In order to join the workshop, please go through the following software installation steps at least a day in advance.
If you run into any errors please send an email to Sean Perez (seanperez@berkeley.edu) for help and include:
- Which step you’re running,
- The error message(s) (screenshot preferred), and
- Your operating system + Python version + Jupyter Notebook version.
It is important to get everything working before the training because we will not likely have sufficient time to investigate fixes during the training itself.
There are two options to get started on this workshop:
-
If you have a berkeley.edu email account, you likely have access to dlab's datahub!
-
Click this link
-
That's it. You are ready!
-
When you want to refer to the workshop in the future, just find the folder Python-Deep-Learning within your datahub environment at https://dlab.datahub.berkeley.edu/
- This option is for attendees without a berkeley.edu email account.
- Simply open each notebook using the following links:
- 01-Vanilla-Neural-Networks.ipynb
- 02-Vanilla-Convolutional-Neural-Network-Comparison.ipynb
- 03-Convolutional-Neural-Networks.ipynb
- 01-Vanilla-Neural-Networks.ipynb
- This will take approximately 30-60 minutes.
- Download the Python-Deep-Learning repository here: https://github.com/dlab-berkeley/Python-Deep-Learning
- Download Anaconda here: https://www.anaconda.com/products/individual
- Open the Anaconda Navigaor and launch the Powershell Prompt.
- Execute the following lines of code one at a time (note that this may take a while...):
conda create --name pydeeplearning python=3.9 keras=2.6 matplotlib=3.4 numpy=1.21 jupyterlab ipykernel
Creates a conda environment called pydeeplearning with only the essential packages for our workshop. You will need to reply y to allow installation.
conda activate pydeeplearning
Activates the environment.
python -m ipykernel install --user --name pydeeplearning
Creates a kernel so that we can use this python environment in our Jupyter notebook.
conda install -c conda-forge tensorflow
Installs tensorflow, the backend for Keras. This may take a while!
- Launch Jupyter Notebook, navigate to the repository you downloaded, and open up your notebook.
- In the toolbar below the Jupyter logo go to Kernel > change Kernel > pydeeplearning
- You are set to go!
- Download the Python-Deep-Learning repository here: https://github.com/dlab-berkeley/Python-Deep-Learning
- Download Anaconda here: https://www.anaconda.com/products/individual
- Open your terminal and input the following commands:
conda
Check to see if the conda command is recogenized and shows argument options.
If you get the error message: command not found: conda
Try running the following:
/opt/anaconda3/bin/conda init zsh
Check to see if the conda command now shows argument options.
conda create -n pydeeplearning python=3.9
Creates a conda environment called pydeeplearning
conda activate pydeeplearning
Activates the environment.
python -m pip install tensorflow
Installs tensorflow.
python -m pip install matplotlib
Installs matplotlib for plotting.
python -m pip install jupyterlab
Installs jupyter notebook station.
python -m ipykernel install --user --name pydeeplearning
Creates a kernel so that we can use this python environment in our Jupyter notebook.
- Launch Jupyter Notebook, navigate to the repository you downloaded, and open up your notebook.
- In the toolbar below the Jupyter logo go to Kernel > change Kernel > pydeeplearning
- You are set to go!
- I suggest following this guide: https://github.com/jeffheaton/t81_558_deep_learning/blob/master/install/tensorflow-install-mac-metal-jul-2021.ipynb
- This process is quite difficult, but if successful allows you to use GPUs!
Why do we need to need to do all this conda stuff?
Conda is a package manager, which we can use to deal with versioning conflicts and dependencies for tensorflow/keras! In order for keras/tensorflow to work smoothly, we create a new isolated conda environment and only install our essential packages.