如何使用深度学习平台发布的虚拟机映像获取 NVIDIA 驱动程序?

Question

我遇到了需要安装 NVIDIA 驱动程序的问题。

我最初基于以下内容创建了一个计算引擎 VM：

export IMAGE_FAMILY="pytorch-latest-cu100"
export ZONE="us-west1-b"
export INSTANCE_NAME="my-instance"
gcloud compute instances create $INSTANCE_NAME 
--zone=$ZONE 
--image-family=$IMAGE_FAMILY 
--image-project=deeplearning-platform-release 
--maintenance-policy=TERMINATE 
--accelerator="type=nvidia-tesla-v100,count=1" 
--metadata="install-nvidia-driver=True"`

在此 VM 上部署的代码工作正常。 现在我需要在上面创建一个 REST API 层，因此根据这一点，我需要使用 docker 对应用程序进行容器化。

我尝试从以下位置构建我的 docker 映像： gcr.io/deeplearning-platform-release/pytorch-latest-cu100 (来自上面的命令(但似乎这个图像不存在。

然后我尝试从以下位置构建另一个映像： gcr.io/deeplearning-platform-release/pytorch-gpu.1-1

但是现在当我运行代码时，出现以下错误：

Traceback (most recent call last):
File "model.py", line 297, in run
data = main(filepath)
File "model.py", line 52, in main
model = model.cuda()
File "/root/miniconda3/lib/python3.7/site- 
packages/torch/nn/modules/module.py", line 260, in cuda
return self._apply(lambda t: t.cuda(device))
File "/root/miniconda3/lib/python3.7/site- 
packages/torch/nn/modules/module.py", line 187, in _apply
module._apply(fn)
File "/root/miniconda3/lib/python3.7/site- 
packages/torch/nn/modules/module.py", line 187, in _apply
module._apply(fn)
File "/root/miniconda3/lib/python3.7/site- 
packages/torch/nn/modules/module.py", line 187, in _apply
module._apply(fn)
File "/root/miniconda3/lib/python3.7/site- 
packages/torch/nn/modules/module.py", line 193, in _apply
param.data = fn(param.data)
File "/root/miniconda3/lib/python3.7/site- 
packages/torch/nn/modules/module.py", line 260, in <lambda>
return self._apply(lambda t: t.cuda(device))
File "/root/miniconda3/lib/python3.7/site- 
packages/torch/cuda/__init__.py", line 161, in _lazy_init
_check_driver()
File "/root/miniconda3/lib/python3.7/site- 
packages/torch/cuda/__init__.py", line 82, in _check_driver
http://www.nvidia.com/Download/index.aspx""")
AssertionError: 
Found no NVIDIA driver on your system. Please check that you
have an NVIDIA GPU and installed a driver from
http://www.nvidia.com/Download/index.aspx

我的 Dockerfile：

FROM gcr.io/deeplearning-platform-release/pytorch-gpu.1-1
WORKDIR /app
COPY requirements.txt /app
RUN pip install --no-cache-dir -r requirements.txt
EXPOSE 8080
COPY . /app/
CMD [ "python","main.py" ]

我 main.py：

from flask import Flask, request
import model
app = Flask(__name__)
@app.route('/getduration', methods=['POST'])
def get_duration():
try:
data = request.args.get('param')
except:
data = None
try:
duration = model.run(data)
return duration, 200
except Exception as e:
error = f"There was an error: {e}"
return error, 500
if __name__ == '__main__':
app.run(host='127.0.0.1', port=8080, debug=True)

如何更新我的 Dockerfile，以便我可以使用 Nvidia 驱动程序？

Answer 1

你使用的是 NVIDIA Docker 嗎？如果没有，那可能是你的问题。完全按照docker使用nvidia-docker，这将使 NVIDIA 驱动程序在您的容器内可用。

https://github.com/NVIDIA/nvidia-docker