Merge branch 'main' into dropdefaultscudajob

Author: NicolasHug

.github/workflows/build_ffmpeg.yaml

@@ -48,6 +48,33 @@ jobs:
         mkdir -p "${artifact_dir}"
         mv ffmpeg.tar.gz "${artifact_dir}/${FFMPEG_VERSION}.tar.gz"
 
+  LGPL-Linux-aarch64:
+    strategy:
+      fail-fast: false
+      matrix:
+        ffmpeg-version: ["4.4.4", "5.1.4", "6.1.1", "7.0.1", "8.0"]
+    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
+    permissions:
+      id-token: write
+      contents: read
+    with:
+      job-name: Build
+      upload-artifact: ffmpeg-lgpl-linux_aarch64-${{ matrix.ffmpeg-version }}
+      repository: meta-pytorch/torchcodec
+      runner: linux.arm64.2xlarge
+      docker-image: pytorch/manylinux2_28_aarch64-builder:cpu-aarch64
+      script: |
+        export FFMPEG_VERSION="${{ matrix.ffmpeg-version }}"
+        export FFMPEG_ROOT="${PWD}/ffmpeg"
+
+        packaging/build_ffmpeg.sh
+
+        tar -cf ffmpeg.tar.gz ffmpeg/include ffmpeg/lib
+
+        artifact_dir="${RUNNER_ARTIFACT_DIR}/$(date +%Y-%m-%d)/linux_aarch64"
+        mkdir -p "${artifact_dir}"
+        mv ffmpeg.tar.gz "${artifact_dir}/${FFMPEG_VERSION}.tar.gz"
+
   LGPL-macOS:
     strategy:
       fail-fast: false

.github/workflows/docs.yaml

@@ -84,6 +84,7 @@ jobs:
           ${CONDA_RUN} conda info
           ${CONDA_RUN} nvidia-smi
           ${CONDA_RUN} conda list
+          echo LD_LIBRARY_PATH=$CONDA_PREFIX/lib:/usr/local/cuda/lib64/:${LD_LIBRARY_PATH} >> $GITHUB_ENV
       - name: Assert ffmpeg exists
         run: |
           ${CONDA_RUN} ffmpeg -buildconf

.github/workflows/linux_cuda_aarch64_wheel.yaml

@@ -0,0 +1,58 @@
+name: Build and test Linux CUDA aarch64 wheels
+
+on:
+  pull_request:
+  push:
+    branches:
+      - nightly
+      - main
+      - release/*
+    tags:
+        - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
+  workflow_dispatch:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ github.ref_type == 'branch' && github.sha }}-${{ github.event_name == 'workflow_dispatch' }}
+  cancel-in-progress: true
+
+permissions:
+  id-token: write
+  contents: write
+
+defaults:
+  run:
+    shell: bash -l -eo pipefail {0}
+
+jobs:
+  generate-matrix:
+    uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
+    with:
+      package-type: wheel
+      os: linux-aarch64
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      with-cpu: disable
+      with-xpu: disable
+      with-rocm: disable
+      with-cuda: enable
+      build-python-only: "disable"
+  build:
+    needs: generate-matrix
+    strategy:
+      fail-fast: false
+    name: Build and Upload wheel
+    uses: pytorch/test-infra/.github/workflows/build_wheels_linux.yml@main
+    with:
+      repository: meta-pytorch/torchcodec
+      ref: ""
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      build-matrix: ${{ needs.generate-matrix.outputs.matrix }}
+      pre-script: packaging/pre_build_script.sh
+      post-script: packaging/post_build_script.sh
+      smoke-test-script: packaging/fake_smoke_test.py
+      package-name: torchcodec
+      trigger-event: ${{ github.event_name }}
+      architecture: aarch64
+      build-platform: "python-build-package"
+      build-command: "BUILD_AGAINST_ALL_FFMPEG_FROM_S3=1 ENABLE_CUDA=1 python -m build --wheel -vvv --no-isolation"

.github/workflows/linux_cuda_wheel.yaml

@@ -99,13 +99,14 @@ jobs:
       - name: Install dependencies
         run: |
           # Note: xorg-libxau was addded to fix a problem with ffmpeg 4. We should consider removing it.
-          conda install libnpp cuda-nvrtc=${{ matrix.cuda-version }} cuda-toolkit=${{ matrix.cuda-version }} cuda-cudart=${{ matrix.cuda-version }} cuda-driver-dev=${{ matrix.cuda-version }} ffmpeg=${{ matrix.ffmpeg-version-for-tests }} xorg-libxau
-      - name: Check env
+          default-packages: "nvidia/label/cuda-${{ matrix.cuda-version }}.0::libnpp nvidia::cuda-nvrtc=${{ matrix.cuda-version }} nvidia::cuda-toolkit=${{ matrix.cuda-version }} nvidia::cuda-cudart=${{ matrix.cuda-version }} nvidia::cuda-driver-dev=${{ matrix.cuda-version }} conda-forge::ffmpeg=${{ matrix.ffmpeg-version-for-tests }} conda-forge::xorg-libxau"
+      - name: Check env, set LD_LIBRARY_PATH
         run: |
-          env
-          conda info
-          nvidia-smi
-          conda list
+          ${CONDA_RUN} env
+          ${CONDA_RUN} conda info
+          ${CONDA_RUN} nvidia-smi
+          ${CONDA_RUN} conda list
+          echo LD_LIBRARY_PATH=$CONDA_PREFIX/lib:/usr/local/cuda/lib64/:${LD_LIBRARY_PATH} >> $GITHUB_ENV
       - name: Assert ffmpeg exists
         run: |
            ffmpeg -buildconf

.pre-commit-config.yaml

@@ -14,6 +14,14 @@ repos:
       - id: check-added-large-files
         args: ['--maxkb=1000']
 
+  - repo: https://github.com/asottile/pyupgrade
+    rev: v3.21.2
+    hooks:
+    -   id: pyupgrade
+        args: [--py310-plus]
+        files: ^(test|src)/
+        exclude: ^examples/
+
   - repo: https://github.com/omnilib/ufmt
     rev: v2.6.0
     hooks:

CONTRIBUTING.md

@@ -30,7 +30,7 @@ Start by installing the **nightly** build of PyTorch following the
 Then, the easiest way to install the rest of the dependencies is to run:
 
 ```bash
-conda install cmake pkg-config pybind11 "ffmpeg<8" -c conda-forge
+conda install cmake pkg-config pybind11 "ffmpeg" -c conda-forge
 ```
 
 ### Clone and build

README.md

@@ -1,10 +1,10 @@
-[**Installation**](#installing-torchcodec) | [**Simple Example**](#using-torchcodec) | [**Detailed Example**](https://pytorch.org/torchcodec/stable/generated_examples/) | [**Documentation**](https://pytorch.org/torchcodec) | [**Contributing**](CONTRIBUTING.md) | [**License**](#license)
+[**Installation**](#installing-torchcodec) | [**Simple Example**](#using-torchcodec) | [**Detailed Example**](https://meta-pytorch.org/torchcodec/stable/generated_examples/) | [**Documentation**](https://meta-pytorch.org/torchcodec) | [**Contributing**](CONTRIBUTING.md) | [**License**](#license)
 
 # TorchCodec
 
 TorchCodec is a Python library for decoding video and audio data into PyTorch
-tensors, on CPU and CUDA GPU. It also supports audio encoding, and video
-encoding will come soon!  It aims to be fast, easy to use, and well integrated
+tensors, on CPU and CUDA GPU. It also supports video and audio encoding on CPU!
+It aims to be fast, easy to use, and well integrated
 into the PyTorch ecosystem.  If you want to use PyTorch to train ML models on
 videos and audio, TorchCodec is how you turn these into data.
 
@@ -23,7 +23,7 @@ We achieve these capabilities through:
 
 Here's a condensed summary of what you can do with TorchCodec. For more detailed
 examples, [check out our
-documentation](https://pytorch.org/torchcodec/stable/generated_examples/)!
+documentation](https://meta-pytorch.org/torchcodec/stable/generated_examples/)!
 
 #### Decoding
 
@@ -107,16 +107,16 @@ ffmpeg -f lavfi -i \
    `torch` and `torchcodec`.
 
 2. Install FFmpeg, if it's not already installed. Linux distributions usually
-   come with FFmpeg pre-installed. TorchCodec supports major FFmpeg versions
-   in [4, 7] on all platforms, and FFmpeg version 8 is supported on Mac and Linux.
+   come with FFmpeg pre-installed. TorchCodec supports supports all major FFmpeg versions
+   in [4, 8].
 
    If FFmpeg is not already installed, or you need a more recent version, an
    easy way to install it is to use `conda`:
 
    ```bash
-   conda install "ffmpeg<8"
+   conda install "ffmpeg"
    # or
-   conda install "ffmpeg<8" -c conda-forge
+   conda install "ffmpeg" -c conda-forge
    ```
 
 3. Install TorchCodec:
@@ -130,7 +130,8 @@ The following table indicates the compatibility between versions of
 
 | `torchcodec`       | `torch`            | Python              |
 | ------------------ | ------------------ | ------------------- |
-| `main` / `nightly` | `main` / `nightly` | `>=3.10`, `<=3.13`   |
+| `main` / `nightly` | `main` / `nightly` | `>=3.10`, `<=3.14`   |
+| `0.9`              | `2.9`              | `>=3.10`, `<=3.14`   |
 | `0.8`              | `2.9`              | `>=3.10`, `<=3.13`   |
 | `0.7`              | `2.8`              | `>=3.9`, `<=3.13`   |
 | `0.6`              | `2.8`              | `>=3.9`, `<=3.13`   |
@@ -148,16 +149,15 @@ format you want. Refer to Nvidia's GPU support matrix for more details
 [here](https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new).
 
 1. Install FFmpeg with NVDEC support.
-   TorchCodec with CUDA should work with FFmpeg versions in [4, 7] on all platforms,
-   and FFmpeg version 8 is supported on Linux.
+   TorchCodec with CUDA should work with FFmpeg versions in [4, 8].
 
    If FFmpeg is not already installed, or you need a more recent version, an
    easy way to install it is to use `conda`:
 
    ```bash
-   conda install "ffmpeg<8"
+   conda install "ffmpeg"
    # or
-   conda install "ffmpeg<8" -c conda-forge
+   conda install "ffmpeg" -c conda-forge
    ```
 
    After installing FFmpeg make sure it has NVDEC support when you list the supported
@@ -219,7 +219,7 @@ The bottom row is [promotional video from NASA](https://download.pytorch.org/tor
 that has a resolution of 960x540 at 29.7 fps and is 206 seconds long. Both videos were
 encoded with libx264 and yuv420p pixel format. All decoders, except for TorchVision, used FFmpeg 6.1.2. TorchVision used FFmpeg 4.2.2.
 
-For TorchCodec, the "approx" label means that it was using [approximate mode](https://pytorch.org/torchcodec/stable/generated_examples/approximate_mode.html)
+For TorchCodec, the "approx" label means that it was using [approximate mode](https://meta-pytorch.org/torchcodec/stable/generated_examples/decoding/approximate_mode.html)
 for seeking.
 
 ## Contributing

docs/requirements.txt

@@ -3,6 +3,7 @@ sphinx==5.0.0
 sphinx_design
 sphinx_copybutton
 sphinx-tabs
+sphinx-sitemap
 matplotlib
 torchvision
 ipython

docs/source/api_ref_encoders.rst

@@ -16,3 +16,4 @@ For an audio decoder tutorial, see: :ref:`sphx_glr_generated_examples_encoding_a
     :template: class.rst
 
     AudioEncoder
+    VideoEncoder

docs/source/api_ref_transforms.rst

@@ -14,4 +14,6 @@ For a tutorial, see: TODO_DECODER_TRANSFORMS_TUTORIAL.
     :template: dataclass.rst
 
     DecoderTransform
+    CenterCrop
+    RandomCrop
     Resize