- use python3.6 and Make sure python3-dev tools are installed. If using apt:
sudo apt-get install python3-dev - notebooks are best run in jupyterlab
- need gcc4.9+ (tested on gcc7) to install pippi
- need nodejs5+ for some jupyterlab widgets
- since gcc and nodejs are required for parts of this project, a miniconda environment is ideal https://docs.conda.io/en/latest/miniconda.html
- check nodejs --version and gcc --version, then (if needed) update your conda env with:
conda install -c conda-forge nodejs conda install -c creditx gcc-7 - install the following widgets for jupyter-lab: jupyterlab-plotly and plotlywidget v4.9.0
jupyter labextension install @jupyter-widgets/jupyterlab-manager plotlywidget@4.9.0 jupyter labextension install jupyterlab-plotly@4.9.0 - install pippi and other requirements:
sh install.sh - You will need pytorch, torchvision and torchaudio. Specific version are required based on other packages (torchviz, learn2learn, transformers, etc) and your cuda version. You can install these by hunting down the proper version, but conda can simplify the installation process:
conda install pytorch torchaudio -c pytorch pip install torchvision pip install learn2learn - once installed, check random_gen.ipynb for a start
imilas/Synths_Stacks_Search
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|