Prerequisites¶
gobbli requires Python 3.7+.
First, ensure Docker is installed and your user has permissions to run docker commands. Next, install the gobbli
package and dependencies into your environment:
pip install gobbli
Some of the Data Augmentation methods require extra packages. You can install them all using the following steps:
pip install gobbli[augment]
python -m spacy download en_core_web_sm
Additionally, Document Windowing with the SentencePiece tokenizer requires extra packages. Install them like so:
pip install gobbli[tokenize]
The Streamlit-based Interactive Apps require their own set of dependencies:
pip install gobbli[interactive]
If you want to train models using a GPU, you will additionally need an NVIDIA graphics card and nvidia-docker.