6
Installation recommended within an Ubuntu/Mint VM with 6-8 GB of memory:
- Anaconda https://www.anaconda.com/download/
- Download Github Tensorflow tutorials: https://github.com/Hvass-Labs/TensorFlow-Tutorials
- From the folder where you extracted these files, load a bash shell and install the python packages:
- conda create --name tf python=3
- pip install -r requirements.txt
(you may need sudo -H)
To activate/deactivate the environment:
- source activate tf
- jupyter notebook
- click on "http://localhost:8888/?token=..." link. This may also run automatically
- Browse to the folder where you extracted the github files, click on "16_Reinforcement_Learning.ipynb"
- Shutdown: ctrl-c followed by "source deactivate"
First run through the Jupyter notebook step-by-step with the run command with the default values and ensure that the environment is working correctly. Training with Tensorflow/CPU will require ~150 hours of computations. If you have a nVidia Geforce series 10 (1060/1070/1080) you can reduce this to hours. (edit requirements.txt to install the GPU version of Tensorflow)