Deep Learning Teaching Environments
| Linux | Mac OS X | Windows |
|---|---|---|
 |
NOT SUPPORTED |
Conda environments for Deep Learning lectures.
Usage
First install a Python 3, 64-bits Conda distribution. If you are using Windows make sure you install Anaconda only for your current user, and under a folder that does not contain non-UTF characters (such as ñ or á). After installing follow the instructions for your particular operative system. It is highly recommended that you use a machine with an nVidia GPU.
Linux
CPU environment
Download the environment file to your computer (right click -> Save link as...) and open a terminal in the same folder. Then type
conda env create -f environment-linux.yml
to create an environment named deeplearning-labs with the necessary packages.
After creating the environment, open a terminal in the folder with the notebooks you want to work with. Then log into the environment running
source activate deeplearning-labs
and then type
jupyter notebook
to start the notebook server.
Note: if for any reason Jupyter is unable to find the environment (kernel) you just installed, you can configure it manually by running the following command in the terminal (after activating the environment):
python -m ipykernel install --user --name deeplearning-labs --display-name "Python (deep-learning-labs)"
GPU environment
Make sure you laptop has an nVidia GPU and the appropriate drivers for your operative system and GPU card. After that, download the environment file to your computer (right click -> Save link as...) and open a terminal in the same folder. Then type
conda env create -f environment-linux-gpu.yml
and follow the same steps as for the CPU version. If you find issues when importing tensorflow or keras, you will need the find out what CUDA version is appropriate for your nVidia GPU and driver version, and modify the contents of the environment file to make use of an appropriate cudatoolkit version.
Windows
CPU environment
Download the environment file to your computer (right click -> Save link as...) and open an Anaconda Navigator. Click on the Environments tab, and look for the Import button. After clicking it, select as Specification file the environment file you downloaded. This will create an environment named deeplearning-labs with the necessary packages. Note the creation process might take a while.
After creating the environment, click the play icon at the environment name, and choose the option to launch a Jupyter notebook.
If for whatever reason the Jupyter notebook launch fails, you can also try opening an Anaconda Prompt in the folder with the notebooks you want to work with, then log into the environment running
activate deeplearning-labs
and then type
jupyter notebook
to start the notebook server.
GPU environment
Make sure you laptop has an nVidia GPU and the appropriate drivers for your operative system and GPU card. After that, download the environment file (right click -> Save file as...) and follow the same steps as for the CPU version.
If you find issues when importing tensorflow or keras, you will need the find out what CUDA version is appropriate for your nVidia GPU and driver version, and modify the contents of the environment file to make use of an appropriate cudatoolkit version.
Manual install
If after installing the environment you are unable to run keras or tensorflow, proceed as follows:
-
Remove the previously created conda environment (if any)
conda env remove -n deeplearning-labs
-
Create the environment from scratch
conda env create -n deeplearning-labs
-
Log into the environment
source activate deeplearning-labs (Linux/Mac) activate deeplearning-labs (Windows)
-
Manually install tensorflow. Try first to install the package via conda, and if the install fails for whatever reason, install with pip.
conda install tensorflow pip install tensorflow
-
Manually install all the rest of packages in the corresponding environment file for you distribution. Again, try first to install the package via conda, and if the install fails for whatever reason, install with pip.