Deep Learning Environment

Currently we find that multiple version of CUDA could be installed on windows. And different frameworks could use different CUDA version nicely together.

PyTorch Tensorflow Co-environment

Currently, we can have

  • pytorch 1.3 torchvision 0.4.2and tensorflow 1.15 live together,
  • with
    • Python 3.7
    • CUDA 10.0 and CUDA 10.1 both installed.
    • NVIDIA Driver Version: 432.00.
    • numpy version 1.16.4
  • cudnn version
    • <unknown>::cudnn-7.6.4-cuda10.0_0 --> anaconda::cudnn-7.6.5-cuda10.1_0

Keras Tensorflow co-environment

If we want keras as well, refer to this post for installation guide on keras-gpu installation on windows.

Note, never open 2 tf instances at once on a computer, if so, try to kill the new tf by using nvidia-smi and kill

PyTorch Caffe Co-environment

See Working with Caffe for more info!

The easiest way is using the the conda export file to copy the same environment. See the attached caffe-torch tf-torch.yml for more information.

conda create --name caffe36 --file caffe36_spec-file.txt
conda create -f caffe-torch.yml

Then it will download and install all the packages with the required source and version

conda activate caffe36

To export such files there are usually 2 ways, switch to the environment you wish to export and use these commands.

conda list --explicit > spec-file.txt
conda env export > environment.yml

https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html#

Post on export environment with conda.

GFW evading

If you are in China and you need to use a GFW censorship evading tool, see this,

https://blog.hiaoxui.com/blog/post/hiaoxui/ssr-manual-zh