 +# `load-onnxruntime`の場合、ONNX Runtimeを`dlopen`/`LoadLibraryExW`で開く。 +# `link-onnxruntime`の場合はONNX Runtimeをロード時動的リンクする。またそれに伴い、 # `struct Onnxruntime`の初期化方法も異なる。 -onnxruntime-libloading = ["voicevox-ort/load-dynamic"] -onnxruntime-link-dylib = [] +load-onnxruntime = ["voicevox-ort/load-dynamic"] +link-onnxruntime = [] cuda = ["voicevox-ort/cuda"] directml = ["voicevox-ort/directml"] diff --git a/crates/voicevox_core/src/__internal/doctest_fixtures.rs b/crates/voicevox_core/src/__internal/doctest_fixtures.rs index 1d881856f..8f45cba73 100644 --- a/crates/voicevox_core/src/__internal/doctest_fixtures.rs +++ b/crates/voicevox_core/src/__internal/doctest_fixtures.rs @@ -6,18 +6,18 @@ use crate::{AccelerationMode, InitializeOptions}; pub async fn synthesizer_with_sample_voice_model( voice_model_path: impl AsRef, - #[cfg_attr(feature = "onnxruntime-link-dylib", allow(unused_variables))] onnxruntime_dylib_path: impl Into< + #[cfg_attr(feature = "link-onnxruntime", allow(unused_variables))] onnxruntime_dylib_path: impl Into< OsString, >, open_jtalk_dic_dir: impl AsRef, ) -> anyhow::Result> { let syntesizer = crate::tokio::Synthesizer::new( - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] crate::tokio::Onnxruntime::load_once() .filename(onnxruntime_dylib_path) .exec() .await?, - #[cfg(feature = "onnxruntime-link-dylib")] + #[cfg(feature = "link-onnxruntime")] crate::tokio::Onnxruntime::init_once().await?, crate::tokio::OpenJtalk::new(open_jtalk_dic_dir).await?, &InitializeOptions { diff --git a/crates/voicevox_core/src/blocking.rs b/crates/voicevox_core/src/blocking.rs index a19a88ccc..8d0bc2129 100644 --- a/crates/voicevox_core/src/blocking.rs +++ b/crates/voicevox_core/src/blocking.rs @@ -7,7 +7,7 @@ pub use crate::{ }; pub mod onnxruntime { - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub use crate::infer::runtimes::onnxruntime::blocking::LoadOnce; } diff --git a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs index ed5b03fa3..63a6bba27 100644 --- a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs +++ b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs @@ -262,13 +262,13 @@ pub(crate) mod blocking { impl Onnxruntime { /// ONNX Runtimeのライブラリ名。 - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub const LIB_NAME: &'static str = "onnxruntime"; /// 推奨されるONNX Runtimeのバージョン。 - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub const LIB_VERSION: &'static str = ort::downloaded_version!(); /// [`LIB_NAME`]と[`LIB_VERSION`]からなる動的ライブラリのファイル名。 @@ -278,8 +278,8 @@ pub(crate) mod blocking { /// [`LIB_NAME`]: Self::LIB_NAME /// [`LIB_VERSION`]: Self::LIB_VERSION /// [`LIB_UNVERSIONED_FILENAME`]: Self::LIB_UNVERSIONED_FILENAME - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub const LIB_VERSIONED_FILENAME: &'static str = if cfg!(target_os = "linux") { const_format::concatcp!( "lib", @@ -302,8 +302,8 @@ pub(crate) mod blocking { /// [`LIB_NAME`]からなる動的ライブラリのファイル名。 /// /// [`LIB_NAME`]: Self::LIB_NAME - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub const LIB_UNVERSIONED_FILENAME: &'static str = const_format::concatcp!( std::env::consts::DLL_PREFIX, Onnxruntime::LIB_NAME, @@ -333,8 +333,8 @@ pub(crate) mod blocking { /// ONNX Runtimeをロードして初期化する。 /// /// 一度成功したら、以後は引数を無視して同じ参照を返す。 - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub fn load_once() -> LoadOnce { LoadOnce::default() } @@ -342,15 +342,15 @@ pub(crate) mod blocking { /// ONNX Runtimeを初期化する。 /// /// 一度成功したら以後は同じ参照を返す。 - #[cfg(feature = "onnxruntime-link-dylib")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-link-dylib")))] + #[cfg(feature = "link-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "link-onnxruntime")))] pub fn init_once() -> crate::Result<&'static Self> { Self::once(|| ort::try_init(None)) } #[cfg(test)] pub(crate) fn from_test_util_data() -> anyhow::Result<&'static Self> { - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] { Self::load_once() .filename(test_util::ONNXRUNTIME_DYLIB_PATH) @@ -358,7 +358,7 @@ pub(crate) mod blocking { .map_err(Into::into) } - #[cfg(feature = "onnxruntime-link-dylib")] + #[cfg(feature = "link-onnxruntime")] { Self::init_once().map_err(Into::into) } @@ -371,12 +371,12 @@ pub(crate) mod blocking { } /// [`Onnxruntime::load_once`]のビルダー。 - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] pub struct LoadOnce { filename: std::ffi::OsString, } - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] impl Default for LoadOnce { fn default() -> Self { let filename = Onnxruntime::LIB_VERSIONED_FILENAME.into(); @@ -384,7 +384,7 @@ pub(crate) mod blocking { } } - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] impl LoadOnce { /// ONNX Runtimeのfilenameを指定する。 /// @@ -445,14 +445,14 @@ pub(crate) mod tokio { impl Onnxruntime { /// ONNX Runtimeのライブラリ名。 - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] // ブロッキング版と等しいことはテストで担保 pub const LIB_NAME: &'static str = "onnxruntime"; /// 推奨されるONNX Runtimeのバージョン。 - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] // ブロッキング版と等しいことはテストで担保 pub const LIB_VERSION: &'static str = ort::downloaded_version!(); @@ -463,16 +463,16 @@ pub(crate) mod tokio { /// [`LIB_NAME`]: Self::LIB_NAME /// [`LIB_VERSION`]: Self::LIB_VERSION /// [`LIB_UNVERSIONED_FILENAME`]: Self::LIB_UNVERSIONED_FILENAME - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub const LIB_VERSIONED_FILENAME: &'static str = super::blocking::Onnxruntime::LIB_VERSIONED_FILENAME; /// [`LIB_NAME`]からなる動的ライブラリのファイル名。 /// /// [`LIB_NAME`]: Self::LIB_NAME - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub const LIB_UNVERSIONED_FILENAME: &'static str = super::blocking::Onnxruntime::LIB_UNVERSIONED_FILENAME; @@ -489,8 +489,8 @@ pub(crate) mod tokio { /// ONNX Runtimeをロードして初期化する。 /// /// 一度成功したら、以後は引数を無視して同じ参照を返す。 - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub fn load_once() -> LoadOnce { LoadOnce::default() } @@ -498,8 +498,8 @@ pub(crate) mod tokio { /// ONNX Runtimeを初期化する。 /// /// 一度成功したら以後は同じ参照を返す。 - #[cfg(feature = "onnxruntime-link-dylib")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-link-dylib")))] + #[cfg(feature = "link-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "link-onnxruntime")))] pub async fn init_once() -> crate::Result<&'static Self> { let inner = crate::task::asyncify(super::blocking::Onnxruntime::init_once).await?; Ok(Self::from_blocking(inner)) @@ -519,11 +519,11 @@ pub(crate) mod tokio { } /// [`Onnxruntime::load_once`]のビルダー。 - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] #[derive(Default)] pub struct LoadOnce(super::blocking::LoadOnce); - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] impl LoadOnce { /// ONNX Runtimeのfilenameを指定する。 /// @@ -545,7 +545,7 @@ pub(crate) mod tokio { mod tests { use rstest::rstest; - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] #[test] fn assert_same_lib_names_and_versions() { use pretty_assertions::assert_eq; diff --git a/crates/voicevox_core/src/lib.rs b/crates/voicevox_core/src/lib.rs index 5880f1f37..c7a022eb6 100644 --- a/crates/voicevox_core/src/lib.rs +++ b/crates/voicevox_core/src/lib.rs @@ -2,27 +2,25 @@ #![cfg_attr(docsrs, feature(doc_cfg))] -#[cfg(not(any(feature = "onnxruntime-libloading", feature = "onnxruntime-link-dylib")))] -compile_error!("either `onnxruntime-libloading` or `onnxruntime-link-dylib` must be enabled"); +#[cfg(not(any(feature = "load-onnxruntime", feature = "link-onnxruntime")))] +compile_error!("either `load-onnxruntime` or `link-onnxruntime` must be enabled"); #[cfg(not(doc))] const _: () = { - #[cfg(all(feature = "onnxruntime-libloading", feature = "onnxruntime-link-dylib"))] - compile_error!( - "`onnxruntime-libloading` and `onnxruntime-link-dylib` cannot be enabled at the same time", - ); + #[cfg(all(feature = "load-onnxruntime", feature = "link-onnxruntime"))] + compile_error!("`load-onnxruntime` and `link-onnxruntime` cannot be enabled at the same time"); // Rust APIでvoicevox-ortを他のクレートが利用する可能性を考え、voicevox-ort側とfeatureがズレ // ないようにする - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] ort::assert_load_dynamic_is_enabled!( - "when `onnxruntime-libloading` is enabled,`voicevox-ort/load-dynamic` must be also enabled", + "when `load-onnxruntime` is enabled,`voicevox-ort/load-dynamic` must be also enabled", ); - #[cfg(feature = "onnxruntime-link-dylib")] + #[cfg(feature = "link-onnxruntime")] ort::assert_load_dynamic_is_disabled!( - "when `onnxruntime-link-dylib` is enabled,`voicevox-ort/load-dynamic` must be disabled", + "when `link-onnxruntime` is enabled,`voicevox-ort/load-dynamic` must be disabled", ); }; diff --git a/crates/voicevox_core/src/tokio.rs b/crates/voicevox_core/src/tokio.rs index 0fdccc3c3..1e2fabada 100644 --- a/crates/voicevox_core/src/tokio.rs +++ b/crates/voicevox_core/src/tokio.rs @@ -7,7 +7,7 @@ pub use crate::{ }; pub mod onnxruntime { - #[cfg(feature = "onnxruntime-libloading")] - #[cfg_attr(docsrs, doc(cfg(feature = "onnxruntime-libloading")))] + #[cfg(feature = "load-onnxruntime")] + #[cfg_attr(docsrs, doc(cfg(feature = "load-onnxruntime")))] pub use crate::infer::runtimes::onnxruntime::tokio::LoadOnce; } diff --git a/crates/voicevox_core_c_api/Cargo.toml b/crates/voicevox_core_c_api/Cargo.toml index 4a9110849..29b66e55a 100644 --- a/crates/voicevox_core_c_api/Cargo.toml +++ b/crates/voicevox_core_c_api/Cargo.toml @@ -13,8 +13,8 @@ harness = false name = "e2e" [features] -onnxruntime-libloading = ["voicevox_core/onnxruntime-libloading"] -onnxruntime-link-dylib = ["voicevox_core/onnxruntime-link-dylib"] +load-onnxruntime = ["voicevox_core/load-onnxruntime"] +link-onnxruntime = ["voicevox_core/link-onnxruntime"] cuda = ["voicevox_core/cuda"] directml = ["voicevox_core/directml"] diff --git a/crates/voicevox_core_c_api/cbindgen.toml b/crates/voicevox_core_c_api/cbindgen.toml index 550be7ca1..0ed07d036 100644 --- a/crates/voicevox_core_c_api/cbindgen.toml +++ b/crates/voicevox_core_c_api/cbindgen.toml @@ -57,15 +57,15 @@ after_includes = """ #include #endif // __cplusplus -//#define VOICEVOX_ONNXRUNTIME_LIBLOADING -//#define VOICEVOX_ONNXRUNTIME_LINK_DYLIB +//#define VOICEVOX_LINK_ONNXRUNTIME +//#define VOICEVOX_LOAD_ONNXRUNTIME -#if !(defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) || defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB)) -#error "either `VOICEVOX_ONNXRUNTIME_LIBLOADING` or `VOICEVOX_ONNXRUNTIME_LINK_DYLIB` must be enabled" +#if !(defined(VOICEVOX_LINK_ONNXRUNTIME) || defined(VOICEVOX_LOAD_ONNXRUNTIME)) +#error "either `VOICEVOX_LINK_ONNXRUNTIME` or `VOICEVOX_LOAD_ONNXRUNTIME` must be enabled" #endif -#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) && defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB) -#error "`VOICEVOX_ONNXRUNTIME_LIBLOADING` or `VOICEVOX_ONNXRUNTIME_LINK_DYLIB` cannot be enabled at the same time" +#if defined(VOICEVOX_LINK_ONNXRUNTIME) && defined(VOICEVOX_LOAD_ONNXRUNTIME) +#error "`VOICEVOX_LINK_ONNXRUNTIME` or `VOICEVOX_LOAD_ONNXRUNTIME` cannot be enabled at the same time" #endif""" # Code Style Options @@ -91,5 +91,5 @@ parse_deps = true include = ["voicevox_core"] [defines] -"feature = onnxruntime-libloading" = "VOICEVOX_ONNXRUNTIME_LIBLOADING" -"feature = onnxruntime-link-dylib" = "VOICEVOX_ONNXRUNTIME_LINK_DYLIB" +"feature = load-onnxruntime" = "VOICEVOX_LINK_ONNXRUNTIME" +"feature = link-onnxruntime" = "VOICEVOX_LOAD_ONNXRUNTIME" diff --git a/crates/voicevox_core_c_api/include/voicevox_core.h b/crates/voicevox_core_c_api/include/voicevox_core.h index aa4f049a8..d13ca14f5 100644 --- a/crates/voicevox_core_c_api/include/voicevox_core.h +++ b/crates/voicevox_core_c_api/include/voicevox_core.h @@ -53,15 +53,15 @@ #include #endif // __cplusplus -//#define VOICEVOX_ONNXRUNTIME_LIBLOADING -//#define VOICEVOX_ONNXRUNTIME_LINK_DYLIB +//#define VOICEVOX_LINK_ONNXRUNTIME +//#define VOICEVOX_LOAD_ONNXRUNTIME -#if !(defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) || defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB)) -#error "either `VOICEVOX_ONNXRUNTIME_LIBLOADING` or `VOICEVOX_ONNXRUNTIME_LINK_DYLIB` must be enabled" +#if !(defined(VOICEVOX_LINK_ONNXRUNTIME) || defined(VOICEVOX_LOAD_ONNXRUNTIME)) +#error "either `VOICEVOX_LINK_ONNXRUNTIME` or `VOICEVOX_LOAD_ONNXRUNTIME` must be enabled" #endif -#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) && defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB) -#error "`VOICEVOX_ONNXRUNTIME_LIBLOADING` or `VOICEVOX_ONNXRUNTIME_LINK_DYLIB` cannot be enabled at the same time" +#if defined(VOICEVOX_LINK_ONNXRUNTIME) && defined(VOICEVOX_LOAD_ONNXRUNTIME) +#error "`VOICEVOX_LINK_ONNXRUNTIME` or `VOICEVOX_LOAD_ONNXRUNTIME` cannot be enabled at the same time" #endif /** @@ -289,7 +289,7 @@ typedef struct VoicevoxUserDict VoicevoxUserDict; */ typedef struct VoicevoxVoiceModel VoicevoxVoiceModel; -#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) +#if defined(VOICEVOX_LINK_ONNXRUNTIME) typedef struct VoicevoxLoadOnnxruntimeOptions { const char *filename; } VoicevoxLoadOnnxruntimeOptions; @@ -372,7 +372,7 @@ typedef struct VoicevoxUserDictWord { extern "C" { #endif // __cplusplus -#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) +#if defined(VOICEVOX_LINK_ONNXRUNTIME) #ifdef _WIN32 __declspec(dllimport) #endif @@ -391,7 +391,7 @@ __declspec(dllimport) #endif const struct VoicevoxOnnxruntime *voicevox_onnxruntime_get(void); -#if defined(VOICEVOX_ONNXRUNTIME_LIBLOADING) +#if defined(VOICEVOX_LINK_ONNXRUNTIME) /** * ONNX Runtimeをロードして初期化する。 * @@ -414,7 +414,7 @@ VoicevoxResultCode voicevox_onnxruntime_load_once(struct VoicevoxLoadOnnxruntime const struct VoicevoxOnnxruntime **out_onnxruntime); #endif -#if defined(VOICEVOX_ONNXRUNTIME_LINK_DYLIB) +#if defined(VOICEVOX_LOAD_ONNXRUNTIME) /** * ONNX Runtimeを初期化する。 * diff --git a/crates/voicevox_core_c_api/src/c_impls.rs b/crates/voicevox_core_c_api/src/c_impls.rs index 1b471a6fe..66fe6c621 100644 --- a/crates/voicevox_core_c_api/src/c_impls.rs +++ b/crates/voicevox_core_c_api/src/c_impls.rs @@ -19,7 +19,7 @@ impl VoicevoxOnnxruntime { voicevox_core::blocking::Onnxruntime::get().map(Self::new) } - #[cfg(feature = "onnxruntime-libloading")] + #[cfg(feature = "load-onnxruntime")] pub(crate) fn load_once(filename: &std::ffi::CStr) -> CApiResult<&'static Self> { use crate::helpers::ensure_utf8; @@ -29,7 +29,7 @@ impl VoicevoxOnnxruntime { Ok(Self::new(inner)) } - #[cfg(feature = "onnxruntime-link-dylib")] + #[cfg(feature = "link-onnxruntime")] pub(crate) fn init_once() -> CApiResult<&'static Self> { let inner = voicevox_core::blocking::Onnxruntime::init_once()?; Ok(Self::new(inner)) diff --git a/crates/voicevox_core_c_api/src/lib.rs b/crates/voicevox_core_c_api/src/lib.rs index dce26dbb8..a36626ad9 100644 --- a/crates/voicevox_core_c_api/src/lib.rs +++ b/crates/voicevox_core_c_api/src/lib.rs @@ -4,7 +4,7 @@ mod c_impls; /// cbindgen:ignore -#[cfg(feature = "onnxruntime-libloading")] +#[cfg(feature = "load-onnxruntime")] mod compatible_engine; mod drop_check; mod helpers; @@ -88,13 +88,13 @@ fn init_logger_once() { * voicevox_core/publish.rsにある対応する関数とはこのファイルに定義してある公開関数からvoicevoxプレフィックスを取り除いた名前の関数である */ -#[cfg(feature = "onnxruntime-libloading")] +#[cfg(feature = "load-onnxruntime")] #[repr(C)] pub struct VoicevoxLoadOnnxruntimeOptions { filename: *const c_char, } -#[cfg(feature = "onnxruntime-libloading")] +#[cfg(feature = "load-onnxruntime")] #[no_mangle] pub extern "C" fn voicevox_make_default_load_onnxruntime_options() -> VoicevoxLoadOnnxruntimeOptions { @@ -155,7 +155,7 @@ pub extern "C" fn voicevox_onnxruntime_get() -> Option<&'static VoicevoxOnnxrunt /// - `options.filename`はヌル終端文字列を指し、かつ読み込みについて有効でなければならない。 /// - `out_onnxruntime`は書き込みについて有効でなければならない。 /// } -#[cfg(feature = "onnxruntime-libloading")] +#[cfg(feature = "load-onnxruntime")] #[no_mangle] pub unsafe extern "C" fn voicevox_onnxruntime_load_once( options: VoicevoxLoadOnnxruntimeOptions, @@ -187,7 +187,7 @@ pub unsafe extern "C" fn voicevox_onnxruntime_load_once( /// \safety{ /// - `out_onnxruntime`は書き込みについて有効でなければならない。 /// } -#[cfg(feature = "onnxruntime-link-dylib")] +#[cfg(feature = "link-onnxruntime")] #[no_mangle] pub unsafe extern "C" fn voicevox_onnxruntime_init_once( out_onnxruntime: NonNull<&'static VoicevoxOnnxruntime>, diff --git a/crates/voicevox_core_c_api/tests/e2e/main.rs b/crates/voicevox_core_c_api/tests/e2e/main.rs index a826c02e9..0a7520c76 100644 --- a/crates/voicevox_core_c_api/tests/e2e/main.rs +++ b/crates/voicevox_core_c_api/tests/e2e/main.rs @@ -22,7 +22,7 @@ fn main() -> anyhow::Result<()> { enum TestContext {} impl assert_cdylib::TestContext for TestContext { - const FEATURES: &'static [&'static str] = &["onnxruntime-libloading"]; + const FEATURES: &'static [&'static str] = &["load-onnxruntime"]; const TARGET_DIR: &'static str = "../../target"; const CDYLIB_NAME: &'static str = "voicevox_core"; const RUNTIME_ENVS: &'static [(&'static str, &'static str)] = diff --git a/crates/voicevox_core_java_api/Cargo.toml b/crates/voicevox_core_java_api/Cargo.toml index fd785e17b..e9cced8b2 100644 --- a/crates/voicevox_core_java_api/Cargo.toml +++ b/crates/voicevox_core_java_api/Cargo.toml @@ -25,7 +25,7 @@ serde_json = { workspace = true, features = ["preserve_order"] } tracing = { workspace = true, features = ["log"] } tracing-subscriber = { workspace = true, features = ["env-filter"] } uuid.workspace = true -voicevox_core = { workspace = true, features = ["onnxruntime-libloading"] } +voicevox_core = { workspace = true, features = ["load-onnxruntime"] } [lints.rust] unsafe_code = "allow" # jni-rsが要求 diff --git a/crates/voicevox_core_python_api/Cargo.toml b/crates/voicevox_core_python_api/Cargo.toml index cc83d583e..48c92dfb4 100644 --- a/crates/voicevox_core_python_api/Cargo.toml +++ b/crates/voicevox_core_python_api/Cargo.toml @@ -23,7 +23,7 @@ serde = { workspace = true, features = ["derive"] } serde_json.workspace = true tracing = { workspace = true, features = ["log"] } uuid.workspace = true -voicevox_core = { workspace = true, features = ["onnxruntime-libloading"] } +voicevox_core = { workspace = true, features = ["load-onnxruntime"] } [lints.rust] unsafe_code = "forbid" diff --git a/docs/apis/c_api/doxygen/Doxyfile b/docs/apis/c_api/doxygen/Doxyfile index cd3313463..f9603fbd6 100644 --- a/docs/apis/c_api/doxygen/Doxyfile +++ b/docs/apis/c_api/doxygen/Doxyfile @@ -2257,7 +2257,7 @@ INCLUDE_FILE_PATTERNS = # recursively expanded use the := operator instead of the = operator. # This tag requires that the tag ENABLE_PREPROCESSING is set to YES. -PREDEFINED = VOICEVOX_ONNXRUNTIME_LIBLOADING= VOICEVOX_ONNXRUNTIME_LINK_DYLIB= +PREDEFINED = VOICEVOX_LOAD_ONNXRUNTIME= VOICEVOX_LINK_ONNXRUNTIME= # If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then this # tag can be used to specify a list of macro names that should be expanded. The diff --git a/docs/feature-options.md b/docs/feature-options.md index cae419ac5..c84c6246b 100644 --- a/docs/feature-options.md +++ b/docs/feature-options.md @@ -3,18 +3,18 @@ Rust API（`voicevox_core`）およびC API（`voicevox_core_c_api`）においては、ビルド時に 次のCargoフィーチャのうちどちらかを選択しなければなりません。 詳しくは[voicevox_core/Cargo.toml](../crates/voicevox_core/Cargo.toml)のコメントを参照して -下さい。Python APIやJava APIでは`onnxruntime-libloading`のみに限定しています。 +下さい。Python APIやJava APIでは`load-onnxruntime`のみに限定しています。 -- `onnxruntime-libloading` -- `onnxruntime-link-dylib` +- `load-onnxruntime` +- `link-onnxruntime` ```console -❯ cargo build --release -p voicevox_core_c_api --features onnxruntime-libloading +❯ cargo build --release -p voicevox_core_c_api --features load-onnxruntime ``` ```console -❯ cargo build --release -p voicevox_core_c_api --features onnxruntime-link-dylib +❯ cargo build --release -p voicevox_core_c_api --features link-onnxruntime ``` -C APIのリリースでは`dlopen`の利用が厳しいiOSでのみ`onnxruntime-link-dylib`で、その他 -は`onnxruntime-libloading`でビルドしています。 +C APIのリリースでは`dlopen`の利用が厳しいiOSでのみ`link-onnxruntime`で、その他 +は`load-onnxruntime`でビルドしています。 From f0c720cd1f231cb420a25d4f797ada2ee849d4a6 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 00:07:08 +0900 Subject: [PATCH 06/26] =?UTF-8?q?fixup!=20=E3=83=95=E3=82=A3=E3=83=BC?= =?UTF-8?q?=E3=83=81=E3=83=A3=E3=82=92=E3=83=AA=E3=83=8D=E3=83=BC=E3=83=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .github/workflows/build_and_deploy.yml | 8 ++++---- .github/workflows/test.yml | 8 ++++---- 2 files changed, 8 insertions(+), 8 deletions(-) diff --git a/.github/workflows/build_and_deploy.yml b/.github/workflows/build_and_deploy.yml index fce217748..47a35068f 100644 --- a/.github/workflows/build_and_deploy.yml +++ b/.github/workflows/build_and_deploy.yml @@ -266,8 +266,8 @@ jobs: shell: bash run: | case ${{ matrix.c_release_format }} in - plain-cdylib) linking=onnxruntime-libloading ;; - ios-xcframework) linking=onnxruntime-link-dylib ;; + plain-cdylib) linking=load-onnxruntime ;; + ios-xcframework) linking=link-onnxruntime ;; esac function build() { cargo build -p voicevox_core_c_api -vv --features "$linking",${{ matrix.features }} --target ${{ matrix.target }} --release @@ -311,8 +311,8 @@ jobs: run: | mkdir -p "artifact/${{ env.ASSET_NAME }}" case ${{ matrix.c_release_format }} in - plain-cdylib) feature=VOICEVOX_ONNXRUNTIME_LIBLOADING ;; - ios-xcframework) feature=VOICEVOX_ONNXRUNTIME_LINK_DYLIB ;; + plain-cdylib) feature=VOICEVOX_LOAD_ONNXRUNTIME ;; + ios-xcframework) feature=VOICEVOX_LINK_ONNXRUNTIME ;; esac sed 's:^//\(#define VOICEVOX_'"$feature"'\)$:\1:' crates/voicevox_core_c_api/include/voicevox_core.h \ > "artifact/${{ env.ASSET_NAME }}/voicevox_core.h" diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml index c87fd727e..ccd1dbd5c 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test.yml @@ -74,8 +74,8 @@ jobs: - uses: Swatinem/rust-cache@v2 - run: cargo clippy -vv --features directml,cuda --tests -- -D clippy::all -D warnings --no-deps - run: cargo clippy -vv --features directml,cuda -- -D clippy::all -D warnings --no-deps - - run: cargo clippy -vv -p voicevox_core -p voicevox_core_c_api --features onnxruntime-link-dylib,directml,cuda --tests -- -D clippy::all -D warnings --no-deps - - run: cargo clippy -vv -p voicevox_core -p voicevox_core_c_api --features onnxruntime-link-dylib,directml,cuda -- -D clippy::all -D warnings --no-deps + - run: cargo clippy -vv -p voicevox_core -p voicevox_core_c_api --features link-onnxruntime,directml,cuda --tests -- -D clippy::all -D warnings --no-deps + - run: cargo clippy -vv -p voicevox_core -p voicevox_core_c_api --features link-onnxruntime,directml,cuda -- -D clippy::all -D warnings --no-deps - run: cargo fmt -- --check rust-unit-test: @@ -195,7 +195,7 @@ jobs: - name: Install cargo-binstall uses: taiki-e/install-action@cargo-binstall - name: build voicevox_core_c_api - run: cargo build -p voicevox_core_c_api --features onnxruntime-libloading -vv + run: cargo build -p voicevox_core_c_api --features load-onnxruntime -vv - name: 必要なfileをunix用exampleのディレクトリに移動させる run: | mkdir -p example/cpp/unix/voicevox_core/ @@ -239,7 +239,7 @@ jobs: - name: Install cargo-binstall uses: taiki-e/install-action@cargo-binstall - name: build voicevox_core_c_api - run: cargo build -p voicevox_core_c_api --features onnxruntime-libloading -vv + run: cargo build -p voicevox_core_c_api --features load-onnxruntime -vv - name: 必要なfileをexampleのディレクトリに移動させる shell: bash run: | From 4af36e0003bd621fc525c021ef425f597a879992 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 00:17:31 +0900 Subject: [PATCH 07/26] =?UTF-8?q?fixup!=20=E3=83=95=E3=82=A3=E3=83=BC?= =?UTF-8?q?=E3=83=81=E3=83=A3=E3=82=92=E3=83=AA=E3=83=8D=E3=83=BC=E3=83=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .github/workflows/test.yml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml index ccd1dbd5c..f66fb8a5e 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test.yml @@ -199,7 +199,7 @@ jobs: - name: 必要なfileをunix用exampleのディレクトリに移動させる run: | mkdir -p example/cpp/unix/voicevox_core/ - sed 's:^//\(#define VOICEVOX_ONNXRUNTIME_LIBLOADING\)$:\1:' \ + sed 's:^//\(#define VOICEVOX_LOAD_ONNXRUNTIME\)$:\1:' \ crates/voicevox_core_c_api/include/voicevox_core.h \ > example/cpp/unix/voicevox_core/voicevox_core.h cp -v target/debug/libvoicevox_core.{so,dylib} example/cpp/unix/voicevox_core/ || true @@ -244,7 +244,7 @@ jobs: shell: bash run: | mkdir -p example/cpp/windows/simple_tts/lib/x64 - sed 's:^//\(#define VOICEVOX_ONNXRUNTIME_LIBLOADING\)$:\1:' \ + sed 's:^//\(#define VOICEVOX_LOAD_ONNXRUNTIME\)$:\1:' \ crates/voicevox_core_c_api/include/voicevox_core.h \ > example/cpp/windows/simple_tts/voicevox_core.h cp target/debug/voicevox_core.dll.lib example/cpp/windows/simple_tts/lib/x64/voicevox_core.lib From 249f8f327376194f7883be680cf578b16d4ba964 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 00:25:13 +0900 Subject: [PATCH 08/26] =?UTF-8?q?fixup!=20=E3=83=95=E3=82=A3=E3=83=BC?= =?UTF-8?q?=E3=83=81=E3=83=A3=E3=82=92=E3=83=AA=E3=83=8D=E3=83=BC=E3=83=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- crates/voicevox_core_c_api/cbindgen.toml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/crates/voicevox_core_c_api/cbindgen.toml b/crates/voicevox_core_c_api/cbindgen.toml index 0ed07d036..28aee5706 100644 --- a/crates/voicevox_core_c_api/cbindgen.toml +++ b/crates/voicevox_core_c_api/cbindgen.toml @@ -91,5 +91,5 @@ parse_deps = true include = ["voicevox_core"] [defines] -"feature = load-onnxruntime" = "VOICEVOX_LINK_ONNXRUNTIME" -"feature = link-onnxruntime" = "VOICEVOX_LOAD_ONNXRUNTIME" +"feature = load-onnxruntime" = "VOICEVOX_LOAD_ONNXRUNTIME" +"feature = link-onnxruntime" = "VOICEVOX_LINK_ONNXRUNTIME" From 7b572686b72813a7a7d020ba72eddb49c3b1729a Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 00:26:09 +0900 Subject: [PATCH 09/26] =?UTF-8?q?fixup!=20=E3=83=95=E3=82=A3=E3=83=BC?= =?UTF-8?q?=E3=83=81=E3=83=A3=E3=82=92=E3=83=AA=E3=83=8D=E3=83=BC=E3=83=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- crates/voicevox_core_c_api/include/voicevox_core.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/crates/voicevox_core_c_api/include/voicevox_core.h b/crates/voicevox_core_c_api/include/voicevox_core.h index d13ca14f5..07d58b036 100644 --- a/crates/voicevox_core_c_api/include/voicevox_core.h +++ b/crates/voicevox_core_c_api/include/voicevox_core.h @@ -289,7 +289,7 @@ typedef struct VoicevoxUserDict VoicevoxUserDict; */ typedef struct VoicevoxVoiceModel VoicevoxVoiceModel; -#if defined(VOICEVOX_LINK_ONNXRUNTIME) +#if defined(VOICEVOX_LOAD_ONNXRUNTIME) typedef struct VoicevoxLoadOnnxruntimeOptions { const char *filename; } VoicevoxLoadOnnxruntimeOptions; @@ -372,7 +372,7 @@ typedef struct VoicevoxUserDictWord { extern "C" { #endif // __cplusplus -#if defined(VOICEVOX_LINK_ONNXRUNTIME) +#if defined(VOICEVOX_LOAD_ONNXRUNTIME) #ifdef _WIN32 __declspec(dllimport) #endif @@ -391,7 +391,7 @@ __declspec(dllimport) #endif const struct VoicevoxOnnxruntime *voicevox_onnxruntime_get(void); -#if defined(VOICEVOX_LINK_ONNXRUNTIME) +#if defined(VOICEVOX_LOAD_ONNXRUNTIME) /** * ONNX Runtimeをロードして初期化する。 * @@ -414,7 +414,7 @@ VoicevoxResultCode voicevox_onnxruntime_load_once(struct VoicevoxLoadOnnxruntime const struct VoicevoxOnnxruntime **out_onnxruntime); #endif -#if defined(VOICEVOX_LOAD_ONNXRUNTIME) +#if defined(VOICEVOX_LINK_ONNXRUNTIME) /** * ONNX Runtimeを初期化する。 * From e8dc442f0d12af491eb23de14bcc98d006d3561d Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 00:35:09 +0900 Subject: [PATCH 10/26] =?UTF-8?q?fixup!=20=E3=83=95=E3=82=A3=E3=83=BC?= =?UTF-8?q?=E3=83=81=E3=83=A3=E3=82=92=E3=83=AA=E3=83=8D=E3=83=BC=E3=83=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- crates/test_util/build.rs | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/crates/test_util/build.rs b/crates/test_util/build.rs index 5df791c58..3cdc88d78 100644 --- a/crates/test_util/build.rs +++ b/crates/test_util/build.rs @@ -220,7 +220,7 @@ fn generate_c_api_rs_bindings(out_dir: &Utf8Path) -> anyhow::Result<()> { .header(C_BINDINGS_PATH) .header(ADDITIONAL_C_BINDINGS_PATH) // we test for `--feature load-onnxruntime` - .clang_arg("-DVOICEVOX_LINK_ONNXRUNTIME=") + .clang_arg("-DVOICEVOX_LOAD_ONNXRUNTIME=") .parse_callbacks(Box::new(bindgen::CargoCallbacks::new())) .dynamic_library_name("CApi") .generate()? From 9cfc7b175190687f6acdcb737e9541dd9edad324 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 02:15:06 +0900 Subject: [PATCH 11/26] =?UTF-8?q?voicevox-ort=E3=82=92=E6=9B=B4=E6=96=B0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- Cargo.lock | 4 ++-- Cargo.toml | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/Cargo.lock b/Cargo.lock index 9875b69db..e54fab37a 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -3623,7 +3623,7 @@ checksum = "49874b5167b65d7193b8aba1567f5c7d93d001cafc34600cee003eda787e483f" [[package]] name = "voicevox-ort" version = "2.0.0-rc.2" -source = "git+https://github.com/qryxip/ort.git?rev=7642ebc93831e65afa114a19ebd70f9c09dee2f8#7642ebc93831e65afa114a19ebd70f9c09dee2f8" +source = "git+https://github.com/qryxip/ort.git?rev=82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d#82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d" dependencies = [ "anyhow", "half", @@ -3640,7 +3640,7 @@ dependencies = [ [[package]] name = "voicevox-ort-sys" version = "2.0.0-rc.2" -source = "git+https://github.com/qryxip/ort.git?rev=7642ebc93831e65afa114a19ebd70f9c09dee2f8#7642ebc93831e65afa114a19ebd70f9c09dee2f8" +source = "git+https://github.com/qryxip/ort.git?rev=82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d#82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d" dependencies = [ "flate2", "sha2", diff --git a/Cargo.toml b/Cargo.toml index 39043e884..2729dcff7 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -90,7 +90,7 @@ zip = "0.6.3" [workspace.dependencies.voicevox-ort] git = "https://github.com/qryxip/ort.git" -rev = "7642ebc93831e65afa114a19ebd70f9c09dee2f8" +rev = "82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d" [workspace.dependencies.open_jtalk] git = "https://github.com/VOICEVOX/open_jtalk-rs.git" From c18e3356de1ba94a97ce95629c6f97240010c6f4 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 02:26:43 +0900 Subject: [PATCH 12/26] =?UTF-8?q?`sed`=E3=81=AE=E3=81=97=E6=90=8D=E3=81=AD?= =?UTF-8?q?=E3=82=92=E4=BF=AE=E6=AD=A3?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .github/workflows/build_and_deploy.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/workflows/build_and_deploy.yml b/.github/workflows/build_and_deploy.yml index 47a35068f..e9b5e9eb5 100644 --- a/.github/workflows/build_and_deploy.yml +++ b/.github/workflows/build_and_deploy.yml @@ -314,7 +314,7 @@ jobs: plain-cdylib) feature=VOICEVOX_LOAD_ONNXRUNTIME ;; ios-xcframework) feature=VOICEVOX_LINK_ONNXRUNTIME ;; esac - sed 's:^//\(#define VOICEVOX_'"$feature"'\)$:\1:' crates/voicevox_core_c_api/include/voicevox_core.h \ + sed 's:^//\(#define '"$feature"'\)$:\1:' crates/voicevox_core_c_api/include/voicevox_core.h \ > "artifact/${{ env.ASSET_NAME }}/voicevox_core.h" cp -v target/${{ matrix.target }}/release/*voicevox_core.{dll,so,dylib} "artifact/${{ env.ASSET_NAME }}" || true cp -v target/${{ matrix.target }}/release/voicevox_core.dll.lib "artifact/${{ env.ASSET_NAME }}/voicevox_core.lib" || true From 1492995aa6c3684ab086479aae0976282ccf1b4b Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 03:00:13 +0900 Subject: [PATCH 13/26] =?UTF-8?q?=E3=83=89=E3=82=AD=E3=83=A5=E3=83=A1?= =?UTF-8?q?=E3=83=B3=E3=83=88=E3=82=84=E3=82=B3=E3=83=A1=E3=83=B3=E3=83=88?= =?UTF-8?q?=E3=82=92=E6=9B=B4=E6=96=B0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- crates/voicevox_core/Cargo.toml | 5 ++--- .../include/voicevox_core.h | 21 +++++++++++++++---- crates/voicevox_core_c_api/src/lib.rs | 15 +++++++++---- .../python/voicevox_core/_rust/asyncio.pyi | 2 +- .../python/voicevox_core/_rust/blocking.pyi | 2 +- docs/feature-options.md | 4 ++-- 6 files changed, 34 insertions(+), 15 deletions(-) diff --git a/crates/voicevox_core/Cargo.toml b/crates/voicevox_core/Cargo.toml index 782bfbb65..5c04b3009 100644 --- a/crates/voicevox_core/Cargo.toml +++ b/crates/voicevox_core/Cargo.toml @@ -14,9 +14,8 @@ default = [] # ONNX Runtimeのリンク方法を決めるフィーチャ。どちらか片方を有効化しなければならず、両方の有効化は # 許容されない。 # -# `load-onnxruntime`の場合、ONNX Runtimeを`dlopen`/`LoadLibraryExW`で開く。 -# `link-onnxruntime`の場合はONNX Runtimeをロード時動的リンクする。またそれに伴い、 -# `struct Onnxruntime`の初期化方法も異なる。 +# `load-onnxruntime`の場合、ONNX Runtimeを`dlopen`/`LoadLibraryExW`で開く。`link-onnxruntime`の場合 +# はONNX Runtimeをロード時動的リンクする。またそれに伴い、`struct Onnxruntime`の初期化方法も異なる。 load-onnxruntime = ["voicevox-ort/load-dynamic"] link-onnxruntime = [] diff --git a/crates/voicevox_core_c_api/include/voicevox_core.h b/crates/voicevox_core_c_api/include/voicevox_core.h index 07d58b036..b3e47b7b2 100644 --- a/crates/voicevox_core_c_api/include/voicevox_core.h +++ b/crates/voicevox_core_c_api/include/voicevox_core.h @@ -290,7 +290,15 @@ typedef struct VoicevoxUserDict VoicevoxUserDict; typedef struct VoicevoxVoiceModel VoicevoxVoiceModel; #if defined(VOICEVOX_LOAD_ONNXRUNTIME) +/** + * ::voicevox_onnxruntime_load_once のオプション。 + */ typedef struct VoicevoxLoadOnnxruntimeOptions { + /** + * ONNX Runtimeのfilenameを指定する。 + * + * `dlopen`/`LoadLibraryExW`の引数に使われる。 + */ const char *filename; } VoicevoxLoadOnnxruntimeOptions; #endif @@ -373,6 +381,11 @@ extern "C" { #endif // __cplusplus #if defined(VOICEVOX_LOAD_ONNXRUNTIME) +/** + * デフォルトの ::voicevox_onnxruntime_load_once のオプションを生成する。 + * + * @return デフォルトの ::voicevox_onnxruntime_load_once のオプション + */ #ifdef _WIN32 __declspec(dllimport) #endif @@ -380,9 +393,9 @@ struct VoicevoxLoadOnnxruntimeOptions voicevox_make_default_load_onnxruntime_opt #endif /** - * ONNX Runtimeを初期化済みであれば、 ::VoicevoxOnnxruntime のインスタンスを得る。 + * ::VoicevoxOnnxruntime のインスタンスが既に作られているならそれを得る。 * - * 初期化していなければ`NULL`を返す。 + * 作られていなければ`NULL`を返す。 * * @returns ::VoicevoxOnnxruntime のインスタンス */ @@ -395,7 +408,7 @@ const struct VoicevoxOnnxruntime *voicevox_onnxruntime_get(void); /** * ONNX Runtimeをロードして初期化する。 * - * 二度目以降はパラメータに関わらず既存のインスタンスを返す。 + * 一度成功したら、以後は引数を無視して同じ参照を返す。 * * @param [in] options オプション * @param [out] out_onnxruntime ::VoicevoxOnnxruntime のインスタンス @@ -418,7 +431,7 @@ VoicevoxResultCode voicevox_onnxruntime_load_once(struct VoicevoxLoadOnnxruntime /** * ONNX Runtimeを初期化する。 * - * 二度目以降は既存のインスタンスを返す。 + * 一度成功したら以後は同じ参照を返す。 * * @param [out] out_onnxruntime ::VoicevoxOnnxruntime のインスタンス * diff --git a/crates/voicevox_core_c_api/src/lib.rs b/crates/voicevox_core_c_api/src/lib.rs index a36626ad9..071e70d5b 100644 --- a/crates/voicevox_core_c_api/src/lib.rs +++ b/crates/voicevox_core_c_api/src/lib.rs @@ -88,12 +88,19 @@ fn init_logger_once() { * voicevox_core/publish.rsにある対応する関数とはこのファイルに定義してある公開関数からvoicevoxプレフィックスを取り除いた名前の関数である */ +/// ::voicevox_onnxruntime_load_once のオプション。 #[cfg(feature = "load-onnxruntime")] #[repr(C)] pub struct VoicevoxLoadOnnxruntimeOptions { + /// ONNX Runtimeのfilenameを指定する。 + /// + /// `dlopen`/`LoadLibraryExW`の引数に使われる。 filename: *const c_char, } +/// デフォルトの ::voicevox_onnxruntime_load_once のオプションを生成する。 +/// +/// @return デフォルトの ::voicevox_onnxruntime_load_once のオプション #[cfg(feature = "load-onnxruntime")] #[no_mangle] pub extern "C" fn voicevox_make_default_load_onnxruntime_options() -> VoicevoxLoadOnnxruntimeOptions @@ -132,9 +139,9 @@ pub struct VoicevoxOnnxruntime(!); #[repr(transparent)] pub struct VoicevoxOnnxruntime(voicevox_core::blocking::Onnxruntime); -/// ONNX Runtimeを初期化済みであれば、 ::VoicevoxOnnxruntime のインスタンスを得る。 +/// ::VoicevoxOnnxruntime のインスタンスが既に作られているならそれを得る。 /// -/// 初期化していなければ`NULL`を返す。 +/// 作られていなければ`NULL`を返す。 /// /// @returns ::VoicevoxOnnxruntime のインスタンス #[no_mangle] @@ -144,7 +151,7 @@ pub extern "C" fn voicevox_onnxruntime_get() -> Option<&'static VoicevoxOnnxrunt /// ONNX Runtimeをロードして初期化する。 /// -/// 二度目以降はパラメータに関わらず既存のインスタンスを返す。 +/// 一度成功したら、以後は引数を無視して同じ参照を返す。 /// /// @param [in] options オプション /// @param [out] out_onnxruntime ::VoicevoxOnnxruntime のインスタンス @@ -178,7 +185,7 @@ pub unsafe extern "C" fn voicevox_onnxruntime_load_once( /// ONNX Runtimeを初期化する。 /// -/// 二度目以降は既存のインスタンスを返す。 +/// 一度成功したら以後は同じ参照を返す。 /// /// @param [out] out_onnxruntime ::VoicevoxOnnxruntime のインスタンス /// diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi b/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi index 6066a35aa..9c47127ec 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi +++ b/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi @@ -80,7 +80,7 @@ class Onnxruntime: """ インスタンスが既に作られているならそれを得る。 - 作られていなければ `None` を返す。 + 作られていなければ ``None`` を返す。 """ ... @staticmethod diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi b/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi index 5f96c38bb..cfdf73f7f 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi +++ b/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi @@ -80,7 +80,7 @@ class Onnxruntime: """ インスタンスが既に作られているならそれを得る。 - 作られていなければ `None` を返す。 + 作られていなければ ``None`` を返す。 """ ... @staticmethod diff --git a/docs/feature-options.md b/docs/feature-options.md index c84c6246b..e861446dd 100644 --- a/docs/feature-options.md +++ b/docs/feature-options.md @@ -16,5 +16,5 @@ Rust API（`voicevox_core`）およびC API（`voicevox_core_c_api`）におい ❯ cargo build --release -p voicevox_core_c_api --features link-onnxruntime ``` -C APIのリリースでは`dlopen`の利用が厳しいiOSでのみ`link-onnxruntime`で、その他 -は`load-onnxruntime`でビルドしています。 +C APIのリリースでは`dlopen`の利用が厳しいiOSでのみ`link-onnxruntime`で、その他は`load-onnxruntime`で +ビルドしています。 From e619d8b532e857cd15acf26bd66b29e67cd94d36 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 03:38:38 +0900 Subject: [PATCH 14/26] =?UTF-8?q?C=20API=E3=81=AB=E5=AE=9A=E6=95=B0?= =?UTF-8?q?=E3=81=AEgetter=E3=82=92=E8=BF=BD=E5=8A=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../include/voicevox_core.h | 22 ++++++++++++ crates/voicevox_core_c_api/src/c_impls.rs | 23 ++++++++++++ crates/voicevox_core_c_api/src/lib.rs | 35 ++++++++++++++----- 3 files changed, 72 insertions(+), 8 deletions(-) diff --git a/crates/voicevox_core_c_api/include/voicevox_core.h b/crates/voicevox_core_c_api/include/voicevox_core.h index b3e47b7b2..5a31eb728 100644 --- a/crates/voicevox_core_c_api/include/voicevox_core.h +++ b/crates/voicevox_core_c_api/include/voicevox_core.h @@ -380,6 +380,28 @@ typedef struct VoicevoxUserDictWord { extern "C" { #endif // __cplusplus +#if defined(VOICEVOX_LOAD_ONNXRUNTIME) +/** + * ONNX Runtimeの動的ライブラリの、バージョン付きのファイル名。 + * + * WindowsとAndroidでは ::voicevox_get_onnxruntime_lib_unversioned_filename と同じ。 + */ +#ifdef _WIN32 +__declspec(dllimport) +#endif +const char *voicevox_get_onnxruntime_lib_versioned_filename(void); +#endif + +#if defined(VOICEVOX_LOAD_ONNXRUNTIME) +/** + * ONNX Runtimeの動的ライブラリの、バージョン無しのファイル名。 + */ +#ifdef _WIN32 +__declspec(dllimport) +#endif +const char *voicevox_get_onnxruntime_lib_unversioned_filename(void); +#endif + #if defined(VOICEVOX_LOAD_ONNXRUNTIME) /** * デフォルトの ::voicevox_onnxruntime_load_once のオプションを生成する。 diff --git a/crates/voicevox_core_c_api/src/c_impls.rs b/crates/voicevox_core_c_api/src/c_impls.rs index 66fe6c621..1adc402cd 100644 --- a/crates/voicevox_core_c_api/src/c_impls.rs +++ b/crates/voicevox_core_c_api/src/c_impls.rs @@ -12,6 +12,16 @@ use crate::{ // 欠いている impl VoicevoxOnnxruntime { + #[cfg(feature = "load-onnxruntime")] + pub(crate) fn lib_versioned_filename() -> &'static std::ffi::CStr { + to_cstr!(voicevox_core::blocking::Onnxruntime::LIB_VERSIONED_FILENAME) + } + + #[cfg(feature = "load-onnxruntime")] + pub(crate) fn lib_unversioned_filename() -> &'static std::ffi::CStr { + to_cstr!(voicevox_core::blocking::Onnxruntime::LIB_UNVERSIONED_FILENAME) + } + #[ref_cast_custom] fn new(rust: &voicevox_core::blocking::Onnxruntime) -> &Self; @@ -36,6 +46,19 @@ impl VoicevoxOnnxruntime { } } +#[cfg(feature = "load-onnxruntime")] +macro_rules! to_cstr { + ($s:expr) => {{ + const __RUST_STR: &str = $s; + static __C_STR: &[u8] = const_format::concatcp!(__RUST_STR, '\0').as_bytes(); + + std::ffi::CStr::from_bytes_with_nul(__C_STR) + .unwrap_or_else(|e| panic!("{__RUST_STR:?} should not contain `\\0`: {e}")) + }}; +} +#[cfg(feature = "load-onnxruntime")] +use to_cstr; + impl OpenJtalkRc { pub(crate) fn new(open_jtalk_dic_dir: impl AsRef) -> Result { Ok(Self { diff --git a/crates/voicevox_core_c_api/src/lib.rs b/crates/voicevox_core_c_api/src/lib.rs index 071e70d5b..bbedc0dbf 100644 --- a/crates/voicevox_core_c_api/src/lib.rs +++ b/crates/voicevox_core_c_api/src/lib.rs @@ -88,6 +88,32 @@ fn init_logger_once() { * voicevox_core/publish.rsにある対応する関数とはこのファイルに定義してある公開関数からvoicevoxプレフィックスを取り除いた名前の関数である */ +// TODO: https://github.com/mozilla/cbindgen/issues/927 +//#[cfg(feature = "load-onnxruntime")] +//pub const VOICEVOX_ONNXRUNTIME_LIB_NAME: &CStr = ..; +//#[cfg(feature = "load-onnxruntime")] +//pub const VOICEVOX_ONNXRUNTIME_LIB_VERSION: &CStr = ..; + +/// ONNX Runtimeの動的ライブラリの、バージョン付きのファイル名。 +/// +/// WindowsとAndroidでは ::voicevox_get_onnxruntime_lib_unversioned_filename と同じ。 +#[cfg(feature = "load-onnxruntime")] +#[no_mangle] +pub extern "C" fn voicevox_get_onnxruntime_lib_versioned_filename() -> *const c_char { + init_logger_once(); + let filename = VoicevoxOnnxruntime::lib_versioned_filename(); + C_STRING_DROP_CHECKER.blacklist(filename).as_ptr() +} + +/// ONNX Runtimeの動的ライブラリの、バージョン無しのファイル名。 +#[cfg(feature = "load-onnxruntime")] +#[no_mangle] +pub extern "C" fn voicevox_get_onnxruntime_lib_unversioned_filename() -> *const c_char { + init_logger_once(); + let filename = VoicevoxOnnxruntime::lib_unversioned_filename(); + C_STRING_DROP_CHECKER.blacklist(filename).as_ptr() +} + /// ::voicevox_onnxruntime_load_once のオプション。 #[cfg(feature = "load-onnxruntime")] #[repr(C)] @@ -105,15 +131,8 @@ pub struct VoicevoxLoadOnnxruntimeOptions { #[no_mangle] pub extern "C" fn voicevox_make_default_load_onnxruntime_options() -> VoicevoxLoadOnnxruntimeOptions { - use const_format::concatcp; - - const RUST_FILENAME: &str = voicevox_core::blocking::Onnxruntime::LIB_VERSIONED_FILENAME; - static C_FILENAME: &[u8] = concatcp!(RUST_FILENAME, "\0").as_bytes(); - init_logger_once(); - - let filename = CStr::from_bytes_with_nul(C_FILENAME) - .unwrap_or_else(|e| panic!("{RUST_FILENAME:?} should not contain `\\0`: {e}")); + let filename = VoicevoxOnnxruntime::lib_versioned_filename(); let filename = C_STRING_DROP_CHECKER.blacklist(filename).as_ptr(); VoicevoxLoadOnnxruntimeOptions { filename } } From 29eb302f02d1905929e117b9c111b3aa9b8da61d Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 16:00:33 +0900 Subject: [PATCH 15/26] =?UTF-8?q?voicevox-ort=E3=82=92=E6=9B=B4=E6=96=B0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- Cargo.lock | 4 ++-- Cargo.toml | 2 +- crates/voicevox_core/src/lib.rs | 6 ++++-- 3 files changed, 7 insertions(+), 5 deletions(-) diff --git a/Cargo.lock b/Cargo.lock index e54fab37a..2d08d5cd5 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -3623,7 +3623,7 @@ checksum = "49874b5167b65d7193b8aba1567f5c7d93d001cafc34600cee003eda787e483f" [[package]] name = "voicevox-ort" version = "2.0.0-rc.2" -source = "git+https://github.com/qryxip/ort.git?rev=82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d#82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d" +source = "git+https://github.com/qryxip/ort.git?rev=2ab092ec0e874932bf579b51d28aa389f2fa8ea9#2ab092ec0e874932bf579b51d28aa389f2fa8ea9" dependencies = [ "anyhow", "half", @@ -3640,7 +3640,7 @@ dependencies = [ [[package]] name = "voicevox-ort-sys" version = "2.0.0-rc.2" -source = "git+https://github.com/qryxip/ort.git?rev=82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d#82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d" +source = "git+https://github.com/qryxip/ort.git?rev=2ab092ec0e874932bf579b51d28aa389f2fa8ea9#2ab092ec0e874932bf579b51d28aa389f2fa8ea9" dependencies = [ "flate2", "sha2", diff --git a/Cargo.toml b/Cargo.toml index 2729dcff7..b79872321 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -90,7 +90,7 @@ zip = "0.6.3" [workspace.dependencies.voicevox-ort] git = "https://github.com/qryxip/ort.git" -rev = "82c348b22bbb0bf5f3a264a210ec3ce2ee80aa9d" +rev = "2ab092ec0e874932bf579b51d28aa389f2fa8ea9" [workspace.dependencies.open_jtalk] git = "https://github.com/VOICEVOX/open_jtalk-rs.git" diff --git a/crates/voicevox_core/src/lib.rs b/crates/voicevox_core/src/lib.rs index c7a022eb6..3dbed0756 100644 --- a/crates/voicevox_core/src/lib.rs +++ b/crates/voicevox_core/src/lib.rs @@ -14,12 +14,14 @@ const _: () = { // ないようにする #[cfg(feature = "load-onnxruntime")] - ort::assert_load_dynamic_is_enabled!( + ort::assert_feature!( + cfg(feature = "load-dynamic"), "when `load-onnxruntime` is enabled,`voicevox-ort/load-dynamic` must be also enabled", ); #[cfg(feature = "link-onnxruntime")] - ort::assert_load_dynamic_is_disabled!( + ort::assert_feature!( + cfg(not(feature = "load-dynamic")), "when `link-onnxruntime` is enabled,`voicevox-ort/load-dynamic` must be disabled", ); }; From ae1c71add0f9aebf7ac675ab8f589f5e797d11c0 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Sun, 30 Jun 2024 17:14:14 +0900 Subject: [PATCH 16/26] =?UTF-8?q?"filename"=E3=82=92"=E3=83=95=E3=82=A1?= =?UTF-8?q?=E3=82=A4=E3=83=AB=E5=90=8D=EF=BC=88=E3=83=A2=E3=82=B8=E3=83=A5?= =?UTF-8?q?=E3=83=BC=E3=83=AB=E5=90=8D=EF=BC=89=E3=82=82=E3=81=97=E3=81=8F?= =?UTF-8?q?=E3=81=AF=E3=83=95=E3=82=A1=E3=82=A4=E3=83=AB=E3=83=91=E3=82=B9?= =?UTF-8?q?"=E3=81=AB?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../src/infer/runtimes/onnxruntime.rs | 18 ++++++++++++------ .../include/voicevox_core.h | 4 ++-- crates/voicevox_core_c_api/src/lib.rs | 4 ++-- .../jp/hiroshiba/voicevoxcore/Onnxruntime.java | 7 ++++--- .../python/voicevox_core/_rust/asyncio.pyi | 5 ++++- .../python/voicevox_core/_rust/blocking.pyi | 5 ++++- example/kotlin/README.md | 2 +- example/kotlin/app/src/main/kotlin/app/App.kt | 2 +- 8 files changed, 30 insertions(+), 17 deletions(-) diff --git a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs index 63a6bba27..253ed7fe8 100644 --- a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs +++ b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs @@ -386,10 +386,13 @@ pub(crate) mod blocking { #[cfg(feature = "load-onnxruntime")] impl LoadOnce { - /// ONNX Runtimeのfilenameを指定する。 + /// ONNX Runtimeのファイル名（モジュール名）もしくはファイルパスを指定する。 /// - /// デフォルトは[`Onnxruntime::LIB_VERSIONED_FILENAME`]。filenameは - /// `dlopen`/`LoadLibraryExW`の引数に使われる。 + /// `dlopen`/[`LoadLibraryExW`]の引数に使われる。デフォルト + /// は[`Onnxruntime::LIB_VERSIONED_FILENAME`]。 + /// + /// [`LoadLibraryExW`]: + /// https://learn.microsoft.com/en-us/windows/win32/api/libloaderapi/nf-libloaderapi-loadlibraryexw pub fn filename(mut self, filename: impl Into) -> Self { self.filename = filename.into(); self @@ -525,10 +528,13 @@ pub(crate) mod tokio { #[cfg(feature = "load-onnxruntime")] impl LoadOnce { - /// ONNX Runtimeのfilenameを指定する。 + /// ONNX Runtimeのファイル名（モジュール名）もしくはファイルパスを指定する。 + /// + /// `dlopen`/[`LoadLibraryExW`]の引数に使われる。デフォルト + /// は[`Onnxruntime::LIB_VERSIONED_FILENAME`]。 /// - /// デフォルトは[`Onnxruntime::LIB_VERSIONED_FILENAME`]。filenameは - /// `dlopen`/`LoadLibraryExW`の引数に使われる。 + /// [`LoadLibraryExW`]: + /// https://learn.microsoft.com/en-us/windows/win32/api/libloaderapi/nf-libloaderapi-loadlibraryexw pub fn filename(self, filename: impl Into) -> Self { Self(self.0.filename(filename)) } diff --git a/crates/voicevox_core_c_api/include/voicevox_core.h b/crates/voicevox_core_c_api/include/voicevox_core.h index 5a31eb728..c25be166e 100644 --- a/crates/voicevox_core_c_api/include/voicevox_core.h +++ b/crates/voicevox_core_c_api/include/voicevox_core.h @@ -295,9 +295,9 @@ typedef struct VoicevoxVoiceModel VoicevoxVoiceModel; */ typedef struct VoicevoxLoadOnnxruntimeOptions { /** - * ONNX Runtimeのfilenameを指定する。 + * ONNX Runtimeのファイル名（モジュール名）もしくはファイルパスを指定する。 * - * `dlopen`/`LoadLibraryExW`の引数に使われる。 + * `dlopen`/[`LoadLibraryExW`](https://learn.microsoft.com/en-us/windows/win32/api/libloaderapi/nf-libloaderapi-loadlibraryexw)の引数に使われる。デフォルトは ::voicevox_get_onnxruntime_lib_versioned_filename と同じ。 */ const char *filename; } VoicevoxLoadOnnxruntimeOptions; diff --git a/crates/voicevox_core_c_api/src/lib.rs b/crates/voicevox_core_c_api/src/lib.rs index bbedc0dbf..f9f2152c6 100644 --- a/crates/voicevox_core_c_api/src/lib.rs +++ b/crates/voicevox_core_c_api/src/lib.rs @@ -118,9 +118,9 @@ pub extern "C" fn voicevox_get_onnxruntime_lib_unversioned_filename() -> *const #[cfg(feature = "load-onnxruntime")] #[repr(C)] pub struct VoicevoxLoadOnnxruntimeOptions { - /// ONNX Runtimeのfilenameを指定する。 + /// ONNX Runtimeのファイル名（モジュール名）もしくはファイルパスを指定する。 /// - /// `dlopen`/`LoadLibraryExW`の引数に使われる。 + /// `dlopen`/[`LoadLibraryExW`](https://learn.microsoft.com/en-us/windows/win32/api/libloaderapi/nf-libloaderapi-loadlibraryexw)の引数に使われる。デフォルトは ::voicevox_get_onnxruntime_lib_versioned_filename と同じ。 filename: *const c_char, } diff --git a/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Onnxruntime.java b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Onnxruntime.java index d0c45e6a4..9a9cbe133 100644 --- a/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Onnxruntime.java +++ b/crates/voicevox_core_java_api/lib/src/main/java/jp/hiroshiba/voicevoxcore/Onnxruntime.java @@ -74,10 +74,11 @@ public static LoadOnce loadOnce() { /** {@link #loadOnce}のビルダー。 */ public static class LoadOnce { /** - * ONNX Runtimeのfilenameを指定する。 + * ONNX Runtimeのファイル名（モジュール名）もしくはファイルパスを指定する。 * - * @param filename {@code dlopen}/{@code LoadLibraryExW}の引数に使われるfilename。デフォルトは{@link - * LIB_VERSIONED_FILENAME}。 + * @param filename {@code dlopen}/{@code + * LoadLibraryExW}の引数に使われる。デフォルトは{@link LIB_VERSIONED_FILENAME}。 * @return このオブジェクト。 */ public LoadOnce filename(@Nonnull String filename) { diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi b/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi index 9c47127ec..7652a7d2c 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi +++ b/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi @@ -93,7 +93,10 @@ class Onnxruntime: Parameters ---------- filename - ``dlopen``/``LoadLibraryExW`` の引数に使われるfilename。 + ONNX Runtimeのファイル名（モジュール名）もしくはファイルパス。 + ``dlopen``/`LoadLibraryExW + `_ + の引数に使われる。 """ ... def supported_devices(self) -> SupportedDevices: diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi b/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi index cfdf73f7f..602ff31bc 100644 --- a/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi +++ b/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi @@ -93,7 +93,10 @@ class Onnxruntime: Parameters ---------- filename - ``dlopen``/``LoadLibraryExW`` の引数に使われるfilename。 + ONNX Runtimeのファイル名（モジュール名）もしくはファイルパス。 + ``dlopen``/`LoadLibraryExW + `_ + の引数に使われる。 """ ... def supported_devices(self) -> SupportedDevices: diff --git a/example/kotlin/README.md b/example/kotlin/README.md index 9a244dd2c..7dc5b637c 100644 --- a/example/kotlin/README.md +++ b/example/kotlin/README.md @@ -47,7 +47,7 @@ Usage: voicevoxcoreexample options_list Options: --mode [AUTO] -> モード { Value should be one of [auto, cpu, gpu] } --vvm -> vvmファイルへのパス (always required) { String } - --onnxruntime [libonnxruntime.so.1.17.3] -> ONNX Runtimeのファイルネーム { String } + --onnxruntime [libonnxruntime.so.1.17.3] -> ONNX Runtimeのファイル名（モジュール名）もしくはファイルパス { String } --dictDir [./open_jtalk_dic_utf_8-1.11] -> Open JTalkの辞書ディレクトリ { String } --text [この音声は、ボイスボックスを使用して、出力されています。] -> 読み上げさせたい文章 { String } --out [./output.wav] -> 出力wavファイルのパス { String } diff --git a/example/kotlin/app/src/main/kotlin/app/App.kt b/example/kotlin/app/src/main/kotlin/app/App.kt index 1c517c582..7f2651020 100644 --- a/example/kotlin/app/src/main/kotlin/app/App.kt +++ b/example/kotlin/app/src/main/kotlin/app/App.kt @@ -17,7 +17,7 @@ fun main(args: Array) { parser.option(ArgType.String, fullName = "vvm", description = "vvmファイルへのパス").required() val onnxruntime by parser - .option(ArgType.String, description = "ONNX Runtimeのファイルネーム") + .option(ArgType.String, description = "ONNX Runtimeのファイル名（モジュール名）もしくはファイルパス") .default(Onnxruntime.LIB_VERSIONED_FILENAME) val dictDir by parser From c79c3f22c5f1f058419fa3cafac142541898ab99 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Mon, 1 Jul 2024 02:29:29 +0900 Subject: [PATCH 17/26] =?UTF-8?q?C=20API=E3=81=AE=E3=83=89=E3=82=AD?= =?UTF-8?q?=E3=83=A5=E3=83=A1=E3=83=B3=E3=83=88=E3=81=AB"Availability"?= =?UTF-8?q?=E3=81=AE=E3=82=BB=E3=82=AF=E3=82=B7=E3=83=A7=E3=83=B3=E3=82=92?= =?UTF-8?q?=E8=BF=BD=E5=8A=A0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../include/voicevox_core.h | 24 +++++++++++++++++++ crates/voicevox_core_c_api/src/lib.rs | 24 +++++++++++++++++++ docs/apis/c_api/doxygen/Doxyfile | 1 + 3 files changed, 49 insertions(+) diff --git a/crates/voicevox_core_c_api/include/voicevox_core.h b/crates/voicevox_core_c_api/include/voicevox_core.h index c25be166e..7cf68a1e5 100644 --- a/crates/voicevox_core_c_api/include/voicevox_core.h +++ b/crates/voicevox_core_c_api/include/voicevox_core.h @@ -292,6 +292,10 @@ typedef struct VoicevoxVoiceModel VoicevoxVoiceModel; #if defined(VOICEVOX_LOAD_ONNXRUNTIME) /** * ::voicevox_onnxruntime_load_once のオプション。 + * + * \availability{ + * ::voicevox_onnxruntime_load_once が利用可能なときのみ、この型定義も利用可能になる。 + * } */ typedef struct VoicevoxLoadOnnxruntimeOptions { /** @@ -385,6 +389,10 @@ extern "C" { * ONNX Runtimeの動的ライブラリの、バージョン付きのファイル名。 * * WindowsとAndroidでは ::voicevox_get_onnxruntime_lib_unversioned_filename と同じ。 + * + * \availability{ + * ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 + * } */ #ifdef _WIN32 __declspec(dllimport) @@ -395,6 +403,10 @@ const char *voicevox_get_onnxruntime_lib_versioned_filename(void); #if defined(VOICEVOX_LOAD_ONNXRUNTIME) /** * ONNX Runtimeの動的ライブラリの、バージョン無しのファイル名。 + * + * \availability{ + * ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 + * } */ #ifdef _WIN32 __declspec(dllimport) @@ -407,6 +419,10 @@ const char *voicevox_get_onnxruntime_lib_unversioned_filename(void); * デフォルトの ::voicevox_onnxruntime_load_once のオプションを生成する。 * * @return デフォルトの ::voicevox_onnxruntime_load_once のオプション + * + * \availability{ + * ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 + * } */ #ifdef _WIN32 __declspec(dllimport) @@ -437,6 +453,10 @@ const struct VoicevoxOnnxruntime *voicevox_onnxruntime_get(void); * * @returns 結果コード * + * \availability{ + * バイナリによって、この関数か ::voicevox_onnxruntime_init_once のどちらかのみが利用可能。 + * } + * * \safety{ * - `options.filename`はヌル終端文字列を指し、かつ読み込みについて有効でなければならない。 * - `out_onnxruntime`は書き込みについて有効でなければならない。 @@ -459,6 +479,10 @@ VoicevoxResultCode voicevox_onnxruntime_load_once(struct VoicevoxLoadOnnxruntime * * @returns 結果コード * + * \availability{ + * バイナリによって、この関数か ::voicevox_onnxruntime_load_once のどちらかのみが利用可能。 + * } + * * \safety{ * - `out_onnxruntime`は書き込みについて有効でなければならない。 * } diff --git a/crates/voicevox_core_c_api/src/lib.rs b/crates/voicevox_core_c_api/src/lib.rs index f9f2152c6..d68ba5d8e 100644 --- a/crates/voicevox_core_c_api/src/lib.rs +++ b/crates/voicevox_core_c_api/src/lib.rs @@ -97,6 +97,10 @@ fn init_logger_once() { /// ONNX Runtimeの動的ライブラリの、バージョン付きのファイル名。 /// /// WindowsとAndroidでは ::voicevox_get_onnxruntime_lib_unversioned_filename と同じ。 +/// +/// \availability{ +/// ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 +/// } #[cfg(feature = "load-onnxruntime")] #[no_mangle] pub extern "C" fn voicevox_get_onnxruntime_lib_versioned_filename() -> *const c_char { @@ -106,6 +110,10 @@ pub extern "C" fn voicevox_get_onnxruntime_lib_versioned_filename() -> *const c_ } /// ONNX Runtimeの動的ライブラリの、バージョン無しのファイル名。 +/// +/// \availability{ +/// ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 +/// } #[cfg(feature = "load-onnxruntime")] #[no_mangle] pub extern "C" fn voicevox_get_onnxruntime_lib_unversioned_filename() -> *const c_char { @@ -115,6 +123,10 @@ pub extern "C" fn voicevox_get_onnxruntime_lib_unversioned_filename() -> *const } /// ::voicevox_onnxruntime_load_once のオプション。 +/// +/// \availability{ +/// ::voicevox_onnxruntime_load_once が利用可能なときのみ、この型定義も利用可能になる。 +/// } #[cfg(feature = "load-onnxruntime")] #[repr(C)] pub struct VoicevoxLoadOnnxruntimeOptions { @@ -127,6 +139,10 @@ pub struct VoicevoxLoadOnnxruntimeOptions { /// デフォルトの ::voicevox_onnxruntime_load_once のオプションを生成する。 /// /// @return デフォルトの ::voicevox_onnxruntime_load_once のオプション +/// +/// \availability{ +/// ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 +/// } #[cfg(feature = "load-onnxruntime")] #[no_mangle] pub extern "C" fn voicevox_make_default_load_onnxruntime_options() -> VoicevoxLoadOnnxruntimeOptions @@ -177,6 +193,10 @@ pub extern "C" fn voicevox_onnxruntime_get() -> Option<&'static VoicevoxOnnxrunt /// /// @returns 結果コード /// +/// \availability{ +/// バイナリによって、この関数か ::voicevox_onnxruntime_init_once のどちらかのみが利用可能。 +/// } +/// /// \safety{ /// - `options.filename`はヌル終端文字列を指し、かつ読み込みについて有効でなければならない。 /// - `out_onnxruntime`は書き込みについて有効でなければならない。 @@ -210,6 +230,10 @@ pub unsafe extern "C" fn voicevox_onnxruntime_load_once( /// /// @returns 結果コード /// +/// \availability{ +/// バイナリによって、この関数か ::voicevox_onnxruntime_load_once のどちらかのみが利用可能。 +/// } +/// /// \safety{ /// - `out_onnxruntime`は書き込みについて有効でなければならない。 /// } diff --git a/docs/apis/c_api/doxygen/Doxyfile b/docs/apis/c_api/doxygen/Doxyfile index f9603fbd6..c845adbd4 100644 --- a/docs/apis/c_api/doxygen/Doxyfile +++ b/docs/apis/c_api/doxygen/Doxyfile @@ -2660,6 +2660,7 @@ GENERATE_LEGEND = YES DOT_CLEANUP = YES +ALIASES += availability{1}="

Availability
\1
" ALIASES += example{1}="
Example
\1
" ALIASES += examples{1}="
Examples
\1
" From dc587aed478053358a007a3936dcf6b3d563f47b Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Tue, 2 Jul 2024 02:45:46 +0900 Subject: [PATCH 18/26] =?UTF-8?q?=E3=83=95=E3=82=A3=E3=83=BC=E3=83=81?= =?UTF-8?q?=E3=83=A3=E3=81=AB=E9=96=A2=E3=81=99=E3=82=8B=E3=83=89=E3=82=AD?= =?UTF-8?q?=E3=83=A5=E3=83=A1=E3=83=B3=E3=83=88=E3=82=92=E6=9B=B4=E6=96=B0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit https://github.com/VOICEVOX/voicevox_core/pull/802#issuecomment-2198751933 --- crates/voicevox_core/Cargo.toml | 7 ++-- crates/voicevox_core/src/lib.rs | 36 +++++++++++++++++++ crates/voicevox_core_c_api/cbindgen.toml | 13 +++++++ .../include/voicevox_core.h | 25 +++++++++---- crates/voicevox_core_c_api/src/lib.rs | 12 +++---- 5 files changed, 76 insertions(+), 17 deletions(-) diff --git a/crates/voicevox_core/Cargo.toml b/crates/voicevox_core/Cargo.toml index 5c04b3009..b1d95c0f2 100644 --- a/crates/voicevox_core/Cargo.toml +++ b/crates/voicevox_core/Cargo.toml @@ -11,14 +11,11 @@ rustdoc-args = ["--cfg", "docsrs"] [features] default = [] -# ONNX Runtimeのリンク方法を決めるフィーチャ。どちらか片方を有効化しなければならず、両方の有効化は -# 許容されない。 -# -# `load-onnxruntime`の場合、ONNX Runtimeを`dlopen`/`LoadLibraryExW`で開く。`link-onnxruntime`の場合 -# はONNX Runtimeをロード時動的リンクする。またそれに伴い、`struct Onnxruntime`の初期化方法も異なる。 +# ONNX Runtimeのリンク方法を決めるフィーチャ（rustdocを参照）。 load-onnxruntime = ["voicevox-ort/load-dynamic"] link-onnxruntime = [] +# GPUを利用可能にするフィーチャ（rustdocを参照）。 cuda = ["voicevox-ort/cuda"] directml = ["voicevox-ort/directml"] diff --git a/crates/voicevox_core/src/lib.rs b/crates/voicevox_core/src/lib.rs index 3dbed0756..910f9eeef 100644 --- a/crates/voicevox_core/src/lib.rs +++ b/crates/voicevox_core/src/lib.rs @@ -1,4 +1,40 @@ //! 無料で使える中品質なテキスト読み上げソフトウェア、VOICEVOXのコア。 +//! +//! # Feature flags +//! +//! ## ONNX Runtimeのリンク方法を決めるフィーチャ +//! +//! このクレートの利用にあたっては以下の二つの[Cargoフィーチャ]のうちどちらかを有効にしなければなり +//! ません。両方の有効化はコンパイルエラーとなります。[`Onnxruntime`]の初期化方法はこれらの +//! フィーチャによって決まります。 +//! +//! - **`load-onnxruntime`**: ONNX Runtimeを`dlopen`/`LoadLibraryExW`で開きます。 +//! - **`link-onnxruntime`**: ONNX Runtimeをロード時動的リンクします。iOSのような`dlopen`の利用が +//! 困難な環境でのみこちらを利用するべきです。_Note_: +//! [動的リンク対象のライブラリ名]は`onnxruntime`で固定です。変更 +//! は`patchelf(1)`や`install_name_tool(1)`で行ってください。 +//! +//! ## GPUを利用可能にするフィーチャ +//! +//! - **`cuda`** +//! - **`directml`** +// TODO: こんな感じ(↓)で書く +////! - **`cuda`**: [CUDAを用いた機械学習推論]を可能にします。 +////! - ❗ [acceleration\_mode]={Gpu,Auto}のときの挙動が変化します。`directml`と共に +////! 有効化したときの挙動は未規定です。 +////! - **`directml`**: [DirectMLを用いた機械学習推論]を可能にします。 +////! - ❗ 〃 +////! +////! [CUDAを用いた機械学習推論]: +////! https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html +////! [DirectMLを用いた機械学習推論]: +////! https://onnxruntime.ai/docs/execution-providers/DirectML-ExecutionProvider.html +////! [acceleration\_mode]: InitializeOptions::acceleration_mode +//! +//! [Cargoフィーチャ]: https://doc.rust-lang.org/stable/cargo/reference/features.html +//! [動的リンク対象のライブラリ名]: +//! https://doc.rust-lang.org/cargo/reference/build-scripts.html#rustc-link-lib +//! [`Onnxruntime`]: blocking::Onnxruntime #![cfg_attr(docsrs, feature(doc_cfg))] diff --git a/crates/voicevox_core_c_api/cbindgen.toml b/crates/voicevox_core_c_api/cbindgen.toml index 28aee5706..c377daa20 100644 --- a/crates/voicevox_core_c_api/cbindgen.toml +++ b/crates/voicevox_core_c_api/cbindgen.toml @@ -9,6 +9,19 @@ header = """ * 無料で使える中品質なテキスト読み上げソフトウェア、VOICEVOXのコア。 * *
+ *
+ * Availability + *
+ * + *
+ * ヘッダによって次の二つのマクロのうちどちらかが存在する。[リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSでのみ`VOICEVOX_LINK_ONNXRUNTIME`が、他のプラットフォームでは`VOICEVOX_LOAD_ONNXRUNTIME`が存在する。 + * + * - `VOICEVOX_LOAD_ONNXRUNTIME`: ::voicevox_onnxruntime_load_once と、それに付属するアイテムが利用可能になる。 + * - `VOICEVOX_LINK_ONNXRUNTIME`: ::voicevox_onnxruntime_init_once が利用可能になる。またこのマクロが存在するなら、このライブラリはONNX Runtimeをロード時動的リンクする。 + *
+ *
+ * + *
*
* ⚠️ Safety *
diff --git a/crates/voicevox_core_c_api/include/voicevox_core.h b/crates/voicevox_core_c_api/include/voicevox_core.h index 7cf68a1e5..fe19a4c2e 100644 --- a/crates/voicevox_core_c_api/include/voicevox_core.h +++ b/crates/voicevox_core_c_api/include/voicevox_core.h @@ -4,6 +4,19 @@ * 無料で使える中品質なテキスト読み上げソフトウェア、VOICEVOXのコア。 * *
+ *
+ * Availability + *
+ * + *
+ * ヘッダによって次の二つのマクロのうちどちらかが存在する。[リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSでのみ`VOICEVOX_LINK_ONNXRUNTIME`が、他のプラットフォームでは`VOICEVOX_LOAD_ONNXRUNTIME`が存在する。 + * + * - `VOICEVOX_LOAD_ONNXRUNTIME`: ::voicevox_onnxruntime_load_once と、それに付属するアイテムが利用可能になる。 + * - `VOICEVOX_LINK_ONNXRUNTIME`: ::voicevox_onnxruntime_init_once が利用可能になる。またこのマクロが存在するなら、このライブラリはONNX Runtimeをロード時動的リンクする。 + *
+ *
+ * + *
*
* ⚠️ Safety *
@@ -294,7 +307,7 @@ typedef struct VoicevoxVoiceModel VoicevoxVoiceModel; * ::voicevox_onnxruntime_load_once のオプション。 * * \availability{ - * ::voicevox_onnxruntime_load_once が利用可能なときのみ、この型定義も利用可能になる。 + * [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 * } */ typedef struct VoicevoxLoadOnnxruntimeOptions { @@ -391,7 +404,7 @@ extern "C" { * WindowsとAndroidでは ::voicevox_get_onnxruntime_lib_unversioned_filename と同じ。 * * \availability{ - * ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 + * [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 * } */ #ifdef _WIN32 @@ -405,7 +418,7 @@ const char *voicevox_get_onnxruntime_lib_versioned_filename(void); * ONNX Runtimeの動的ライブラリの、バージョン無しのファイル名。 * * \availability{ - * ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 + * [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 * } */ #ifdef _WIN32 @@ -421,7 +434,7 @@ const char *voicevox_get_onnxruntime_lib_unversioned_filename(void); * @return デフォルトの ::voicevox_onnxruntime_load_once のオプション * * \availability{ - * ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 + * [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 * } */ #ifdef _WIN32 @@ -454,7 +467,7 @@ const struct VoicevoxOnnxruntime *voicevox_onnxruntime_get(void); * @returns 結果コード * * \availability{ - * バイナリによって、この関数か ::voicevox_onnxruntime_init_once のどちらかのみが利用可能。 + * [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 * } * * \safety{ @@ -480,7 +493,7 @@ VoicevoxResultCode voicevox_onnxruntime_load_once(struct VoicevoxLoadOnnxruntime * @returns 結果コード * * \availability{ - * バイナリによって、この関数か ::voicevox_onnxruntime_load_once のどちらかのみが利用可能。 + * [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSでのみ利用可能。詳細はファイルレベルの"Availability"の節を参照。 * } * * \safety{ diff --git a/crates/voicevox_core_c_api/src/lib.rs b/crates/voicevox_core_c_api/src/lib.rs index d68ba5d8e..964e635ab 100644 --- a/crates/voicevox_core_c_api/src/lib.rs +++ b/crates/voicevox_core_c_api/src/lib.rs @@ -99,7 +99,7 @@ fn init_logger_once() { /// WindowsとAndroidでは ::voicevox_get_onnxruntime_lib_unversioned_filename と同じ。 /// /// \availability{ -/// ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 +/// [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 /// } #[cfg(feature = "load-onnxruntime")] #[no_mangle] @@ -112,7 +112,7 @@ pub extern "C" fn voicevox_get_onnxruntime_lib_versioned_filename() -> *const c_ /// ONNX Runtimeの動的ライブラリの、バージョン無しのファイル名。 /// /// \availability{ -/// ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 +/// [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 /// } #[cfg(feature = "load-onnxruntime")] #[no_mangle] @@ -125,7 +125,7 @@ pub extern "C" fn voicevox_get_onnxruntime_lib_unversioned_filename() -> *const /// ::voicevox_onnxruntime_load_once のオプション。 /// /// \availability{ -/// ::voicevox_onnxruntime_load_once が利用可能なときのみ、この型定義も利用可能になる。 +/// [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 /// } #[cfg(feature = "load-onnxruntime")] #[repr(C)] @@ -141,7 +141,7 @@ pub struct VoicevoxLoadOnnxruntimeOptions { /// @return デフォルトの ::voicevox_onnxruntime_load_once のオプション /// /// \availability{ -/// ::voicevox_onnxruntime_load_once が利用可能なときのみ、この関数も利用可能になる。 +/// [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 /// } #[cfg(feature = "load-onnxruntime")] #[no_mangle] @@ -194,7 +194,7 @@ pub extern "C" fn voicevox_onnxruntime_get() -> Option<&'static VoicevoxOnnxrunt /// @returns 結果コード /// /// \availability{ -/// バイナリによって、この関数か ::voicevox_onnxruntime_init_once のどちらかのみが利用可能。 +/// [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSを除くプラットフォームで利用可能。詳細はファイルレベルの"Availability"の節を参照。 /// } /// /// \safety{ @@ -231,7 +231,7 @@ pub unsafe extern "C" fn voicevox_onnxruntime_load_once( /// @returns 結果コード /// /// \availability{ -/// バイナリによって、この関数か ::voicevox_onnxruntime_load_once のどちらかのみが利用可能。 +/// [リリース](https://github.com/voicevox/voicevox_core/releases)されているライブラリではiOSでのみ利用可能。詳細はファイルレベルの"Availability"の節を参照。 /// } /// /// \safety{ From e99b2936058153429d6a4e27a36bf1aea73bda33 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Tue, 2 Jul 2024 02:53:29 +0900 Subject: [PATCH 19/26] =?UTF-8?q?C=20API=E3=81=AE=E3=83=93=E3=83=AB?= =?UTF-8?q?=E3=83=89=E3=81=AB=E9=96=A2=E3=81=99=E3=82=8B=E3=83=89=E3=82=AD?= =?UTF-8?q?=E3=83=A5=E3=83=A1=E3=83=B3=E3=83=88=E3=81=AE=E3=83=98=E3=83=83?= =?UTF-8?q?=E3=83=80=E3=83=95=E3=82=A1=E3=82=A4=E3=83=AB=E5=91=A8=E3=82=8A?= =?UTF-8?q?=E3=82=92=E6=9B=B4=E6=96=B0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- README.md | 10 +++++++++- docs/feature-options.md | 6 ++++++ 2 files changed, 15 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index 67ef5d6b2..ae60f587b 100644 --- a/README.md +++ b/README.md @@ -153,7 +153,15 @@ model フォルダにある onnx モデルはダミーのため、ノイズの cargo build --release -p voicevox_core_c_api --features load-onnxruntime ``` -DLL 用のヘッダファイルは [crates/voicevox_core_c_api/include/voicevox_core.h](https://github.com/VOICEVOX/voicevox_core/tree/main/crates/voicevox_core_c_api/include/voicevox_core.h) にあります。 +DLL 用のヘッダファイルの雛形は [crates/voicevox_core_c_api/include/voicevox_core.h](https://github.com/VOICEVOX/voicevox_core/tree/main/crates/voicevox_core_c_api/include/voicevox_core.h) にあります。 +詳しくは[feature-options.md](./docs/feature-options.md)を参照してください。 + +```bash +# ヘッダファイルを加工し、マクロ`VOICEVOX_LOAD_ONNXRUNTIME`を宣言 +sed 's:^//\(#define VOICEVOX_LOAD_ONNXRUNTIME\)$:\1:' \ + crates/voicevox_core_c_api/include/voicevox_core.h \ + > ./voicevox_core.h +``` ## コアライブラリのテスト diff --git a/docs/feature-options.md b/docs/feature-options.md index e861446dd..16487d3bf 100644 --- a/docs/feature-options.md +++ b/docs/feature-options.md @@ -10,10 +10,16 @@ Rust API（`voicevox_core`）およびC API（`voicevox_core_c_api`）におい ```console ❯ cargo build --release -p voicevox_core_c_api --features load-onnxruntime +❯ sed 's:^//\(#define VOICEVOX_LOAD_ONNXRUNTIME\)$:\1:' \ + crates/voicevox_core_c_api/include/voicevox_core.h \ + > ./voicevox_core.h ``` ```console ❯ cargo build --release -p voicevox_core_c_api --features link-onnxruntime +❯ sed 's:^//\(#define VOICEVOX_LINK_ONNXRUNTIME\)$:\1:' \ + crates/voicevox_core_c_api/include/voicevox_core.h \ + > ./voicevox_core.h ``` C APIのリリースでは`dlopen`の利用が厳しいiOSでのみ`link-onnxruntime`で、その他は`load-onnxruntime`で From cc1f6ffeb6fe82e0d0f65212132fa0cc69716cc0 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Tue, 2 Jul 2024 04:34:25 +0900 Subject: [PATCH 20/26] =?UTF-8?q?voicevox-ort=E3=82=92=E6=9B=B4=E6=96=B0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- Cargo.lock | 4 ++-- Cargo.toml | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/Cargo.lock b/Cargo.lock index 2d08d5cd5..d6929942d 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -3623,7 +3623,7 @@ checksum = "49874b5167b65d7193b8aba1567f5c7d93d001cafc34600cee003eda787e483f" [[package]] name = "voicevox-ort" version = "2.0.0-rc.2" -source = "git+https://github.com/qryxip/ort.git?rev=2ab092ec0e874932bf579b51d28aa389f2fa8ea9#2ab092ec0e874932bf579b51d28aa389f2fa8ea9" +source = "git+https://github.com/VOICEVOX/ort.git?rev=07c047c449b959d8f76593046e139bae520d59c3#07c047c449b959d8f76593046e139bae520d59c3" dependencies = [ "anyhow", "half", @@ -3640,7 +3640,7 @@ dependencies = [ [[package]] name = "voicevox-ort-sys" version = "2.0.0-rc.2" -source = "git+https://github.com/qryxip/ort.git?rev=2ab092ec0e874932bf579b51d28aa389f2fa8ea9#2ab092ec0e874932bf579b51d28aa389f2fa8ea9" +source = "git+https://github.com/VOICEVOX/ort.git?rev=07c047c449b959d8f76593046e139bae520d59c3#07c047c449b959d8f76593046e139bae520d59c3" dependencies = [ "flate2", "sha2", diff --git a/Cargo.toml b/Cargo.toml index b79872321..b467c583e 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -89,8 +89,8 @@ windows = "0.43.0" zip = "0.6.3" [workspace.dependencies.voicevox-ort] -git = "https://github.com/qryxip/ort.git" -rev = "2ab092ec0e874932bf579b51d28aa389f2fa8ea9" +git = "https://github.com/VOICEVOX/ort.git" +rev = "07c047c449b959d8f76593046e139bae520d59c3" [workspace.dependencies.open_jtalk] git = "https://github.com/VOICEVOX/open_jtalk-rs.git" From c301b11dc8fb286b3c8035b90291cbef4688d782 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Tue, 2 Jul 2024 14:21:08 +0900 Subject: [PATCH 21/26] =?UTF-8?q?`JavaStr`=E3=82=92`CStr::to=5Fstr`?= =?UTF-8?q?=E3=81=97=E3=81=AA=E3=81=84?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- crates/voicevox_core_java_api/src/common.rs | 6 ------ crates/voicevox_core_java_api/src/onnxruntime.rs | 2 +- 2 files changed, 1 insertion(+), 7 deletions(-) diff --git a/crates/voicevox_core_java_api/src/common.rs b/crates/voicevox_core_java_api/src/common.rs index b38fdf648..1231bf3c2 100644 --- a/crates/voicevox_core_java_api/src/common.rs +++ b/crates/voicevox_core_java_api/src/common.rs @@ -149,9 +149,6 @@ where env.throw_new("java/lang/IllegalArgumentException", error.to_string()) ) } - JavaApiError::Utf8(error) => or_panic!( - env.throw_new("java/lang/IllegalArgumentException", error.to_string()) - ), JavaApiError::DeJson(error) => { or_panic!( env.throw_new("java/lang/IllegalArgumentException", error.to_string()) @@ -175,9 +172,6 @@ pub(crate) enum JavaApiError { #[from] Uuid(uuid::Error), - #[from] - Utf8(Utf8Error), - DeJson(serde_json::Error), } diff --git a/crates/voicevox_core_java_api/src/onnxruntime.rs b/crates/voicevox_core_java_api/src/onnxruntime.rs index 0af7dfb60..004ff4d97 100644 --- a/crates/voicevox_core_java_api/src/onnxruntime.rs +++ b/crates/voicevox_core_java_api/src/onnxruntime.rs @@ -31,7 +31,7 @@ unsafe extern "system" fn Java_jp_hiroshiba_voicevoxcore_Onnxruntime_rsNew<'loca filename: JString<'local>, ) { throw_if_err(env, (), |env| { - let filename = env.get_string(&filename)?.to_str()?.to_owned(); + let filename = String::from(env.get_string(&filename)?); let internal = voicevox_core::blocking::Onnxruntime::load_once() .filename(filename) .exec()?; From 8a1e2d1e7b7c4c52e38874f5e3cefeaa918c6cb3 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Tue, 2 Jul 2024 14:27:22 +0900 Subject: [PATCH 22/26] =?UTF-8?q?fixup!=20`JavaStr`=E3=82=92`CStr::to=5Fst?= =?UTF-8?q?r`=E3=81=97=E3=81=AA=E3=81=84?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- crates/voicevox_core_java_api/src/common.rs | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/crates/voicevox_core_java_api/src/common.rs b/crates/voicevox_core_java_api/src/common.rs index 1231bf3c2..c55ce8106 100644 --- a/crates/voicevox_core_java_api/src/common.rs +++ b/crates/voicevox_core_java_api/src/common.rs @@ -1,4 +1,4 @@ -use std::{error::Error as _, iter, str::Utf8Error}; +use std::{error::Error as _, iter}; use derive_more::From; use easy_ext::ext; From 07d70e084496fae70fcf528e1161a1e1471f708f Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Tue, 2 Jul 2024 14:28:57 +0900 Subject: [PATCH 23/26] =?UTF-8?q?=E4=B8=8D=E8=A6=81=E3=81=AA=E4=BE=9D?= =?UTF-8?q?=E5=AD=98=E3=82=92=E5=89=8A=E9=99=A4?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- Cargo.lock | 1 - crates/voicevox_core/Cargo.toml | 1 - 2 files changed, 2 deletions(-) diff --git a/Cargo.lock b/Cargo.lock index d6929942d..a612d57e6 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -3675,7 +3675,6 @@ dependencies = [ "open_jtalk", "ouroboros", "pretty_assertions", - "process_path", "rayon", "ref-cast", "regex", diff --git a/crates/voicevox_core/Cargo.toml b/crates/voicevox_core/Cargo.toml index b1d95c0f2..44341cf04 100644 --- a/crates/voicevox_core/Cargo.toml +++ b/crates/voicevox_core/Cargo.toml @@ -40,7 +40,6 @@ ndarray.workspace = true once_cell.workspace = true open_jtalk.workspace = true ouroboros.workspace = true -process_path.workspace = true rayon.workspace = true ref-cast.workspace = true regex.workspace = true From c545f89e3f56a50e0c2e5361f91f52d52927c8c4 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Tue, 2 Jul 2024 14:49:33 +0900 Subject: [PATCH 24/26] =?UTF-8?q?`link-onnxruntime`=E3=81=A7`cargo=20test`?= =?UTF-8?q?=E3=81=8C=E9=80=9A=E3=82=8B=E3=82=88=E3=81=86=E3=81=AB=E3=81=99?= =?UTF-8?q?=E3=82=8B?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .github/workflows/test.yml | 6 ++++-- crates/voicevox_core/Cargo.toml | 2 +- crates/voicevox_core/src/infer/runtimes/onnxruntime.rs | 6 ++++-- crates/voicevox_core/src/synthesizer.rs | 3 ++- 4 files changed, 11 insertions(+), 6 deletions(-) diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml index f66fb8a5e..a22fc5161 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test.yml @@ -136,8 +136,10 @@ jobs: - uses: Swatinem/rust-cache@v2 with: key: "cargo-integration-test-cache-${{ matrix.features }}-${{ matrix.os }}" - - name: Run cargo integration test - run: RUST_BACKTRACE=full cargo test --test "*" -vv --features ,${{ matrix.features }} -- --include-ignored + - name: Run cargo integration test (load-onnxruntime) + run: RUST_BACKTRACE=full cargo test -vv --features ,${{ matrix.features }} -- --include-ignored + - name: Run cargo integration test (link-onnxruntime) + run: RUST_BACKTRACE=full cargo test -vv -p voicevox_core --features link-onnxruntime,${{ matrix.features }} c-header: runs-on: ubuntu-latest diff --git a/crates/voicevox_core/Cargo.toml b/crates/voicevox_core/Cargo.toml index 44341cf04..527fa7494 100644 --- a/crates/voicevox_core/Cargo.toml +++ b/crates/voicevox_core/Cargo.toml @@ -63,7 +63,7 @@ pretty_assertions.workspace = true rstest.workspace = true rstest_reuse.workspace = true test_util.workspace = true -tokio = { workspace = true, features = ["rt", "macros"] } +tokio = { workspace = true, features = ["rt-multi-thread", "macros"] } [target."cfg(windows)".dependencies] humansize.workspace = true diff --git a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs index 253ed7fe8..db046553a 100644 --- a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs +++ b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs @@ -232,7 +232,8 @@ pub(crate) mod blocking { /// インスタンスは[voicevox-ort]側に作られる。Rustのクレートとしてこのライブラリを利用する場合、 /// Tokio版APIやvoicevox-ortを利用する他クレートともインスタンスが共有される。 /// - /// ``` + #[cfg_attr(feature = "load-onnxruntime", doc = "```")] + #[cfg_attr(feature = "link-onnxruntime", doc = "```compile_fail")] /// # use voicevox_core as another_lib; /// # /// # fn main() -> anyhow::Result<()> { @@ -419,7 +420,8 @@ pub(crate) mod tokio { /// インスタンスは[voicevox-ort]側に作られる。Rustのクレートとしてこのライブラリを利用する場合、 /// ブロッキング版APIやvoicevox-ortを利用する他クレートともインスタンスが共有される。 /// - /// ``` + #[cfg_attr(feature = "load-onnxruntime", doc = "```")] + #[cfg_attr(feature = "link-onnxruntime", doc = "```compile_fail")] /// # use voicevox_core as another_lib; /// # /// # #[tokio::main] diff --git a/crates/voicevox_core/src/synthesizer.rs b/crates/voicevox_core/src/synthesizer.rs index e96ebe9cb..1d6b593dc 100644 --- a/crates/voicevox_core/src/synthesizer.rs +++ b/crates/voicevox_core/src/synthesizer.rs @@ -109,7 +109,8 @@ pub(crate) mod blocking { /// /// # Example /// - /// ``` + #[cfg_attr(feature = "load-onnxruntime", doc = "```")] + #[cfg_attr(feature = "link-onnxruntime", doc = "```compile_fail")] /// # #[tokio::main] /// # async fn main() -> anyhow::Result<()> { /// # use test_util::{ONNXRUNTIME_DYLIB_PATH, OPEN_JTALK_DIC_DIR}; From f60cd0eb2e25b608292465178b5b4f67a2c4cb0f Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Tue, 2 Jul 2024 15:13:02 +0900 Subject: [PATCH 25/26] =?UTF-8?q?fixup!=20`link-onnxruntime`=E3=81=A7`carg?= =?UTF-8?q?o=20test`=E3=81=8C=E9=80=9A=E3=82=8B=E3=82=88=E3=81=86=E3=81=AB?= =?UTF-8?q?=E3=81=99=E3=82=8B?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .github/workflows/test.yml | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml index a22fc5161..8550fbfca 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test.yml @@ -137,9 +137,7 @@ jobs: with: key: "cargo-integration-test-cache-${{ matrix.features }}-${{ matrix.os }}" - name: Run cargo integration test (load-onnxruntime) - run: RUST_BACKTRACE=full cargo test -vv --features ,${{ matrix.features }} -- --include-ignored - - name: Run cargo integration test (link-onnxruntime) - run: RUST_BACKTRACE=full cargo test -vv -p voicevox_core --features link-onnxruntime,${{ matrix.features }} + run: RUST_BACKTRACE=full cargo test --test "*" -vv --features ,${{ matrix.features }} -- --include-ignored c-header: runs-on: ubuntu-latest From c8c7813c75b2ad4e26e3ee9bcdfcea04eb690f85 Mon Sep 17 00:00:00 2001 From: Ryo Yamashita Date: Thu, 4 Jul 2024 07:48:36 +0900 Subject: [PATCH 26/26] =?UTF-8?q?fixup!=20`link-onnxruntime`=E3=81=A7`carg?= =?UTF-8?q?o=20test`=E3=81=8C=E9=80=9A=E3=82=8B=E3=82=88=E3=81=86=E3=81=AB?= =?UTF-8?q?=E3=81=99=E3=82=8B?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- crates/voicevox_core/src/infer/runtimes/onnxruntime.rs | 4 ++-- crates/voicevox_core/src/synthesizer.rs | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs index db046553a..74dc8a601 100644 --- a/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs +++ b/crates/voicevox_core/src/infer/runtimes/onnxruntime.rs @@ -233,7 +233,7 @@ pub(crate) mod blocking { /// Tokio版APIやvoicevox-ortを利用する他クレートともインスタンスが共有される。 /// #[cfg_attr(feature = "load-onnxruntime", doc = "```")] - #[cfg_attr(feature = "link-onnxruntime", doc = "```compile_fail")] + #[cfg_attr(not(feature = "load-onnxruntime"), doc = "```compile_fail")] /// # use voicevox_core as another_lib; /// # /// # fn main() -> anyhow::Result<()> { @@ -421,7 +421,7 @@ pub(crate) mod tokio { /// ブロッキング版APIやvoicevox-ortを利用する他クレートともインスタンスが共有される。 /// #[cfg_attr(feature = "load-onnxruntime", doc = "```")] - #[cfg_attr(feature = "link-onnxruntime", doc = "```compile_fail")] + #[cfg_attr(not(feature = "load-onnxruntime"), doc = "```compile_fail")] /// # use voicevox_core as another_lib; /// # /// # #[tokio::main] diff --git a/crates/voicevox_core/src/synthesizer.rs b/crates/voicevox_core/src/synthesizer.rs index 1d6b593dc..767d27b1c 100644 --- a/crates/voicevox_core/src/synthesizer.rs +++ b/crates/voicevox_core/src/synthesizer.rs @@ -110,7 +110,7 @@ pub(crate) mod blocking { /// # Example /// #[cfg_attr(feature = "load-onnxruntime", doc = "```")] - #[cfg_attr(feature = "link-onnxruntime", doc = "```compile_fail")] + #[cfg_attr(not(feature = "load-onnxruntime"), doc = "```compile_fail")] /// # #[tokio::main] /// # async fn main() -> anyhow::Result<()> { /// # use test_util::{ONNXRUNTIME_DYLIB_PATH, OPEN_JTALK_DIC_DIR};