tink/nerfstudio
4
0
Fork 0
mirror of https://github.com/nerfstudio-project/nerfstudio.git synced 2026-11-10 18:17:31 +08:00
Code

Update dockerfile and add docker build action (#3283)

Browse Source 
* Use github docker registry

* Fix DDP train for GPU in exclusive mode

* Improve docker image - compile gsplat, decrease image size

* Drop unrelated change

* Add build docker image action

* Rename build docker image action

* nit

* Remove commented line from Dockerfile

* Fix dockerfile when explicit source is specified

* Fix failing dynamo build for torch.compile

* Lock dockerfile and tcnn versions

* Add `torch.cuda.is_available()` condition

* Drop set_cuda_device

* Fix build docker image github action

* Docker build save disk space

* Set MAX_JOBS to limit resource usage for docker build

* Try bumping `MAX_JOBS` 2 => 4

* Install fixed gsplat version from nerfstudio's pyproject.toml

* Update docs

* Finish docker build action

* Fix ignores push when PR

---------

Co-authored-by: Gina Wu <ginawu98@gmail.com>
Co-authored-by: Brent Yi <yibrenth@gmail.com>
This commit is contained in:
Jonáš Kulhánek 2024-09-05 10:12:21 +02:00 committed by GitHub
parent 96b7fe2230
commit f86dbe6d45
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 188 additions and 189 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

62
.github/workflows/build_docker_image.yml vendored Normal file
View File

@ -0,0 +1,62 @@
name: Build Docker Image
on:
workflow_dispatch:
workflow_call:
pull_request:
push:
branches:
- main
- master
tags:
- 'v*'
env:
REGISTRY: ghcr.io
IMAGE_NAME: ${{ github.repository }}
jobs:
build-and-publish-docker-image:
runs-on: ubuntu-latest
name: build-and-publish-docker-image
permissions:
packages: write
contents: read
attestations: write
id-token: write
steps:
- uses: actions/checkout@v4
- name: Login to GitHub Container Registry
uses: docker/login-action@v3
with:
registry: ghcr.io
username: ${{ github.repository_owner }}
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 }}
tags: |
type=ref,event=branch
type=ref,event=pr
type=semver,pattern={{version}}
type=semver,pattern={{major}}.{{minor}}
- name: Free root space
uses: almahmoud/free-root-space@main
with:
remove-gcc: false
remove-cplusplus: false
- name: Build and push Docker image
id: push
uses: docker/build-push-action@3b5e8027fcad23fda98b2e3ac259d8d67585f671
with:
context: .
file: ./Dockerfile
push: ${{ github.event_name != 'pull_request' }}
tags: ${{ steps.meta.outputs.tags }}
labels: ${{ steps.meta.outputs.labels }}
- name: Generate artifact attestation
uses: actions/attest-build-provenance@v1
with:
subject-name: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME}}
subject-digest: ${{ steps.push.outputs.digest }}
push-to-registry: ${{ github.event_name != 'pull_request' }}

296
Dockerfile
View File

@ -1,185 +1,133 @@
ARG CUDA_VERSION=11.8.0
ARG OS_VERSION=22.04
# Define base image.
FROM nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${OS_VERSION}
ARG CUDA_VERSION
ARG OS_VERSION
# syntax=docker/dockerfile:1
ARG UBUNTU_VERSION=22.04
ARG NVIDIA_CUDA_VERSION=11.8.0
# CUDA architectures, required by Colmap and tiny-cuda-nn. Use >= 8.0 for faster TCNN.
ARG CUDA_ARCHITECTURES="90;89;86;80;75;70;61"
ARG NERFSTUDIO_VERSION=""
# Define username, user uid and gid
ARG USERNAME=user
ARG USER_UID=1000
ARG USER_GID=$USER_UID
# Pull source either provided or from git.
FROM scratch as source_copy
ONBUILD COPY . /tmp/nerfstudio
FROM alpine/git as source_no_copy
ARG NERFSTUDIO_VERSION
ONBUILD RUN git clone --branch ${NERFSTUDIO_VERSION} --recursive https://github.com/nerfstudio-project/nerfstudio.git /tmp/nerfstudio
ARG NERFSTUDIO_VERSION
FROM source_${NERFSTUDIO_VERSION:+no_}copy as source
FROM nvidia/cuda:${NVIDIA_CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION} as builder
ARG CUDA_ARCHITECTURES
ARG NVIDIA_CUDA_VERSION
ARG UBUNTU_VERSION
ENV DEBIAN_FRONTEND=noninteractive
ENV QT_XCB_GL_INTEGRATION=xcb_egl
RUN apt-get update && \
apt-get install -y --no-install-recommends --no-install-suggests \
git \
cmake \
ninja-build \
build-essential \
libboost-program-options-dev \
libboost-filesystem-dev \
libboost-graph-dev \
libboost-system-dev \
libeigen3-dev \
libflann-dev \
libfreeimage-dev \
libmetis-dev \
libgoogle-glog-dev \
libgtest-dev \
libsqlite3-dev \
libglew-dev \
qtbase5-dev \
libqt5opengl5-dev \
libcgal-dev \
libceres-dev \
python3.10-dev \
python3-pip
# Build and install COLMAP.
RUN git clone https://github.com/colmap/colmap.git && \
cd colmap && \
git checkout "3.9.1" && \
mkdir build && \
cd build && \
mkdir -p /build && \
cmake .. -GNinja "-DCMAKE_CUDA_ARCHITECTURES=${CUDA_ARCHITECTURES}" \
-DCMAKE_INSTALL_PREFIX=/build/colmap && \
ninja install -j1 && \
cd ~
# Upgrade pip and install dependencies.
# pip install torch==2.2.2 torchvision==0.17.2 --index-url https://download.pytorch.org/whl/cu118 && \
RUN pip install --no-cache-dir --upgrade pip 'setuptools<70.0.0' && \
pip install --no-cache-dir torch==2.1.2+cu118 torchvision==0.16.2+cu118 'numpy<2.0.0' --extra-index-url https://download.pytorch.org/whl/cu118 && \
git clone --branch master --recursive https://github.com/cvg/Hierarchical-Localization.git /opt/hloc && \
cd /opt/hloc && git checkout v1.4 && python3.10 -m pip install --no-cache-dir . && cd ~ && \
TCNN_CUDA_ARCHITECTURES="${CUDA_ARCHITECTURES}" pip install --no-cache-dir "git+https://github.com/NVlabs/tiny-cuda-nn.git@b3473c81396fe927293bdfd5a6be32df8769927c#subdirectory=bindings/torch" && \
pip install --no-cache-dir pycolmap==0.6.1 pyceres==2.1 omegaconf==2.3.0
# Install gsplat and nerfstudio.
# NOTE: both are installed jointly in order to prevent docker cache with latest
# gsplat version (we do not expliticly specify the commit hash).
#
# We set MAX_JOBS to reduce resource usage for GH actions:
# - https://github.com/nerfstudio-project/gsplat/blob/db444b904976d6e01e79b736dd89a1070b0ee1d0/setup.py#L13-L23
COPY --from=source /tmp/nerfstudio/ /tmp/nerfstudio
RUN export TORCH_CUDA_ARCH_LIST="$(echo "$CUDA_ARCHITECTURES" | tr ';' '\n' | awk '$0 > 70 {print substr($0,1,1)"."substr($0,2)}' | tr '\n' ' ' | sed 's/ $//')" && \
export MAX_JOBS=4 && \
GSPLAT_VERSION="$(sed -n 's/.*gsplat==\s*\([^," '"'"']*\).*/\1/p' /tmp/nerfstudio/pyproject.toml)" && \
pip install --no-cache-dir git+https://github.com/nerfstudio-project/gsplat.git@v${GSPLAT_VERSION} && \
pip install --no-cache-dir /tmp/nerfstudio 'numpy<2.0.0' && \
rm -rf /tmp/nerfstudio
# Fix permissions
RUN chmod -R go=u /usr/local/lib/python3.10 && \
chmod -R go=u /build
#
# Docker runtime stage.
#
FROM nvidia/cuda:${NVIDIA_CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION} as runtime
ARG CUDA_ARCHITECTURES
ARG NVIDIA_CUDA_VERSION
ARG UBUNTU_VERSION
# metainformation
LABEL org.opencontainers.image.version = "0.1.18"
LABEL org.opencontainers.image.source = "https://github.com/nerfstudio-project/nerfstudio"
LABEL org.opencontainers.image.licenses = "Apache License 2.0"
LABEL org.opencontainers.image.base.name="docker.io/library/nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${OS_VERSION}"
LABEL org.opencontainers.image.base.name="docker.io/library/nvidia/cuda:${NVIDIA_CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}"
LABEL org.opencontainers.image.documentation = "https://docs.nerf.studio/"
# Variables used at build time.
## CUDA architectures, required by Colmap and tiny-cuda-nn.
## NOTE: All commonly used GPU architectures are included and supported here. To speedup the image build process remove all architectures but the one of your explicit GPU. Find details here: https://developer.nvidia.com/cuda-gpus (8.6 translates to 86 in the line below) or in the docs.
ARG CUDA_ARCHITECTURES=90;89;86;80;75;70;61;52;37
# Set environment variables.
## Set non-interactive to prevent asking for user inputs blocking image creation.
ENV DEBIAN_FRONTEND=noninteractive
## Set timezone as it is required by some packages.
ENV TZ=Europe/Berlin
## CUDA Home, required to find CUDA in some packages.
ENV CUDA_HOME="/usr/local/cuda"
# Install required apt packages and clear cache afterwards.
# Minimal dependencies to run COLMAP binary compiled in the builder stage.
# Note: this reduces the size of the final image considerably, since all the
# build dependencies are not needed.
RUN apt-get update && \
apt-get install -y --no-install-recommends \
build-essential \
cmake \
curl \
ffmpeg \
git \
libatlas-base-dev \
libboost-filesystem-dev \
libboost-graph-dev \
libboost-program-options-dev \
libboost-system-dev \
libboost-test-dev \
libhdf5-dev \
libcgal-dev \
libeigen3-dev \
libflann-dev \
libfreeimage-dev \
libgflags-dev \
libglew-dev \
libgoogle-glog-dev \
libmetis-dev \
libprotobuf-dev \
libqt5opengl5-dev \
libsqlite3-dev \
libsuitesparse-dev \
nano \
protobuf-compiler \
python-is-python3 \
python3.10-dev \
python3-pip \
qtbase5-dev \
sudo \
vim-tiny \
wget && \
rm -rf /var/lib/apt/lists/*
apt-get install -y --no-install-recommends --no-install-suggests \
libboost-filesystem1.74.0 \
libboost-program-options1.74.0 \
libc6 \
libceres2 \
libfreeimage3 \
libgcc-s1 \
libgl1 \
libglew2.2 \
libgoogle-glog0v5 \
libqt5core5a \
libqt5gui5 \
libqt5widgets5 \
python3.10 \
python3.10-dev \
build-essential \
python-is-python3 \
ffmpeg
# Copy packages from builder stage.
COPY --from=builder /build/colmap/ /usr/local/
COPY --from=builder /usr/local/lib/python3.10/dist-packages/ /usr/local/lib/python3.10/dist-packages/
COPY --from=builder /usr/local/bin/ns* /usr/local/bin/
# Install GLOG (required by ceres).
RUN git clone --branch v0.6.0 https://github.com/google/glog.git --single-branch && \
cd glog && \
mkdir build && \
cd build && \
cmake .. && \
make -j `nproc` && \
make install && \
cd ../.. && \
rm -rf glog
# Add glog path to LD_LIBRARY_PATH.
ENV LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:/usr/local/lib"
# Install Ceres-solver (required by colmap).
RUN git clone --branch 2.1.0 https://ceres-solver.googlesource.com/ceres-solver.git --single-branch && \
cd ceres-solver && \
git checkout $(git describe --tags) && \
mkdir build && \
cd build && \
cmake .. -DBUILD_TESTING=OFF -DBUILD_EXAMPLES=OFF && \
make -j `nproc` && \
make install && \
cd ../.. && \
rm -rf ceres-solver
# Install colmap.
RUN git clone --branch 3.8 https://github.com/colmap/colmap.git --single-branch && \
cd colmap && \
mkdir build && \
cd build && \
cmake .. -DCUDA_ENABLED=ON \
-DCMAKE_CUDA_ARCHITECTURES=${CUDA_ARCHITECTURES} && \
make -j `nproc` && \
make install && \
cd ../.. && \
rm -rf colmap
# Create non root user, add it to custom group and setup environment.
RUN groupadd --gid $USER_GID $USERNAME \
&& useradd --uid $USER_UID --gid $USER_GID -m $USERNAME -d /home/${USERNAME} --shell /usr/bin/bash
# OPTIONAL
# If sudo privilages are not required comment below line
# Create simple password for user and add it to sudo group
# Update group so that it is not required to type password for commands: apt update/upgrade/install/remove
RUN echo "${USERNAME}:password" | chpasswd \
&& usermod -aG sudo ${USERNAME} \
&& echo "%sudo ALL=NOPASSWD:/usr/bin/apt-get update, /usr/bin/apt-get upgrade, /usr/bin/apt-get install, /usr/bin/apt-get remove" >> /etc/sudoers
# Create workspace folder and change ownership to new user
RUN mkdir /workspace && chown ${USER_UID}:${USER_GID} /workspace
# Switch to new user and workdir.
USER ${USER_UID}
WORKDIR /home/${USERNAME}
# Add local user binary folder to PATH variable.
ENV PATH="${PATH}:/home/${USERNAME}/.local/bin"
# Upgrade pip and install packages.
RUN python3.10 -m pip install --no-cache-dir --upgrade pip setuptools==69.5.1 pathtools promise pybind11 omegaconf
# Install pytorch and submodules
# echo "${CUDA_VERSION}" | sed 's/.$//' | tr -d '.' -- CUDA_VERSION -> delete last digit -> delete all '.'
RUN CUDA_VER=$(echo "${CUDA_VERSION}" | sed 's/.$//' | tr -d '.') && python3.10 -m pip install --no-cache-dir \
torch==2.1.2+cu${CUDA_VER} \
torchvision==0.16.2+cu${CUDA_VER} \
--extra-index-url https://download.pytorch.org/whl/cu${CUDA_VER}
# Install tiny-cuda-nn (we need to set the target architectures as environment variable first).
ENV TCNN_CUDA_ARCHITECTURES=${CUDA_ARCHITECTURES}
RUN python3.10 -m pip install --no-cache-dir git+https://github.com/NVlabs/tiny-cuda-nn.git#subdirectory=bindings/torch
# Install pycolmap, required by hloc.
RUN git clone --branch v0.4.0 --recursive https://github.com/colmap/pycolmap.git && \
cd pycolmap && \
python3.10 -m pip install --no-cache-dir . && \
cd ..
# Install hloc 1.4 as alternative feature detector and matcher option for nerfstudio.
RUN git clone --branch master --recursive https://github.com/cvg/Hierarchical-Localization.git && \
cd Hierarchical-Localization && \
git checkout v1.4 && \
python3.10 -m pip install --no-cache-dir -e . && \
cd ..
# Install pyceres from source
RUN git clone --branch v1.0 --recursive https://github.com/cvg/pyceres.git && \
cd pyceres && \
python3.10 -m pip install --no-cache-dir -e . && \
cd ..
# Install pixel perfect sfm.
RUN git clone --recursive https://github.com/cvg/pixel-perfect-sfm.git && \
cd pixel-perfect-sfm && \
git reset --hard 40f7c1339328b2a0c7cf71f76623fb848e0c0357 && \
git clean -df && \
python3.10 -m pip install --no-cache-dir -e . && \
cd ..
# Copy nerfstudio folder and give ownership to user.
COPY --chown=${USER_UID}:${USER_GID} . /home/${USERNAME}/nerfstudio
# Install nerfstudio dependencies.
RUN cd nerfstudio && \
python3.10 -m pip install --no-cache-dir -e . && \
cd ..
# Switch to workspace folder and install nerfstudio cli auto completion
WORKDIR /workspace
RUN ns-install-cli --mode install
# Install nerfstudio cli auto completion
RUN /bin/bash -c 'ns-install-cli --mode install'
# Bash as default entrypoint.
CMD /bin/bash -l
# Force changing password on first container run
# Change line above: CMD /bin/bash -l -> CMD /bin/bash -l -c passwd && /usr/bin/bash -l

19
docs/quickstart/installation.md
View File

@ -222,10 +222,10 @@ Instead of installing and compiling prerequisites, setting up the environment an
### Prerequisites
Docker ([get docker](https://docs.docker.com/get-docker/)) and nvidia GPU drivers ([get nvidia drivers](https://www.nvidia.de/Download/index.aspx?lang=de)), capable of working with CUDA 11.8, must be installed.
The docker image can then either be pulled from [here](https://hub.docker.com/r/dromni/nerfstudio/tags) (replace <version> with the actual version, e.g. 0.1.18)
The docker image can then either be pulled from [here](https://github.com/nerfstudio-project/nerfstudio/pkgs/container/nerfstudio) (`latest` can be replaced with a fixed version, e.g., `1.1.3`)
```bash
docker pull dromni/nerfstudio:<version>
docker pull ghcr.io/nerfstudio-project/nerfstudio:latest
```
or be built from the repository using
@ -240,22 +240,11 @@ For example, here's how to build with support for GeForce 30xx series GPUs:
```bash
docker build \
--build-arg CUDA_VERSION=11.8.0 \
--build-arg CUDA_ARCHITECTURES=86 \
--build-arg OS_VERSION=22.04 \
--tag nerfstudio-86 \
--file Dockerfile .
```
The user inside the container is called 'user' and is mapped to the local user with ID 1000 (usually the first non-root user on Linux systems).
If you suspect that your user might have a different id, override `USER_ID` during the build as follows:
```bash
docker build \
--build-arg USER_ID=$(id -u) \
--file Dockerfile .
```
### Using an interactive container
The docker container can be launched with an interactive terminal where nerfstudio commands can be entered as usual. Some parameters are required and some are strongly recommended for usage as following:
@ -269,7 +258,7 @@ docker run --gpus all \ # Give the conta
--rm \ # Remove container after it is closed (recommended).
-it \ # Start container in interactive mode.
--shm-size=12gb \ # Increase memory assigned to container to avoid memory limitations, default is 64 MB (recommended).
dromni/nerfstudio:<tag> # Docker image name if you pulled from docker hub.
ghcr.io/nerfstudio-project/nerfstudio:<tag> # Docker image name if you pulled from GitHub.
<--- OR --->
nerfstudio # Docker image tag if you built the image from the Dockerfile by yourself using the command from above.
```
@ -280,7 +269,7 @@ Besides, the container can also directly be used by adding the nerfstudio comman
```bash
docker run --gpus all -u $(id -u) -v /folder/of/your/data:/workspace/ -v /home/<YOUR_USER>/.cache/:/home/user/.cache/ -p 7007:7007 --rm -it --shm-size=12gb # Parameters.
dromni/nerfstudio:<tag> \ # Docker image name
ghcr.io/nerfstudio-project/nerfstudio:<tag> \ # Docker image name if you pulled from GitHub.
ns-process-data video --data /workspace/video.mp4 # Smaple command of nerfstudio.
```