INSTALLATION ISSUES

[drivenmadbyapuzzle]

@Jumpat, thank you for your great research project.

Would it be possible to provide a pre-configured Docker image that encapsulates the necessary dependencies and environment to run the project's code successfully?

The installation process is not straightforward, and I've faced too many errors. I've tried various approaches, considering factors such as:

• gcc and g++ versions
• CUDA version
• NVIDIA drivers
• Python versions
• PyTorch versions
• and...

Despite my efforts, I couldn't successfully install it due to errors with different dependencies, particularly those related to packages installed with pip. A Docker image would greatly simplify the process and allow me to focus on utilizing your project.

[Jumpat]

Hi, due to the machine setting of our lab, we cannot build a docker image. But in practice we find conda is good enough for installation. You can specify the errors you met during the installation so that we can check whether we can help tackle them.

[Fabulani]

Hello! I'm writing a Dockerfile for this project and also facing installation issues. It seems my conda install hangs forever trying to solve the environment:

Collecting package metadata (repodata.json): \ WARNING conda.models.version:get_matcher(541): Using .* with relational operator is superfluous and deprecated and will be removed in a future version of conda. Your spec was 1.9.0.*, but conda is ignoring the .* and treating it as 1.9.0
WARNING conda.models.version:get_matcher(541): Using .* with relational operator is superfluous and deprecated and will be removed in a future version of conda. Your spec was 1.8.0.*, but conda is ignoring the .* and treating it as 1.8.0
WARNING conda.models.version:get_matcher(541): Using .* with relational operator is superfluous and deprecated and will be removed in a future version of conda. Your spec was 1.6.0.*, but conda is ignoring the .* and treating it as 1.6.0
WARNING conda.models.version:get_matcher(541): Using .* with relational operator is superfluous and deprecated and will be removed in a future version of conda. Your spec was 1.7.1.*, but conda is ignoring the .* and treating it as 1.7.1
done
Solving environment: / 

It has been like this for over 2 hours now. I suppose this is not normal.

Here's my current Dockerfile:

FROM pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel


# Build arguments
ARG TORCH_CUDA_ARCH_LIST="7.5+PTX"
ENV TORCH_CUDA_ARCH_LIST=${TORCH_CUDA_ARCH_LIST}
ARG COMMIT_ID="4acdaa6"
ENV COMMIT_ID=${COMMIT_ID}

WORKDIR /workspace

RUN apt-get update && apt-get install -y --no-install-recommends \
    git \
    wget \
    # Cleanup
    && apt-get autoremove -y && apt-get clean -y && rm -rf /var/lib/apt/lists

RUN git clone https://github.com/Jumpat/SegAnyGAussians.git && \
    cd SegAnyGAussians && \
    git checkout ${COMMIT_ID}

# Download ViT-H model for SAM
RUN mkdir -p /workspace/SegAnyGAussians/third_party/segment-anything/sam_ckpt && \
    wget -O /workspace/SegAnyGAussians/third_party/segment-anything/sam_ckpt/sam_vit_h_4b8939.pth \
    https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth

WORKDIR /workspace/SegAnyGAussians

ENV CONDA_ALWAYS_YES=true

# Fix for this issue: https://github.com/Jumpat/SegAnyGAussians/issues/5#issuecomment-1972428432
RUN conda install https://anaconda.org/pytorch3d/pytorch3d/0.7.4/download/linux-64/pytorch3d-0.7.4-py39_cu113_pyt1121.tar.bz2

RUN conda env create --file environment.yml && \
    conda activate gaussian_splatting

[Jumpat]

Hi @Fabulani , unfortunately I'm not sure what happened with the conda. There may be many possible causes for this, like network error, bad conda solver, ...

[Fabulani]

Could you give me an estimate of how long it takes to set up the environment with conda in your setup?

I suspect my issue is due to insufficient RAM. Will do some more tests on my side and hopefully figure it out.