This note records steps to setup coding environment from scratch.
Install Anaconda distribution.
Create virtual environment
It is better to create a virtual environment to avoid possible clash of version requirement for different projects. You can create virtual environment using vanilla python environment or using Anaconda environment.
Vanilla Python Environment
Alternatively, you may use python virtualenv to create virtual environment. To do this, run the following:
cd /Users/elynn/Dropbox/__openAI/openai sudo pip install virtualenv # This may already be installed virtualenv -p python3 .env # Create a virtual environment (python3) # Note: you can also use "virtualenv .env" to use your default python (please note we support 3.6) source .env/bin/activate # Activate the virtual environment pip install -r requirements.txt # Install dependencies # Work on the assignment for a while ... deactivate # Exit the virtual environment
To set up a virtual environment, run in a terminal (option + command + ~ in VSCode):
conda create -n openai python=3.6 anaconda
to create an environment called
Then, to activate and enter the environment, run
conda activate openai
To exit, you can simply close the window, or run
conda deactivate openai
To see a list of all your environments, in your terminal window or an Anaconda Prompt, run:
conda info --envs
conda env list
To remove a specific environment, use the follow command in the terminal window:
conda remove --name myenv --all
You may refer to this page for more detailed instructions on managing virtual environments with Anaconda.
But there is a difference between
pip intall and
conda install, Because Conda introduces a new packaging format, you cannot use pip and Conda interchangeably; pip cannot install the Conda package format. You can use the two tools side by side (by installing pip with conda install pip) but they do not interoperate either. See this post for more details.
# Current release for CPU-only pip install tensorflow # GPU pakage for CUDA-enabled GPU cards pip install tensorflow-gpu
pip unintall <packagename> fo uninstall a package.
conda install tensorflow
conda install pytorch torchvision -c pytorch
More information is provided by PyTorch website.
To ensure that PyTorch was installed correctly, we can verify the installation by running sample PyTorch code. The following code construct a randomly initialized tensor.
from __future__ import print_function import torch x = torch.rand(5,3) print(x)
Well, it did not work for me for the first time. Running the test code returns “ModuleNotFoundError: No module named ‘torch’” error.
One possible reseaon is that conda package is not uptodate. So one solution is to run the following code before installing PyTorch.
conda update conda conda install mkl=2018
For me, however, it is because the default version of Python is somehow set to 2.7 instead of 3.7. One can reset the default python to 3.7 (I don’t know how) or uninstall and reinstall Anaconda3.
Other Useful Commands
- Check version of python in an environment, first activate and enter the environment, then type:
python --version # Python 2.7.10 python3 --version # Python 3.6.7 :: Anaconda, Inc
In most of the cases, it is better to work remotely.