Common Commands

CUDA

• Check version: CUDA check version

Conda

• List envs
  • conda info --envs
• Create env
  • conda create -n tf_115 python=3.7 tensorflow-gpu=1.15
  • conda create -n tf_220 python=3.7 tensorflow=2.2.0=gpu_py37h1a511ff_0 scikit-learn
• Activate env
  • conda activate tf_220
• Install env on Jupyter notebook
  1. Create env and activate it
     • conda create -n tf_220 python=3.7 anaconda ipykernel tensorflow=2.2.0=gpu_py37h1a511ff_0
     • conda activate tf_220
  2. or install if created and activated
     • conda install -c anaconda ipykernel
  3. python -m ipykernel install --user --name=tf_220
• Remove env
  • conda remove -n $name --all
• List packages in env
  • conda list -n tf_200

Jupyter lab

• List jupyter notebook
  • jupyter notebook list
• Run background
  • nohup jupyter lab &
  • jupyter lab --allow-root --ip=0.0.0.0 --port=8888
• Stop
  • jupyter notebook stop 3443

Jupyter kernel

• List kernels
  • jupyter kernelspec list
• Install kernel
  • ipython kernel install --user --name=<name of kernel>
• Uninstall kernel
  • jupyter kernelspec uninstall <name of kernel>

TensorBoard

• tensorboard --logdir logs/log_dir --bind_all
• tensorboard —-logdir logs/log_dir —host 0.0.0.0 —port 6006

TensorFlow

• Release GPU
  • https://github.com/tensorflow/tensorflow/issues/36465
  • for i in $(sudo lsof /dev/nvidia2 | grep python | awk '{print $2}' | sort -u); do sudo kill -9 $i; done