infra: Testing out GHA based infra

Signed-off-by: Naren Dasan <naren@narendasan.com>
Signed-off-by: Naren Dasan <narens@nvidia.com>

.github/workflows/build-wheels-linux.yml

@@ -0,0 +1,53 @@
+name: Build Linux Wheels
+
+on:
+  pull_request:
+  push:
+    branches:
+      - nightly
+      - main
+      - release/*
+    tags:
+        # NOTE: Binary build pipelines should only get triggered on release candidate builds
+        # Release candidate tags look like: v1.11.0-rc1
+        - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
+  workflow_dispatch:
+
+jobs:
+  generate-matrix:
+    uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
+    with:
+      package-type: wheel
+      os: linux
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      with-rocm: false
+  build:
+    needs: generate-matrix
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - repository: pytorch/tensorrt
+            pre-script: packaging/pre_build_script.sh
+            env-var-script: packaging/env_vars.txt
+            post-script: ""
+            smoke-test-script: ""
+            package-name: torch_tensorrt
+    name: ${{ matrix.repository }}
+    uses: pytorch/test-infra/.github/workflows/build_wheels_linux.yml@main
+    with:
+      repository: ${{ matrix.repository }}
+      ref: ""
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      build-matrix: ${{ needs.generate-matrix.outputs.matrix }}
+      pre-script: ${{ matrix.pre-script }}
+      env-var-script: ${{ matrix.env-var-script }}
+      post-script: ${{ matrix.post-script }}
+      package-name: ${{ matrix.package-name }}
+      smoke-test-script: ${{ matrix.smoke-test-script }}
+      trigger-event: ${{ github.event_name }}
+    secrets:
+      AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID: ${{ secrets.AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID }}
+      AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY: ${{ secrets.AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY }}

.github/workflows/docgen.yml

@@ -10,32 +10,33 @@ on:
 
 jobs:
   build-docs:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     container:
       image: docker.io/pytorch/manylinux-builder:cuda12.1
     steps:
       - uses: actions/checkout@v3
         with:
           ref: ${{github.head_ref}}
+      - name: Select python
+        run: |
+          export PATH=/opt/python/cp311-cp311/bin/:$PATH
       - name: Install base deps
         run: |
-          ./packaging/pre_build_script.sh
+          python3 -m pip install pyyaml
+          VERSION_SUFFIX=cu121 ./packaging/pre_build_script.sh
       - name: Get HEAD SHA
         id: vars
         run: echo "sha=$(git rev-parse --short HEAD)" >> $GITHUB_OUTPUT
-      - name: Get Bazel version
-        id: bazel_info
-        run: echo "version=$(cat .bazelversion)" >> $GITHUB_OUTPUT
       - name: Build Python Package
         run: |
-          cp toolchains/ci_workspaces/WORKSPACE.x86_64.cu121.release.rhel WORKSPACE
-          python -m pip install pip<=23
-          python -m pip install --pre -e . --extra-index-url https://download.pytorch.org/whl/nightly/cu121
+          python3 -m pip install pip --upgrade
+          CUDA_HOME=/usr/local/cuda-12.1 CI_BUILD=1 python3 -m pip install --pre -e . --extra-index-url https://download.pytorch.org/whl/nightly/cu121
       - name: Generate New Docs
         run: |
           cd docsrc
-          python -m pip install -r requirements.txt
-          python -c "import torch_tensorrt; print(torch_tensorrt.__version__)"
+          yum install -y doxygen pandoc
+          python3 -m pip install -r requirements.txt
+          python3 -c "import torch_tensorrt; print(torch_tensorrt.__version__)"
           make html
           cd ..
       - uses: stefanzweifel/git-auto-commit-action@v4

.github/workflows/nightlies.yml

@@ -0,0 +1,27 @@
+name: Cut Nightly
+
+on:
+  schedule:
+    - cron: '30 9 * * *'
+  workflow_dispatch:
+
+jobs:
+  cut_nightly:
+    runs-on: ubuntu-latest
+    timeout-minutes: 60
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          ref: main
+      - name: Create nightly commit
+        shell: bash
+        run: |
+          git config --global user.email "torch-tensorrt.github.bot@nvidia.com"
+          git config --global user.name "Torch-TensorRT Github Bot"
+          git fetch origin nightly
+          HEAD_COMMIT_HASH=$(git rev-parse HEAD)
+          NIGHTLY_DATE=$(date +"%Y-%m-%d")
+          # shellcheck disable=SC1083
+          NIGHTLY_RELEASE_COMMIT=$(git commit-tree -p FETCH_HEAD HEAD^{tree} -m "${NIGHTLY_DATE} nightly release (${HEAD_COMMIT_HASH})")
+          # shellcheck disable=SC1083
+          git push -f origin "${NIGHTLY_RELEASE_COMMIT}:nightly"

packaging/env_vars.txt

@@ -0,0 +1,2 @@
+export CI_BUILD="1"
+export RELEASE="1"

py/ci/build_whl.sh

@@ -6,11 +6,23 @@ export CXX=g++
 export CUDA_HOME=/usr/local/cuda-12.1
 export PROJECT_DIR=/workspace/project
 
-cp -r $CUDA_HOME /usr/local/cuda
+rm -rf /usr/local/cuda
+
+if [[ $CUDA_HOME == "/usr/local/cuda-12.1" ]]; then
+    cp -r /usr/local/cuda-11.8 /usr/local/cuda
+    cp -r /usr/local/cuda-12.0/ /usr/local/cuda/
+    rsync -a /usr/local/cuda-12.1/ /usr/local/cuda/
+    export CUDA_HOME=/usr/local/cuda
+else
+    ln -s $CUDA_HOME /usr/local/cuda
+fi
 
 build_wheel() {
-    $1/bin/python -m pip install --upgrade pip
-    $1/bin/python -m pip wheel . --config-setting="--build-option=--release" --config-setting="--build-option=--ci" -w dist
+    $1/bin/python -m pip install --upgrade pip setuptools
+    $1/bin/python -m pip install -r py/requirements.txt
+    #$1/bin/python -m pip wheel . -w dist
+    export BUILD_VERSION=$(cd ${PROJECT_DIR} && $1/bin/python3 -c "import versions; versions.torch_tensorrt_version_release()")
+    CI_BUILD=1 RELEASE=1 $1/bin/python setup.py bdist_wheel
 }
 
 patch_wheel() {
@@ -86,7 +98,7 @@ libtorchtrt() {
     ${PY_DIR}/bin/python -m pip install setuptools wheel auditwheel
     bazel build //:libtorchtrt --platforms //toolchains:ci_rhel_x86_64_linux -c opt --noshow_progress
     CUDA_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.cuda_version()")
-    TORCHTRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.torch_tensorrt_version()")
+    TORCHTRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.torch_tensorrt_version_release()")
     TRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.tensorrt_version()")
     CUDNN_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.cudnn_version()")
     TORCH_VERSION=$(${PY_DIR}/bin/python -c "from torch import __version__;print(__version__.split('+')[0])")
@@ -106,7 +118,7 @@ libtorchtrt_pre_cxx11_abi() {
     ${PY_DIR}/bin/python -m pip install setuptools wheel auditwheel
     bazel build //:libtorchtrt --config pre_cxx11_abi --platforms //toolchains:ci_rhel_x86_64_linux -c opt --noshow_progress
     CUDA_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.cuda_version()")
-    TORCHTRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.torch_tensorrt_version()")
+    TORCHTRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.torch_tensorrt_version_release()")
     TRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.tensorrt_version()")
     CUDNN_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.cudnn_version()")
     TORCH_VERSION=$(${PY_DIR}/bin/python -c "from torch import __version__;print(__version__.split('+')[0])")

pyproject.toml

@@ -9,8 +9,8 @@ requires = [
     "typing-extensions>=4.7.0",
     "future>=0.18.3",
     "tensorrt>=8.6,<8.7",
-    #"torch >=2.0.0,<2.1.0",
-    "torch==2.1.0.dev20230703+cu121",
+    #"torch >=2.1.0.dev,<2.2.0",
+    "torch==2.1.0.dev20230731",
     "pybind11==2.6.2",
     "numpy",
 ]
@@ -42,8 +42,8 @@ readme = {file = "py/README.md", content-type = "text/markdown"}
 requires-python = ">=3.8"
 keywords = ["pytorch", "torch", "tensorrt", "trt", "ai", "artificial intelligence", "ml", "machine learning", "dl", "deep learning", "compiler", "dynamo", "torchscript", "inference"]
 dependencies = [
-    #"torch>=2.0.0,<2.1.0",
-    "torch==2.1.0.dev20230703+cu121",
+    #"torch >=2.1.0.dev,<2.2.0",
+    "torch==2.1.0.dev20230731",
     "tensorrt>=8.6,<8.7",
     "packaging>=23",
     "numpy",

setup.py

@@ -19,35 +19,29 @@
 import platform
 import warnings
 
-dir_path = os.path.dirname(os.path.realpath(__file__)) + "/py"
+import yaml
+import re
+import os
+import subprocess
 
-CXX11_ABI = False
-JETPACK_VERSION = None
-FX_ONLY = False
-LEGACY = False
-RELEASE = False
-CI_RELEASE = False
+from datetime import datetime
+from pathlib import Path
+from typing import List
 
 __version__: str = "0.0.0"
 __cuda_version__: str = "0.0"
 __cudnn_version__: str = "0.0"
 __tensorrt_version__: str = "0.0"
 
-
-def load_version_info():
-    global __version__
-    global __cuda_version__
-    global __cudnn_version__
-    global __tensorrt_version__
-    with open("versions.yml", "r") as stream:
-        versions = yaml.safe_load(stream)
-        __version__ = versions["__version__"]
-        __cuda_version__ = versions["__cuda_version__"]
-        __cudnn_version__ = versions["__cudnn_version__"]
-        __tensorrt_version__ = versions["__tensorrt_version__"]
+LEGACY_BASE_VERSION_SUFFIX_PATTERN = re.compile("a0$")
 
 
-load_version_info()
+def get_root_dir() -> Path:
+    return Path(
+        subprocess.check_output(["git", "rev-parse", "--show-toplevel"])
+        .decode("ascii")
+        .strip()
+    )
 
 
 def get_git_revision_short_hash() -> str:
@@ -58,6 +52,45 @@ def get_git_revision_short_hash() -> str:
     )
 
 
+def get_base_version() -> str:
+    root = get_root_dir()
+    try:
+        dirty_version = open(root / "version.txt", "r").read().strip()
+    except FileNotFoundError:
+        print("# WARNING: Base version not found defaulting BUILD_VERSION to 0.1.0")
+        dirty_version = "0.1.0"
+    # Strips trailing a0 from version.txt, not too sure why it's there in the
+    # first place
+    return re.sub(LEGACY_BASE_VERSION_SUFFIX_PATTERN, "", dirty_version)
+
+
+def load_dep_info():
+    global __cuda_version__
+    global __cudnn_version__
+    global __tensorrt_version__
+    with open("dev_dep_versions.yml", "r") as stream:
+        versions = yaml.safe_load(stream)
+        if (gpu_arch_version := os.environ.get("CU_VERSION")) is not None:
+            __cuda_version__ = (
+                (gpu_arch_version[2:])[:-1] + "." + (gpu_arch_version[2:])[-1:]
+            )
+        else:
+            __cuda_version__ = versions["__cuda_version__"]
+        __cudnn_version__ = versions["__cudnn_version__"]
+        __tensorrt_version__ = versions["__tensorrt_version__"]
+
+
+load_dep_info()
+
+dir_path = str(get_root_dir()) + "/py"
+
+CXX11_ABI = False
+JETPACK_VERSION = None
+FX_ONLY = False
+LEGACY = False
+RELEASE = False
+CI_BUILD = False
+
 if "--fx-only" in sys.argv:
     FX_ONLY = True
     sys.argv.remove("--fx-only")
@@ -66,16 +99,31 @@ def get_git_revision_short_hash() -> str:
     LEGACY = True
     sys.argv.remove("--legacy")
 
-if "--release" not in sys.argv:
-    __version__ = __version__ + "+" + get_git_revision_short_hash()
-else:
+if "--release" in sys.argv:
     RELEASE = True
     sys.argv.remove("--release")
 
+if (release_env_var := os.environ.get("RELEASE")) is not None:
+    if release_env_var == "1":
+        RELEASE = True
+
+if (gpu_arch_version := os.environ.get("CU_VERSION")) is None:
+    gpu_arch_version = f"cu{__cuda_version__.replace('.','')}"
+
+
+if RELEASE:
+    __version__ = os.environ.get("BUILD_VERSION")
+else:
+    __version__ = f"{get_base_version()}.dev0+{get_git_revision_short_hash()}"
+
 if "--ci" in sys.argv:
     sys.argv.remove("--ci")
     if RELEASE:
-        CI_RELEASE = True
+        CI_BUILD = True
+
+if (ci_env_var := os.environ.get("CI_BUILD")) is not None:
+    if ci_env_var == "1":
+        CI_BUILD = True
 
 if "--use-cxx11-abi" in sys.argv:
     sys.argv.remove("--use-cxx11-abi")
@@ -159,7 +207,7 @@ def build_libtorchtrt_pre_cxx11_abi(develop=True, use_dist_dir=True, cxx11_abi=F
         cmd.append("--platforms=//toolchains:jetpack_5.0")
         print("Jetpack version: 5.0")
 
-    if CI_RELEASE:
+    if CI_BUILD:
         cmd.append("--platforms=//toolchains:ci_rhel_x86_64_linux")
         print("CI based build")
 
@@ -412,6 +460,7 @@ def run(self):
                 dir_path + "/../bazel-TensorRT/external/tensorrt/include",
                 dir_path + "/../bazel-tensorrt/external/tensorrt/include",
                 dir_path + "/../",
+                "/usr/local/cuda",
             ],
             extra_compile_args=[
                 "-Wno-deprecated",

toolchains/ci_workspaces/WORKSPACE.x86_64.release.rhel → toolchains/ci_workspaces/WORKSPACE.x86_64.cu118.release.rhel

@@ -43,7 +43,7 @@ local_repository(
 new_local_repository(
     name = "cuda",
     build_file = "@//third_party/cuda:BUILD",
-    path = "/usr/local/cuda-12.1",
+    path = "/usr/local/cuda-11.8",
 )
 
 new_local_repository(
@@ -58,17 +58,17 @@ new_local_repository(
 http_archive(
     name = "libtorch",
     build_file = "@//third_party/libtorch:BUILD",
-    sha256 = "1ae8366aaf7af7f68f142ba644fe26c837c6fa8347ec6bd9ce605ac60e7f7e5e",
+    sha256 = "174579a7ee2a506d063714160c5fc57da428f7935311ef511c8f19820eb14c86",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/nightly/cu121/libtorch-cxx11-abi-shared-with-deps-2.1.0.dev20230703%2Bcu121.zip"],
+    urls = ["https://download.pytorch.org/libtorch/nightly/cu121/libtorch-cxx11-abi-shared-with-deps-2.1.0.dev20230731%2Bcu121.zip"],
 )
 
 http_archive(
     name = "libtorch_pre_cxx11_abi",
     build_file = "@//third_party/libtorch:BUILD",
-    sha256 = "9add4832f4da9223866d85810820b816ab3319d5a227066101eeb6cbb76adb4b",
+    sha256 = "532217063c65354d5534211badadc9c370d889cb1c3fdb295c9b3d0f181bc0ba",
     strip_prefix = "libtorch",
-    urls = ["https://download.pytorch.org/libtorch/nightly/cu121/libtorch-shared-with-deps-2.1.0.dev20230703%2Bcu121.zip"],
+    urls = ["https://download.pytorch.org/libtorch/nightly/cu121/libtorch-shared-with-deps-2.1.0.dev20230731%2Bcu121.zip"],
 )
 
 ####################################################################################