[Docker] add image-building workflow and update docker file (#18)

This PR incudes:
1. a workflow for building docker image and release it to github.
2. some docker file updates

.github/workflows/docker_image.yml

@@ -0,0 +1,47 @@
+name: Docker Image CI
+
+on:
+  workflow_dispatch:
+
+env:
+  REGISTRY: ghcr.io
+  IMAGE_NAME: ${{ github.repository }}
+
+jobs:
+  build-and-push-image:
+    runs-on: [self-hosted, Linux, X64, Docker]
+
+    permissions:
+      contents: read
+      packages: write
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Log in to the Container registry
+        uses: docker/login-action@65b78e6e13532edd9afa3aa52ac7964289d1a9c1
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract metadata (tags, labels) for Docker
+        id: meta
+        uses: docker/metadata-action@9ec57ed1fcdbf14dcef7dfbe97b2010124a938b7
+        with:
+          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+          flavor: |
+            latest=auto
+          tags: |
+            type=sha,format=long
+            type=raw,value=latest,enable={{is_default_branch}}
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@f2a1d5e99d037542a71f64918e516c093c6f3fc4
+        with:
+          context: .
+          push: true
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+          build-args: MAX_JOBS=64

Dockerfile

@@ -1,36 +1,45 @@
 FROM nvidia/cuda:12.1.1-cudnn8-devel-ubuntu22.04
 
+ARG MAX_JOBS=32
+
 WORKDIR /app
 
 COPY ./patches /app
 
 ENV TZ=Etc/UTC \
     DEBIAN_FRONTEND=noninteractive \
-    NINJA_MAX_JOBS=32 \
-    MAX_JOBS=32 \
-    TORCH_CUDA_ARCH_LIST="8.0 8.6 8.9 9.0" 
+    NINJA_MAX_JOBS=$MAX_JOBS \
+    MAX_JOBS=$MAX_JOBS \
+    TORCH_CUDA_ARCH_LIST="8.0 8.6 8.9 9.0"\
+    TORCH_CXX_FLAGS="-D_GLIBCXX_USE_CXX11_ABI=1"\
+    _GLIBCXX_USE_CXX11_ABI=1\
+    PYTORCH_VERSION="2.2.0a0+git6c8c5ad"
 
 
 RUN apt-get update \
      && apt-get install -y software-properties-common \
-     && apt-get install -y python3.10 python3-pip git git-lfs cmake ninja-build \
+     && apt-get install -y python3.10 python3-pip git git-lfs cmake ninja-build build-essential gcc g++ \
      && apt-get clean \
      && rm -rf /var/lib/apt/lists/* \
      && ln -sf /usr/bin/python3.10 /usr/bin/python3 \
      && ln -sf /usr/bin/pip3.10 /usr/bin/pip3 \
      && pip3 install --upgrade pip setuptools wheel \
      && apt update \
-     && apt install -y python-is-python3
+     && apt install -y python-is-python3 openssh-server \
+     && apt clean
 
 RUN bash build_pytorch_w_patch.sh
 
+RUN bash build_torchvision.sh
+
 RUN bash build_torchdistX_w_patch.sh
 
 
 RUN pip3 install --no-cache-dir packaging \
     && pip3 install --no-cache-dir mpmath==1.3.0 \
     && pip3 install --no-cache-dir "setuptools>=69.0.0" \
     && pip3 install --no-cache-dir regex \
+    && pip3 install --no-cache-dir pillow \
     && pip3 install --no-cache-dir pybind11 \
     && pip3 install --no-cache-dir einops \
     && pip3 install --no-cache-dir expecttest \

patches/build_pytorch_w_patch.sh

@@ -1,4 +1,4 @@
-#! usr/bin/bash
+#!/usr/bin/bash
 set -e
 
 SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )"
@@ -13,7 +13,7 @@ fi
 
 git clone --branch v2.2.1-rc3 --depth 1 https://github.com/pytorch/pytorch.git
 pushd pytorch 
-git apply $PATCH_PATH
+git apply "$PATCH_PATH"
 git submodule sync 
 git submodule update --init --recursive --depth 1
 pip3 install -r requirements.txt

patches/build_torchdistX_w_patch.sh

@@ -1,4 +1,4 @@
-#! usr/bin/bash
+#!/usr/bin/bash
 set -e
 
 SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )"
@@ -15,7 +15,7 @@ git clone --depth 1 https://github.com/pytorch/torchdistx.git
 pushd torchdistx
 git pull
 git checkout 9c1b9f5cb2fa36bfb8b70ec07c40ed42a33cc87a
-git apply $PATCH_PATH
+git apply "$PATCH_PATH"
 git submodule sync
 git submodule update --init --recursive --depth 1
 cmake -DTORCHDIST_INSTALL_STANDALONE=ON  -GNinja -DCAKE_CXX_COMPILER_LAUNCHER=ccache -B build

patches/build_torchvision.sh

@@ -0,0 +1,11 @@
+#!/usr/bin/bash
+set -e
+
+# build torchvision
+git clone -b v0.17.0 https://github.com/pytorch/vision.git
+pushd vision
+python3 setup.py install
+popd
+rm -rf vision
+python3 -c "import torchvision"
+pip3 list | grep torch