136
votes

the default location for packages is .conda folder in my home directory. however, on the server I am using, there is a very strict limit of how much space I can use, which basically avoids me from putting anything under my home directory. how can I specify the location for the virtual environment that I want to create? Thanks! server is running Ubuntu.

7
Is there something wrong with the answer or why have you not accepted it? You can always leave comments if there is something you don't like. - Charlie Parker
@linX Always be sure to upvote answers you like, and accept the answer which is best or which solves your issue. This gives the authors a minor award and incentivizes them to continue to give well-thought and helpful answers. - Mike Williamson

7 Answers

146
votes

Use the --prefix or -p option to specify where to write the environment files. For example:

conda create --prefix /tmp/test-env python=2.7

Will create the environment named /tmp/test-env which resides in /tmp/ instead of the default .conda.

63
votes

like Paul said, use

conda create --prefix=/users/.../yourEnvName python=x.x

if you are located in the folder in which you want to create your virtual environment, just omit the path and use

conda create --prefix=yourEnvName python=x.x

conda only keep track of the environments included in the folder envs inside the anaconda folder. The next time you will need to activate your new env, move to the folder where you created it and activate it with

source activate yourEnvName
20
votes

While using the --prefix option works, you have to explicitly use it every time you create an environment. If you just want your environments stored somewhere else by default, you can configure it in your .condarc file.

Please see: https://conda.io/docs/user-guide/configuration/use-condarc.html#specify-environment-directories-envs-dirs

19
votes

If you want to use the --prefix or -p arguments, but want to avoid having to use the environment's full path to activate it, you need to edit the .condarc config file before you create the environment.

The .condarc file is in the home directory; C:\Users\<user> on Windows. Edit the values under the envs_dirs key to include the custom path for your environment. Assuming the custom path is D:\envs, the file should end up looking something like this:

ssl_verify: true
channels:
  - defaults
envs_dirs:
  - C:\Users\<user>\Anaconda3\envs
  - D:\envs

Then, when you create a new environment on that path, its name will appear along with the path when you run conda env list, and you should be able to activate it using only the name, and not the full path.

Command line screenshot

In summary, if you edit .condarc to include D:\envs, and then run conda env create -p D:\envs\myenv python=x.x, then activate myenv (or source activate myenv on Linux) should work.

Hope that helps!

P.S. I stumbled upon this through trial and error. I think what happens is when you edit the envs_dirs key, conda updates ~\.conda\environments.txt to include the environments found in all the directories specified under the envs_dirs, so they can be accessed without using absolute paths.

11
votes

You can create it like this

conda create --prefix C:/tensorflow2 python=3.7

and you don't have to move to that folder to activate it.

# To activate this environment, use:
# > activate C:\tensorflow2

As you see I do it like this.

D:\Development_Avector\PycharmProjects\TensorFlow>activate C:\tensorflow2

(C:\tensorflow2) D:\Development_Avector\PycharmProjects\TensorFlow>

(C:\tensorflow2) D:\Development_Avector\PycharmProjects\TensorFlow>conda --version
conda 4.5.13
4
votes

I ran into a similar situation. I did have access to a larger data drive. Depending on your situation, and the access you have to the server you can consider

ln -s /datavol/path/to/your/.conda /home/user/.conda

Then subsequent conda commands will put data to the symlinked dir in datavol

0
votes

Use -p option to specify the path to your env.

For Linux/macOS, env location info is stored in ~/.conda/environments.txt.

Use conda info --envs to list all your envs.