Problem in removing a conda created environment

I installed tensorflow using conda create -n tf-gp tensorflow-gpu. Then I activated it using: conda activate tf-gpuThen I uninstalled tensorflow from this environment using:conda uninstall tensorflow

Then I installed tensorflow on the base environment using: conda install -c conda-forge tensorflow-gpu. that installed tensorflow-gpu-1.13.1.

Then I imported python and checked if tensorflow has been installed or not uisng tensorflow.version and I got the following message:

<module 'tensorflow._api.v1.version' from '/home/..../anaconda3/lib/python3.7/site-packages/tensorflow/_api/v1/version/__init__.py'>

Then I uninstalled it using conda remove tensorflow-gpu, I uninstalled tensorflow.

Then I tried to install it using pip via the following command:pip install --upgrade pip pip install tensorflow

Then I checked if tensorflow has bben installed using pip or not by the following command:pip freeze

Then I tried reinstall tensorflow using:conda create -n tf-gpu tensorflow-gpu

Bot the following error drooped out:WARNING: A conda environment already exists at '/home/..../anaconda3/envs/tf-gpu' Remove existing environment (y/[n])? y

And at the end I got the error:Proceed ([y]/n)? y Preparing transaction: done Verifying transaction: failed NoSpaceLeftError: No space left on devices.

and when I checked if it has already been installed I got the following message:ModuleNotFoundError: No module named 'tensorflow'So, I tried to reinstall it using:

conda create -n tf-gpu tensorflow-gpu

but I got the following error:

WARNING: A directory already exists at the target location '/home/..../anaconda3/envs/tf-gpu' but it is not a conda environment. Continue creating environment (y/[n])? y`

When I choose y I got the following error:CondaValueError: prefix already exists: /home/..../anaconda3/envs/tf-gpu

When I try to create a new_env using:

conda create --name new_env tensorflow-gpu

Then I got the following error:

CondaError: Unable to create prefix directory '/home/..../anaconda3/envs/new_env'. Check that you have sufficient permissions.

I tried to install only tensorflow-gpu in this new environment, but I got the following error:

`conda install tensorflow-gpu

DirectoryNotACondaEnvironmentError: The target directory exists, but it is not a conda environment. Use 'conda create' to convert the directory to a conda environment. target directory: /home/.../anaconda3/envs/tf-gp`

And when I try the way that it was explained I got the following error:

CondaValueError: prefix already exists: /home/..../anaconda3/envs/tf-gpu

CondaValueError: prefix already exists: /home/..../anaconda3/envs/tf-gpu`

Now I don't know How I should resolve this problem. Indeed my goal was installing tensorflow using conda but not in a new environment because I had pythorch installed and I had some jupyter notebooks that I wanted to import tensorflow and use some specific libraries. But I fall into a loop of installing and uninstalling that led to this wired problem. The problem is that I can't reinstall tensorflow in the environment that I had created and removed tensorflow from it because it says the environment is not a conda environment and when I try to convert the environment to a conda environment I get the error of prefix. And when I try to create a new environment I get the error of permission.

Any help is really appreciated.

1 Answer

Conda creates folders for the environments under ~/miniconda3/envs/ by default. Try navigating to this folder and manually deleting the tensorflow-gpu folder then try again. (If you installed a different version of conda or put it in a different location, navigate to the appropriate location and delete the tensorflow-gpu folder in .../envs/.)

Sometimes I have the issue where conda doesn't quite remove everything and manually deleting the folder fixes it.