Visualize Your Deep Learning Training in Static Graphics.
Why? Analyzing learning curves are a standard way to evaluate the learning performances of machine learning models. There exist several tools for creating live plots. This Python package focuses on producing beautiful static graphics only.
Currently, only plotting from Keras CSV log file format is supported.
For creating the graphics, plotnine is used which is build upon Matplotlib.
pip install deep_plotsUnfortunately, you may need to:
pip install numpybefore because a depedency implicitly assumes numpy is installed.
# create a Keras callback to log your training
csv_logger = keras.callbacks.CSVLogger('log.csv')
# train your model
model.fit(X, y, ..., callbacks=[csv_logger, ...])
# after finishing training, plot the learning curves with Deep Plots
deep_plots.from_keras_log('log.csv', 'output_dir')- TensorBoard: Live plots for TensorFlow, Keras.
- tensorboardX: Live plots for PyTorch, Chainer etc..
- Live Loss Plot: Live plots in Jupyter Notebooks for Keras, PyTorch etc..
If you have a question, found a bug or want to propose a new feature, have a look at the issues page.
Pull requests are especially welcomed when they fix bugs or improve the code quality.
MIT.