Built-in Examples

Jupyter Notebook Examples

You can run the included examples (CNN, PICAE, PIMLTurb on 3D data, PIMLTurb1D,and KRR on 2D data). There is also a notebook with examples of plotting routines and ML network visualization. To copy the examples, type:

1	`sapsan get_examples`

This will create a folder ./sapsan_examples with appropriate example jupyter notebooks and GUI. For starters, to launch a CNN example:

1	`jupyter notebook ./sapsan_examples/cnn_example.ipynb`

GUI Examples

In order to try out Sapsan's GUI, start a streamlit instance to open in a browser. After the examples have been compied into your working directory as described above, you will be able to find the GUI example. The entry point:

1	`streamlit run ./sapsan_examples/GUI/Welcome.py`

The scripts for the pages you see (welcome and examples) are located in the subsequent directory: ./sapsan_examples/GUI/pages/.

If you want to build your own demo, then look into Examples.py to get started. Ideally, you would only need to import your Estimator, EstimatorConfig, EstimatorModel and adjust the run_experiment() function, which has a nearly identical setup to a standard Sapsan's jupyter notebook interface.

Sample Data

The data for the CNN and KRR examples has been sourced from JHTDB. Specifically the Forced MHD Dataset (1024³) has been used as a starting point.

Data	Description
u_dim128_2d.h5	velocity field sampled down to [128,128,128] and using the 1^st slice
u_dim32_fm15.h5	velocity field sampled down to [32,32,32], and spectrally filtered down to 15 modes