pylace 0.9.0

A probabalistic programming ML tool for science

pylace

Python bindings to lace

Install

Install latest from PyPI

$ python3 -m pip install pylace

Install latest from GitHub

Building from source requires the Rust compiler (git it here).

$ python3 -m pip install git+https://[e-mail]/promised-ai/lace.git#egg=pylace&subdirectory=pylace

Use

Note that the engine currently only supports loading from an existing metadata file. The following lace functions are supported:

• rowsim
• depprob
• predict
• logp
• simulate
• append_rows
• update

import polars as pl
import lace

# The required files can be found here: https://github.com/promised-ai/lace/tree/master/pylace/lace/resources/datasets/satellites
df = pl.read_csv("./data.csv")
engine = lace.Engine.from_df(df, codebook="codebook.yaml")

# Train the model for 10_000 steps
engine.update(10_000)

# Predict the orbit's class based on the orbit's period.
engine.predict('Class_of_Orbit', given={'Period_minutes': 1436.0})
# ('GEO', 0.13583714831550336)

Tests

To run tests, use pytest

$ pytest -x

To run doctets:

$ python tests/test_docs.py

To prevent plotly from displaying

$ LACE_DOCTEST_NOPLOT=1 python tests/test_docs.py