Merge branch 'master' into rtx5000_fix

Author: ExtReMLapin

.github/workflows/ci.yml

@@ -17,7 +17,7 @@ jobs:
       CT2_VERBOSE: 1
     strategy:
       matrix:
-        os: [ubuntu-20.04]
+        os: [ubuntu-22.04]
         backend: [mkl, dnnl]
 
     steps:
@@ -74,7 +74,7 @@ jobs:
 
 
   build-and-test-cpp-aarch64:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     env:
       CT2_VERBOSE: 1
     strategy:
@@ -132,13 +132,15 @@ jobs:
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: [ubuntu-20.04, windows-2019]
+        os: [ubuntu-22.04, windows-2019]
         arch: [auto64]
         include:
-        - os: ubuntu-20.04
+        - os: ubuntu-22.04
           arch: aarch64
         - os: macos-13
           arch: arm64
+        - os: macos-13
+          arch: x86_64
 
     steps:
       - uses: actions/checkout@v4
@@ -150,14 +152,14 @@ jobs:
         name: Set up QEMU
 
       - name: Build wheels
-        uses: pypa/cibuildwheel@v2.21.3
+        uses: pypa/cibuildwheel@v2.22.0
         with:
           package-dir: python
           output-dir: python/wheelhouse
         env:
           CIBW_ENVIRONMENT_PASS_LINUX: CIBW_ARCHS
           CIBW_ENVIRONMENT_WINDOWS: CTRANSLATE2_ROOT='${{ github.workspace }}\install'
-          CIBW_ENVIRONMENT_MACOS: CTRANSLATE2_ROOT='/usr/local'
+          CIBW_ENVIRONMENT_MACOS: "CTRANSLATE2_ROOT='/usr/local' MACOSX_DEPLOYMENT_TARGET=10.13"
           CIBW_BEFORE_ALL_LINUX: python/tools/prepare_build_environment_linux.sh
           CIBW_BEFORE_ALL_MACOS: python/tools/prepare_build_environment_macos.sh
           CIBW_BEFORE_ALL_WINDOWS: bash python/tools/prepare_build_environment_windows.sh
@@ -181,13 +183,13 @@ jobs:
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: [ubuntu-20.04, windows-2019]
+        os: [ubuntu-22.04, windows-2019]
 
     steps:
-      - name: Set up Python 3.8
+      - name: Set up Python 3.9
         uses: actions/setup-python@v5
         with:
-          python-version: 3.8
+          python-version: 3.9
 
       - uses: actions/checkout@v4
 
@@ -206,18 +208,18 @@ jobs:
       - name: Install wheel
         if: startsWith(matrix.os, 'ubuntu')
         run: |
-          pip install *cp38*manylinux*x86_64.whl
+          pip install *cp39*manylinux*x86_64.whl
 
       - name: Install wheel
         if: startsWith(matrix.os, 'windows')
         shell: bash
         run: |
-          pip install *cp38*win*.whl
+          pip install *cp39*win*.whl
 
       - name: Run tests
         shell: bash
         run: |
-          pytest -v python/tests/
+          pytest -v python/tests/ --ignore=python/tests/test_opennmt_tf.py
 
 
   check-python-style:
@@ -226,10 +228,10 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - name: Set up Python 3.8
+      - name: Set up Python 3.9
         uses: actions/setup-python@v5
         with:
-          python-version: 3.8
+          python-version: 3.9
 
       - name: Install dependencies
         run: |
@@ -255,7 +257,7 @@ jobs:
   publish-python-wheels-on-pypi:
     if: github.event_name == 'push' && startsWith(github.ref, 'refs/tags')
     needs: [build-and-test-cpp-x86_64, build-python-wheels, test-python-wheels, check-python-style]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
 
     steps:
       - name: Download Python wheels
@@ -274,7 +276,7 @@ jobs:
 
 
   build-and-push-docker-images:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     steps:
       - uses: actions/checkout@v4
         with:
@@ -305,10 +307,10 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
-      - name: Set up Python 3.8
+      - name: Set up Python 3.9
         uses: actions/setup-python@v5
         with:
-          python-version: 3.8
+          python-version: 3.9
 
       - name: Download CTranslate2 wheels
         uses: actions/download-artifact@v4
@@ -319,7 +321,7 @@ jobs:
 
       - name: Install CTranslate2 wheel
         run: |
-          pip install *cp38*manylinux*x86_64.whl
+          pip install *cp39*manylinux*x86_64.whl
 
       - name: Install dependencies to build docs
         working-directory: docs

CHANGELOG.md

@@ -4,8 +4,23 @@
 
 ### Fixes and improvements
 
+## [v4.6.0](https://github.com/OpenNMT/CTranslate2/releases/tag/v4.6.0) (2025-04-08)
+Note: The Ctranslate2 Python package now supports python 3.13, drop the support for python 3.8.
+
+### New features
+* Pyhton 3.13 support (#1858)
+* Support returning hidden vector in Wav2Vec2 and Wav2Vec2Bert Models (#1867)
+* Add noexecstack linker flags (#1852 + #1861)
+* Support Qwen2 (#1820)
+* Eoleconv (#1832)
+* Add support RobertModel (#1864)
+
+### Fixes and improvements
+* Fix github action (#1871)
+* Prevent double library def (#1818)
+
 ## [v4.5.0](https://github.com/OpenNMT/CTranslate2/releases/tag/v4.5.0) (2024-10-22)
-Note: The Ctranslate2 Python package now supports CUDNN 9 and is no longer compatible with CUDNN 8.  
+Note: The Ctranslate2 Python package now supports CUDNN 9 and is no longer compatible with CUDNN 8.
 
 ### New features
 * Support Phi3 (#1800)
@@ -18,8 +33,8 @@ Note: The Ctranslate2 Python package now supports CUDNN 9 and is no longer compa
 * Update doc AWQ (#1795)
 
 ## [v4.4.0](https://github.com/OpenNMT/CTranslate2/releases/tag/v4.4.0) (2024-09-09)
-**Removed**: Flash Attention support in the Python package due to significant package size increase with minimal performance gain.  
-Note: Flash Attention remains supported in the C++ package with the `WITH_FLASH_ATTN` option.  
+**Removed**: Flash Attention support in the Python package due to significant package size increase with minimal performance gain.
+Note: Flash Attention remains supported in the C++ package with the `WITH_FLASH_ATTN` option.
 Flash Attention may be re-added in the future if substantial improvements are made.
 
 ### New features

docker/Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:12.2.2-cudnn8-devel-ubuntu20.04 as builder
+FROM nvidia/cuda:12.2.2-cudnn8-devel-ubuntu22.04 as builder
 
 RUN apt-get update && \
     apt-get install -y --no-install-recommends \
@@ -77,7 +77,7 @@ RUN cd python && \
     python3 -m pip --no-cache-dir install -r install_requirements.txt && \
     python3 setup.py bdist_wheel --dist-dir $CTRANSLATE2_ROOT
 
-FROM nvidia/cuda:12.2.2-base-ubuntu20.04
+FROM nvidia/cuda:12.2.2-base-ubuntu22.04
 
 # We remove the cuda-compat package because it conflicts with the CUDA Enhanced Compatibility.
 # See e.g. https://github.com/NVIDIA/nvidia-docker/issues/1515
@@ -86,7 +86,7 @@ RUN apt-get update && \
         libcublas-12-2 \
         libcudnn8=8.9.7.29-1+cuda12.2 \
         libnccl2=2.19.3-1+cuda12.2 \
-        libopenmpi3=4.0.3-0ubuntu1 \
+        libopenmpi3=4.1.2-2ubuntu1 \
         openmpi-bin \
         libgomp1 \
         python3-pip \

docker/build_all.sh

@@ -42,4 +42,4 @@ build()
     fi
 }
 
-build Dockerfile ubuntu20.04-cuda12.2
+build Dockerfile ubuntu22.04-cuda12.2

docs/installation.md

@@ -29,7 +29,7 @@ On Windows [the Visual C++ runtime](https://www.microsoft.com/en-US/download/det
 Docker images can be downloaded from the [GitHub Container registry](https://github.com/OpenNMT/CTranslate2/pkgs/container/ctranslate2):
 
 ```bash
-docker pull ghcr.io/opennmt/ctranslate2:latest-ubuntu20.04-cuda11.2
+docker pull ghcr.io/opennmt/ctranslate2:latest-ubuntu22.04-cuda11.2
 ```
 
 The images include:
@@ -40,7 +40,7 @@ The images include:
 * the translator executable, which is the image entrypoint:
 
 ```bash
-docker run --rm ghcr.io/opennmt/ctranslate2:latest-ubuntu20.04-cuda11.2 --help
+docker run --rm ghcr.io/opennmt/ctranslate2:latest-ubuntu22.04-cuda11.2 --help
 ```
 
 To update to the new version that supports CUDA 12.

docs/requirements.txt

@@ -1,3 +1,3 @@
-myst-parser==0.17.*
+myst-parser==0.18.*
 sphinx-rtd-theme==1.0.*
-sphinx==4.5.*
+sphinx==5.3.*

docs/translation.md

@@ -46,7 +46,7 @@ The translation client can be used via the Docker image:
 
 ```bash
 echo "▁H ello ▁world !" | docker run -i --rm -v $PWD:/data \
-    ghcr.io/opennmt/ctranslate2:latest-ubuntu20.04-cuda11.2 --model /data/ende_ctranslate2 --device cpu
+    ghcr.io/opennmt/ctranslate2:latest-ubuntu22.04-cuda11.2 --model /data/ende_ctranslate2 --device cpu
 ```
 
 To translate on GPU, use `docker run --gpus all` and set the option `--device cuda`. Use `--help` to see the list of available options.

python/ctranslate2/converters/opennmt_tf.py

@@ -291,9 +291,11 @@ def set_multi_head_attention(self, spec, module, self_attention=False):
     def set_layer_norm_from_wrapper(self, spec, module):
         self.set_layer_norm(
             spec,
-            module.output_layer_norm
-            if module.input_layer_norm is None
-            else module.input_layer_norm,
+            (
+                module.output_layer_norm
+                if module.input_layer_norm is None
+                else module.input_layer_norm
+            ),
         )
 
     def set_layer_norm(self, spec, module):