Installation
Requirements
Python 3.9 or newer is required. The test suite is regularly exercised on current Python 3.10/3.11 environments.
Basic installation
This installs the core package and required dependencies. It is enough when you train from local files or pandas DataFrames and do not need optional plotting, Hugging Face dataset loading, or large-model device mapping.
Recommended
For normal research use, install the recommended extras:
This adds support for common plotting workflows, Hugging Face datasets,
safetensors checkpoints, tokenizer backends, and large-model loading with
device_map / base_model_device_map.
| Package | Purpose |
|---|---|
accelerate |
Hugging Face device_map / base_model_device_map loading for large models |
matplotlib |
Static plots such as convergence curves, encoder distributions, strip plots, and heatmaps |
seaborn |
Higher-level statistical visualizations |
datasets |
Loading Hugging Face datasets by id |
safetensors |
Preferred model serialization format |
sentencepiece |
Tokenizer backend required by many T5/LLaMA-style tokenizers |
Optional: data creation (spaCy)
To create training data from raw text with morphological filtering, install:
This adds:
| Package | Purpose |
|---|---|
spacy |
Morphological filtering via spacy_tags in TextFilterConfig |
datasets |
Loading Hugging Face datasets as base data for TextPredictionDataCreator |
spaCy also needs a language model. For German filtering, for example:
You can combine extras:
Interactive encoder scatter (Plotly)
The interactive encoder scatter plot (trainer.plot_encoder_scatter()) uses
Plotly for hover labels, zooming, and notebook exploration. Plotly is installed
by gradiend[recommended] and gradiend[plot]. Without Plotly, the function
returns None and logs a warning.
To enable the interactive scatter in a minimal install:
Choosing extras
| Task | Suggested install |
|---|---|
| Run the quick start from local data | pip install gradiend[recommended] |
| Generate text-prediction data from raw corpora | pip install gradiend[recommended,data] |
| Use interactive scatter plots in notebooks | pip install gradiend[recommended] |
| Develop GRADIEND itself | pip install -e ".[recommended,data,dev]" |
Dev (contributors)
For building docs and running tests:
From source
git clone https://github.com/aieng-lab/gradiend.git
cd gradiend
pip install -e .
# With recommended extras:
pip install -e ".[recommended]"