HemingwayLee / pytorch-cheatsheet

pytorch-cheatsheet

Tips

• UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow

# Slow 
torch.tensor([1, 2, 3])
# Better
torch.tensor(np.array([1, 2, 3]))

• Run with docker on GPU machine

docker run -it --rm --gpus all ...

• Size of runtime images < size of devel images

  • runtime: extends the base image by adding all the shared libraries from the CUDA toolkit. Use this image if you have a pre-built application using multiple CUDA libraries.
  • devel: extends the runtime image by adding the compiler toolchain, the debugging tools, the headers and the static libraries. Use this image to compile a CUDA application from sources.

• Check if run with GPU

import torch
torch.cuda.is_available()

• If it show False

  • run docker with --gpus all options
  • install driver by executing sudo apt install nvidia-driver-410
  • install cuda version of pytorch with pip
    • pip3 install torch==1.10.0+cu113 torchvision==0.11.0+cu113 --extra-index-url https://download.pytorch.org/whl/cu113

• Check if installed pytroch cpu only

$> conda list
pytorch                   1.13.1          cpu_py310hd11e9c7_1    conda-forge