Merge remote-tracking branch 'origin/main' into faster_to

.github/ISSUE_TEMPLATE/bug_report.md

@@ -59,5 +59,5 @@ If you know or suspect the reason for this bug, paste the code lines and suggest
 ## Checklist
 
 - [ ] I have checked that there is no similar issue in the repo (**required**)
-- [ ] I have read the [documentation](https://github.com/pytorch/rl/tree/main/docs/) (**required**)
+- [ ] I have read the [documentation](https://github.com/pytorch/tensordict/tree/main/docs/) (**required**)
 - [ ] I have provided a minimal working example to reproduce the bug (**required**)

.github/PULL_REQUEST_TEMPLATE.md

@@ -8,7 +8,7 @@ Why is this change required? What problem does it solve?
 If it fixes an open issue, please link to the issue here.
 You can use the syntax `close #15213` if this solves the issue #15213
 
-- [ ] I have raised an issue to propose this change ([required](https://github.com/pytorch/rl/issues) for new features and bug fixes)
+- [ ] I have raised an issue to propose this change ([required](https://github.com/pytorch/tensordict/issues) for new features and bug fixes)
 
 ## Types of changes
 
@@ -25,7 +25,7 @@ What types of changes does your code introduce? Remove all that do not apply:
 Go over all the following points, and put an `x` in all the boxes that apply.
 If you are unsure about any of these, don't hesitate to ask. We are here to help!
 
-- [ ] I have read the [CONTRIBUTION](https://github.com/pytorch/rl/blob/main/CONTRIBUTING.md) guide (**required**)
+- [ ] I have read the [CONTRIBUTION](https://github.com/pytorch/tensordict/blob/main/CONTRIBUTING.md) guide (**required**)
 - [ ] My change requires a change to the documentation.
 - [ ] I have updated the tests accordingly (*required for a bug fix or a new feature*).
 - [ ] I have updated the documentation accordingly.

.circleci/unittest/linux/scripts/install.sh → .github/unittest/linux/scripts/install.sh

@@ -6,6 +6,7 @@ unset PYTORCH_VERSION
 # In fact, keeping PYTORCH_VERSION forces us to hardcode PyTorch version in config.
 
 set -e
+set -v
 
 eval "$(./conda/bin/conda shell.bash hook)"
 conda activate ./env
@@ -25,17 +26,28 @@ fi
 git submodule sync && git submodule update --init --recursive
 
 printf "Installing PyTorch with %s\n" "${CU_VERSION}"
-if [ "${CU_VERSION:-}" == cpu ] ; then
-    pip3 install --pre torch --extra-index-url https://download.pytorch.org/whl/nightly/cpu
+if [[ "$TORCH_VERSION" == "nightly" ]]; then
+  if [ "${CU_VERSION:-}" == cpu ] ; then
+      python -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cpu
+  else
+      python -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/$CU_VERSION
+  fi
+elif [[ "$TORCH_VERSION" == "stable" ]]; then
+    if [ "${CU_VERSION:-}" == cpu ] ; then
+      python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
+  else
+      python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/$CU_VERSION
+  fi
 else
-    pip3 install --pre torch --extra-index-url https://download.pytorch.org/whl/nightly/cu113
+  printf "Failed to install pytorch"
+  exit 1
 fi
 
 printf "* Installing tensordict\n"
-pip3 install -e .
+python setup.py develop
 
 # install torchsnapshot nightly
-pip3 install git+https://github.com/pytorch/torchsnapshot
+python -m pip install git+https://github.com/pytorch/torchsnapshot --no-build-isolation
 
 # smoke test
 python -c "import functorch;import torchsnapshot"

.circleci/unittest/linux_torchrec/scripts/setup_env.sh → .github/unittest/linux/scripts/setup_env.sh

@@ -6,6 +6,8 @@
 # Do not install PyTorch and torchvision here, otherwise they also get cached.
 
 set -e
+set -v
+
 
 this_dir="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
 # Avoid error: "fatal: unsafe repository"
@@ -24,7 +26,7 @@ esac
 # 1. Install conda at ./conda
 if [ ! -d "${conda_dir}" ]; then
     printf "* Installing conda\n"
-    wget -O miniconda.sh "http://repo.continuum.io/miniconda/Miniconda3-latest-${os}-x86_64.sh"
+    wget -O miniconda.sh "http://repo.continuum.io/miniconda/Miniconda3-latest-${os}-${ARCH}.sh"
     bash ./miniconda.sh -b -f -p "${conda_dir}"
 fi
 eval "$(${conda_dir}/bin/conda shell.bash hook)"
@@ -45,3 +47,8 @@ cat "${this_dir}/environment.yml"
 pip install pip --upgrade
 
 conda env update --file "${this_dir}/environment.yml" --prune
+
+#if [[ $OSTYPE == 'darwin'* ]]; then
+#  printf "* Installing C++ for OSX\n"
+#  conda install -c conda-forge cxx-compiler -y
+#fi

.github/workflows/lint.yml

@@ -0,0 +1,75 @@
+name: Lint
+
+on:
+  pull_request:
+  push:
+    branches:
+      - nightly
+      - main
+      - release/*
+  workflow_dispatch:
+
+concurrency:
+  # Documentation suggests ${{ github.head_ref }}, but that's only available on pull_request/pull_request_target triggers, so using ${{ github.ref }}.
+  # On master, we want all builds to complete even if merging happens faster to make it easier to discover at which point something broke.
+  group: ${{ github.workflow }}-${{ github.ref == 'refs/heads/main' && format('ci-master-{0}', github.sha) || format('ci-{0}', github.ref) }}
+  cancel-in-progress: true
+
+jobs:
+  python-source-and-configs:
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      repository: pytorch/tensordict
+      script: |
+        set -euo pipefail
+
+        echo '::group::Setup environment'
+        CONDA_PATH=$(which conda)
+        eval "$(${CONDA_PATH} shell.bash hook)"
+        conda create --name ci --quiet --yes python=3.8 pip
+        conda activate ci
+        echo '::endgroup::'
+        
+        echo '::group::Install lint tools'
+        pip install --progress-bar=off pre-commit
+        echo '::endgroup::'
+        
+        echo '::group::Lint Python source and configs'
+        set +e
+        pre-commit run --all-files
+        
+        if [ $? -ne 0 ]; then
+          git --no-pager diff
+          exit 1
+        fi
+        echo '::endgroup::'
+
+  c-source:
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      repository: pytorch/tensordict
+      script: |
+        set -euo pipefail
+        
+        echo '::group::Setup environment'
+        CONDA_PATH=$(which conda)
+        eval "$(${CONDA_PATH} shell.bash hook)"
+        conda create --name ci --quiet --yes -c conda-forge python=3.8 ncurses=5 libgcc
+        conda activate ci
+        export LD_LIBRARY_PATH="${CONDA_PREFIX}/lib:${LD_LIBRARY_PATH}"
+        echo '::endgroup::'
+        
+        echo '::group::Install lint tools'
+        curl https://oss-clang-format.s3.us-east-2.amazonaws.com/linux64/clang-format-linux64 -o ./clang-format
+        chmod +x ./clang-format
+        echo '::endgroup::'
+
+        echo '::group::Lint C source'
+        set +e
+        ./.github/unittest/linux/scripts/run-clang-format.py -r torchrl/csrc --clang-format-executable ./clang-format
+        
+        if [ $? -ne 0 ]; then
+          git --no-pager diff
+          exit 1
+        fi
+        echo '::endgroup::'

.github/workflows/nightly_build.yml

@@ -30,7 +30,7 @@ concurrency:
 jobs:
   build-wheel-linux:
     # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch-labs'
+    if: github.repository_owner == 'pytorch'
     runs-on: ubuntu-20.04
     strategy:
       matrix:
@@ -69,7 +69,7 @@ jobs:
 
   build-wheel-mac:
     # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch-labs'
+    if: github.repository_owner == 'pytorch'
     runs-on: macos-latest
     strategy:
       matrix:
@@ -101,7 +101,7 @@ jobs:
 
   test-wheel-mac:
     # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch-labs'
+    if: github.repository_owner == 'pytorch'
     needs: build-wheel-mac
     runs-on: macos-latest
     strategy:
@@ -150,7 +150,7 @@ jobs:
 
   upload-wheel-linux:
     # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch-labs'
+    if: github.repository_owner == 'pytorch'
     needs: test-wheel-linux
     runs-on: ubuntu-20.04
     strategy:
@@ -181,7 +181,7 @@ jobs:
 
   upload-wheel-mac:
     # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch-labs'
+    if: github.repository_owner == 'pytorch'
     needs: test-wheel-mac
     runs-on: macos-latest
     strategy:
@@ -209,7 +209,7 @@ jobs:
 
   test-wheel-linux:
     # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch-labs'
+    if: github.repository_owner == 'pytorch'
     needs: build-wheel-linux
     runs-on: ubuntu-20.04
     strategy:
@@ -269,7 +269,7 @@ jobs:
 
   build-wheel-windows:
     # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch-labs'
+    if: github.repository_owner == 'pytorch'
     runs-on: windows-latest
     strategy:
       matrix:
@@ -301,7 +301,7 @@ jobs:
 
   test-wheel-windows:
     # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch-labs'
+    if: github.repository_owner == 'pytorch'
     needs: build-wheel-windows
     runs-on: windows-latest
     strategy:
@@ -355,7 +355,7 @@ jobs:
 
   upload-wheel-windows:
     # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch-labs'
+    if: github.repository_owner == 'pytorch'
     needs: test-wheel-windows
     runs-on: windows-latest
     strategy:

.github/workflows/test-linux.yml

@@ -0,0 +1,137 @@
+name: Unit-tests on Linux
+
+on:
+  pull_request:
+  push:
+    branches:
+      - nightly
+      - main
+      - release/*
+  workflow_dispatch:
+
+env:
+  CHANNEL: "nightly"
+
+concurrency:
+  # Documentation suggests ${{ github.head_ref }}, but that's only available on pull_request/pull_request_target triggers, so using ${{ github.ref }}.
+  # On master, we want all builds to complete even if merging happens faster to make it easier to discover at which point something broke.
+  group: ${{ github.workflow }}-${{ github.ref == 'refs/heads/main' && format('ci-master-{0}', github.sha) || format('ci-{0}', github.ref) }}
+  cancel-in-progress: true
+
+jobs:
+  test-gpu:
+    strategy:
+      matrix:
+        python_version: ["3.8"]
+        cuda_arch_version: ["12.1"]
+      fail-fast: false
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      runner: linux.g5.4xlarge.nvidia.gpu
+      repository: pytorch/tensordict
+      gpu-arch-type: cuda
+      gpu-arch-version: ${{ matrix.cuda_arch_version }}
+      script: |
+        # Set env vars from matrix
+        export PYTHON_VERSION=${{ matrix.python_version }}
+        # Commenting these out for now because the GPU test are not working inside docker
+        export CUDA_ARCH_VERSION=${{ matrix.cuda_arch_version }}
+        export CU_VERSION="cu${CUDA_ARCH_VERSION:0:2}${CUDA_ARCH_VERSION:3:1}"
+        export TORCH_VERSION=nightly
+        # Remove the following line when the GPU tests are working inside docker, and uncomment the above lines
+        #export CU_VERSION="cpu"
+        export ARCH=x86_64
+
+        echo "PYTHON_VERSION: $PYTHON_VERSION"
+        echo "CU_VERSION: $CU_VERSION"
+
+        ## setup_env.sh
+        bash .github/unittest/linux/scripts/setup_env.sh
+        bash .github/unittest/linux/scripts/install.sh
+        bash .github/unittest/linux/scripts/run_test.sh
+        bash .github/unittest/linux/scripts/post_process.sh
+
+  test-cpu:
+    strategy:
+      matrix:
+        python_version: ["3.8", "3.9", "3.10", "3.11"]
+      fail-fast: false
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      runner: linux.12xlarge
+      repository: pytorch/tensordict
+      timeout: 90
+      script: |
+        # Set env vars from matrix
+        export PYTHON_VERSION=${{ matrix.python_version }}
+        export CU_VERSION="cpu"
+        export TORCH_VERSION=nightly
+        export ARCH=x86_64
+
+        echo "PYTHON_VERSION: $PYTHON_VERSION"
+        echo "CU_VERSION: $CU_VERSION"
+
+        ## setup_env.sh
+        bash .github/unittest/linux/scripts/setup_env.sh
+        bash .github/unittest/linux/scripts/install.sh
+        bash .github/unittest/linux/scripts/run_test.sh
+        bash .github/unittest/linux/scripts/post_process.sh
+
+  test-stable-gpu:
+    strategy:
+      matrix:
+        python_version: ["3.8"]
+        cuda_arch_version: ["12.1"]
+      fail-fast: false
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      runner: linux.g5.4xlarge.nvidia.gpu
+      repository: pytorch/tensordict
+      gpu-arch-type: cuda
+      gpu-arch-version: ${{ matrix.cuda_arch_version }}
+      timeout: 90
+      script: |
+        # Set env vars from matrix
+        export PYTHON_VERSION=${{ matrix.python_version }}
+        # Commenting these out for now because the GPU test are not working inside docker
+        export CUDA_ARCH_VERSION=${{ matrix.cuda_arch_version }}
+        export CU_VERSION="cu${CUDA_ARCH_VERSION:0:2}${CUDA_ARCH_VERSION:3:1}"
+        export TORCH_VERSION=stable
+        # Remove the following line when the GPU tests are working inside docker, and uncomment the above lines
+        #export CU_VERSION="cpu"
+        export ARCH=x86_64
+
+        echo "PYTHON_VERSION: $PYTHON_VERSION"
+        echo "CU_VERSION: $CU_VERSION"
+
+        ## setup_env.sh
+        bash .github/unittest/linux/scripts/setup_env.sh
+        bash .github/unittest/linux/scripts/install.sh
+        bash .github/unittest/linux/scripts/run_test.sh
+        bash .github/unittest/linux/scripts/post_process.sh
+
+  test-stable-cpu:
+    strategy:
+      matrix:
+        python_version: ["3.8", "3.11"]
+      fail-fast: false
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      runner: linux.12xlarge
+      repository: pytorch/tensordict
+      timeout: 90
+      script: |
+        # Set env vars from matrix
+        export PYTHON_VERSION=${{ matrix.python_version }}
+        export CU_VERSION="cpu"
+        export TORCH_VERSION=stable
+        export ARCH=x86_64
+
+        echo "PYTHON_VERSION: $PYTHON_VERSION"
+        echo "CU_VERSION: $CU_VERSION"
+
+        ## setup_env.sh
+        bash .github/unittest/linux/scripts/setup_env.sh
+        bash .github/unittest/linux/scripts/install.sh
+        bash .github/unittest/linux/scripts/run_test.sh
+        bash .github/unittest/linux/scripts/post_process.sh