search

det-lab / jupyterhub-deploy-kubernetes-jetstream

CDMS JupyterHub deployment on XSEDE Jetstream
0 stars 1 forks source link

Making an environment for a tf_slim project #80

Open pibion opened 1 year ago

pibion commented 1 year ago

We're interested in working on the code in the repository https://github.com/jsyoon0823/TimeGAN.git but are having trouble building the python environment on Jetstream. The steps I've taken are

git clone git@github.com:jsyoon0823/TimeGAN.git
python -m venv gan
source gan/bin/activate
pip install -r requirements.txt

And it seems like it's failing on wheel installation.

The ultimate goal is to build the environment and then create a jupyter kernel, we're trying to find a simpler example to see if we can do that.

pibion commented 1 year ago

The odd thing is that I can build the environment successfully when working on my local WSL environment. The versions are a little different,

Jetstream python is 3.8.3 and my local version is 3.8.10.

But the big difference is at python 3.7 and both are beyond that.

zonca commented 1 year ago

Can you paste the error?

zonca commented 1 year ago

Before running any command, please check that the path is correct.

which python which pip

Also be careful about python vs python3, pip vs pip3.

zonca commented 1 year ago

it is a complicated environment, I would try with conda.

I am testing this, it is taking forever, it says there are conflicts:

conda create --prefix condatgan -y -c conda-forge --file=TimeGAN/requirements.txt python==3.9
zonca commented 1 year ago

it fails with a CUDA driver error:

Your installed CUDA driver is: not available

not sure why we cannot have a CPU only version.