RunpodR
Runpod•10mo ago
canxerian

No CUDA runtime is found, using CUDA_HOME='/usr/local/cuda

Hey everyone 👋

I'm trying to use Runpod serverless to run the https://github.com/nerfstudio-project/gsplat/ gaussian-splatting implementation. However, when building the project from source (pip install . of my Dockerfile below), I get the error:

No CUDA runtime is found, using CUDA_HOME='/usr/local/cuda

Is the CUDA runtime available during the build stage of a docker image on runpod serverless workers?

Thanks in advance!

Dockerfile:
FROM runpod/pytorch:2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04

WORKDIR /

# Python dependencies
COPY builder/requirements.txt /requirements.txt
RUN python -m pip install --upgrade pip && \
    python -m pip install --upgrade -r /requirements.txt --no-cache-dir && \
    rm /requirements.txt


# Install gsplat deps
RUN apt update && apt install libglm-dev g++-9 gcc-9 -y
RUN python -m pip install ninja numpy jaxtyping rich

# Install gsplat
RUN git clone https://github.com/nerfstudio-project/gsplat.git && \
    cd gsplat && \
    python -m pip install . && \
    cd examples && \
    python -m pip install -r requirements.txt
GitHub
CUDA accelerated rasterization of gaussian splatting - nerfstudio-project/gsplat
Was this page helpful?