Merge pull request #1 from AIPowerGrid/urek

Upgrading

.github/workflows/maintests.yml

@@ -26,6 +26,19 @@ jobs:
         with:
           extra_args: --all-files
 
+  dockerfile-lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+      - name: Lint CUDA Dockerfile
+        uses: hadolint/hadolint-action@master
+        with:
+          dockerfile: "Dockerfiles/Dockerfile.cuda"
+      - name: Lint RoCM Dockerfile
+        uses: hadolint/hadolint-action@master
+        with:
+          dockerfile: "Dockerfiles/Dockerfile.rocm"
+
   unit-tests:
     runs-on: ubuntu-latest
     strategy:

.github/workflows/prtests.yml

@@ -31,6 +31,19 @@ jobs:
         with:
           extra_args: --all-files
 
+  dockerfile-lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+      - name: Lint CUDA Dockerfile
+        uses: hadolint/hadolint-action@master
+        with:
+          dockerfile: "Dockerfiles/Dockerfile.cuda"
+      - name: Lint RoCM Dockerfile
+        uses: hadolint/hadolint-action@master
+        with:
+          dockerfile: "Dockerfiles/Dockerfile.rocm"
+
   unit-tests:
     runs-on: ubuntu-latest
     strategy:

.hadolint.yaml

@@ -0,0 +1,5 @@
+ignored:
+  - DL3008 # Pin versions in apt get install
+  - DL3042 # Avoid cache directory with `pip install`
+  - DL3002 # Last USER should not be root
+failure-threshold: warning

.pre-commit-config.yaml

@@ -1,20 +1,20 @@
 repos:
 -   repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v4.6.0
+    rev: v5.0.0
     hooks:
     -   id: check-yaml
     -   id: end-of-file-fixer
     -   id: trailing-whitespace
 -   repo: https://github.com/psf/black
-    rev: 24.4.2
+    rev: 24.10.0
     hooks:
     -   id: black
 -   repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.5.4
+    rev: v0.7.3
     hooks:
     -    id: ruff
 -   repo: https://github.com/pre-commit/mirrors-mypy
-    rev: 'v1.11.0'
+    rev: 'v1.13.0'
     hooks:
     -   id: mypy
         args: []
@@ -38,9 +38,9 @@ repos:
         - python-dotenv
         - aiohttp
         - horde_safety==0.2.3
-        - torch==2.3.1
+        - torch==2.5.0
         - ruamel.yaml
-        - horde_engine==2.15.3
-        - horde_sdk==0.14.11
-        - horde_model_reference==0.9.0
+        - horde_engine==2.18.1
+        - horde_sdk==0.16.4
+        - horde_model_reference==0.9.1
         - semver

Dockerfiles/Dockerfile.cuda

@@ -0,0 +1,71 @@
+# Stage 1: Base environment setup
+ARG CUDA_VERSION=12.4.1
+FROM nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu22.04 AS base
+
+SHELL ["/bin/bash", "-o", "pipefail", "-c"]
+
+ARG DEBIAN_FRONTEND=noninteractive
+ARG PYTHON_VERSION=3.11
+ENV PYTHON_VERSION=${PYTHON_VERSION}
+ENV APP_HOME=/horde-worker-reGen
+
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends software-properties-common && \
+    add-apt-repository ppa:deadsnakes/ppa && \
+    apt-get install -y --no-install-recommends \
+    python${PYTHON_VERSION} \
+    python3-pip \
+    python${PYTHON_VERSION}-venv \
+    libgl1 \
+    git && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+# Extract CUDA version for PyTorch
+ARG CUDA_VERSION
+RUN CUDA_VERSION_SHORT=$(echo "${CUDA_VERSION}" | cut -d. -f1-2 | tr -d '.') && \
+    echo "${CUDA_VERSION_SHORT}" && \
+    echo "export CUDA_VERSION_SHORT=${CUDA_VERSION_SHORT}" >> /env_vars
+
+# Stage 2: Clone repository and install dependencies
+FROM base AS builder
+
+ARG GIT_BRANCH=main
+ARG GIT_OWNER=Haidra-Org
+
+RUN echo "export GIT_BRANCH=${GIT_BRANCH}" >> /env_vars && \
+    echo "export GIT_OWNER=${GIT_OWNER}" >> /env_vars
+
+WORKDIR "${APP_HOME}"
+
+# Clone the repository
+RUN git clone "https://github.com/${GIT_OWNER}/horde-worker-reGen.git" . && \
+    git switch "${GIT_BRANCH}"
+
+# Create virtual environment
+RUN python"${PYTHON_VERSION}" -m venv "${APP_HOME}/venv"
+ENV PATH="${APP_HOME}/venv/bin:$PATH"
+
+# Install dependencies
+ARG PIP_CACHE_DIR=/pip-cache
+ARG USE_PIP_CACHE=true
+
+RUN --mount=type=cache,target="${PIP_CACHE_DIR}",sharing=locked,id=pip-cache \
+    . /env_vars && \
+    if [ "${USE_PIP_CACHE}" = "true" ]; then \
+        pip install --cache-dir="${PIP_CACHE_DIR}" opencv-python-headless -r requirements.txt -U --extra-index-url "https://download.pytorch.org/whl/cu${CUDA_VERSION_SHORT}"; \
+    else \
+        pip install opencv-python-headless -r requirements.txt -U --extra-index-url "https://download.pytorch.org/whl/cu${CUDA_VERSION_SHORT}"; \
+    fi
+
+# Stage 3: Final stage
+FROM builder AS final
+
+WORKDIR "${APP_HOME}"
+COPY entrypoint.sh /entrypoint.sh
+COPY setup_*.sh "${APP_HOME}"
+RUN chmod +x /entrypoint.sh
+
+STOPSIGNAL SIGINT
+
+ENTRYPOINT ["/entrypoint.sh"]

Dockerfiles/Dockerfile.rocm

@@ -0,0 +1,75 @@
+# Stage 1: Base environment setup
+ARG ROCM_VERSION=6.1.2
+FROM rocm/rocm-terminal:${ROCM_VERSION} AS base
+
+USER root
+WORKDIR /
+SHELL ["/bin/bash", "-o", "pipefail", "-c"]
+
+ARG DEBIAN_FRONTEND=noninteractive
+ARG PYTHON_VERSION=3.11
+ENV PYTHON_VERSION=${PYTHON_VERSION}
+ENV APP_HOME=/horde-worker-reGen
+
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends software-properties-common && \
+    add-apt-repository ppa:deadsnakes/ppa && \
+    apt-get install -y --no-install-recommends \
+    python${PYTHON_VERSION} \
+    python3-pip \
+    python${PYTHON_VERSION}-venv \
+    python${PYTHON_VERSION}-dev \
+    python${PYTHON_VERSION}-distutils \
+    ninja-build \
+    rocm \
+    git && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+# Extract ROCm version for PyTorch
+ARG ROCM_VERSION
+RUN ROCM_VERSION_SHORT=$(echo "${ROCM_VERSION}" | cut -d. -f1-2) && \
+    echo "export ROCM_VERSION_SHORT=${ROCM_VERSION_SHORT}" >> /env_vars
+
+# Stage 2: Clone repository and install dependencies
+FROM base AS builder
+
+ARG GIT_BRANCH=main
+ARG GIT_OWNER=Haidra-Org
+
+RUN echo "export GIT_BRANCH=${GIT_BRANCH}" >> /env_vars && \
+    echo "export GIT_OWNER=${GIT_OWNER}" >> /env_vars
+
+WORKDIR "${APP_HOME}"
+
+# Clone the repository
+RUN git clone "https://github.com/${GIT_OWNER}/horde-worker-reGen.git" . && \
+    git switch "${GIT_BRANCH}"
+
+# Create virtual environment
+RUN python"${PYTHON_VERSION}" -m venv "${APP_HOME}/venv"
+ENV PATH="${APP_HOME}/venv/bin:$PATH"
+
+# Install dependencies
+ARG PIP_CACHE_DIR=/pip-cache
+ARG USE_PIP_CACHE=true
+
+RUN --mount=type=cache,target="${PIP_CACHE_DIR}",sharing=locked,id=pip-cache \
+    . /env_vars && \
+    if [ "${USE_PIP_CACHE}" = "true" ]; then \
+        pip install --cache-dir="${PIP_CACHE_DIR}" opencv-python-headless -r requirements.rocm.txt -U --extra-index-url "https://download.pytorch.org/whl/rocm${ROCM_VERSION_SHORT}"; \
+    else \
+        pip install opencv-python-headless -r requirements.rocm.txt -U --extra-index-url "https://download.pytorch.org/whl/rocm${ROCM_VERSION_SHORT}"; \
+    fi && \
+    pip uninstall -y pynvml nvidia-ml-py
+# Stage 3: Final stage
+FROM builder AS final
+
+WORKDIR "${APP_HOME}"
+COPY entrypoint.sh /entrypoint.sh
+COPY setup_*.sh "${APP_HOME}"
+RUN chmod +x /entrypoint.sh
+
+STOPSIGNAL SIGINT
+
+ENTRYPOINT ["/entrypoint.sh"]