NVIDIA-Linux-x86_64-470.63.01-grid.run部署到云运行的问题

Question

我正试图在google云上部署。我的Dockerfile：

FROM ubuntu:20.04
RUN apt update 
RUN apt -y install sudo
RUN apt -y install curl
RUN sudo apt install -y build-essential 
RUN curl -O https://storage.googleapis.com/nvidia-drivers-us-public/GRID/GRID13.0/NVIDIA-Linux-x86_64-470.63.01-grid.run
RUN chmod +x NVIDIA-Linux-x86_64-470.63.01-grid.run
RUN sudo /bin/bash NVIDIA-Linux-x86_64-470.63.01-grid.run

FROM python:3.8

RUN adduser meat
RUN passwd -d meat
USER meat
WORKDIR /home/meat
RUN python3 -m venv meat-env
RUN /bin/bash -c "source meat-env/bin/activate"
RUN /usr/local/bin/python -m pip install --upgrade pip
COPY requirements.txt .
RUN pip3 install -r requirements.txt
COPY . .
CMD ["python", "main.py"]

我得到了一个错误：

Step 8/20 : RUN sudo /bin/bash NVIDIA-Linux-x86_64-470.63.01-grid.run
---> Running in 811998f9cea8
Verifying archive integrity...
OK
Uncompressing NVIDIA Accelerated Graphics Driver for Linux-x86_64 470.63.01

...............

和几个点(.)后来

[91mError opening terminal: unknown.
[0m
unable to stream build output: The command '/bin/sh -c sudo /bin/bash NVIDIA-Linux-x86_64-470.63.01-grid.run' returned a non-zero code: 1
Failed to build the app. Error: unable to stream build output: The command '/bin/sh -c sudo /bin/bash NVIDIA-Linux-x86_64-470.63.01-grid.run' returned a non-zero code: 1

我的Dockerfile出了什么问题？还是有不同的方式来执行命令？

Answer 1

@DazWilkin的评论是正确的：

不可能将图形驱动程序安装到完全管理的云运行中，因为它不公开任何GPU。

您应该在(GKE)中部署它。但是，您需要配置GKE集群来安装GPU，根据文档，您应该遵循以下步骤：

1. Add a GPU-enabled node pool to your GKE cluster。

在此步骤中，您可以启用虚拟工作站(NVIDIA网格)。请选择GPU，如NVIDIA T4、P4或P100来启用NVIDIA网格。

1. Install NVIDIA's device drivers to the nodes。

现在，您可以创建一个将使用GPU的服务，并将映像部署到为Anthos运行的云中：

1. Setting up your service to consume GPUs