kdu4108 / brunoflow Goto Github PK

We recommend installing brunoflow into a fresh conda environment (Python3.9+). To create and then activate a conda environment, run conda create --name bflow python=3.9 and then conda activate bflow.

Then, to install brunoflow, clone the repo and run pip install -e . (make sure this is within your new conda env). Since this is an editable install, you should be able to edit files directly and then those changes should automatically update when you use the library.

import brunoflow as bf
x = bf.Node(val=10, name="x")
output = x ** 2
output.backprop()
print(x.grad) # expected: 20.0
print(x.compute_entropy().val) # expected: 0.0

Run pre-commit install to get nice pre-commit hooks for linting/formatting.

See the README inside of the brunoflow directory.