Skip to main content

Kedro

  • Kedro is the foundation for clean data science code. It borrows concepts from software engineering and applies them to machine-learning projects.
  • https://kedro.org/

How​

Pip​

  • install via pip pip install kedro
  • test if it's working kedro info

Anaconda​

brew install --cask anaconda
export PATH=/usr/local/anaconda3/bin:"$PATH"
conda list

conda create --name kedro-environment python=3.7 -y
conda activate kedro-environment
conda install -c conda-forge kedro
kedro info

Code​

"""Contents of hello_kedro.py"""
from kedro.io import DataCatalog, MemoryDataSet
from kedro.pipeline import node, Pipeline
from kedro.runner import SequentialRunner

# Prepare a data catalog
data_catalog = DataCatalog({"my_salutation": MemoryDataSet()})


# Prepare first node
def return_greeting():
    return "Hello"


return_greeting_node = node(return_greeting, inputs=None, outputs="my_salutation")


# Prepare second node
def join_statements(greeting):
    return f"{greeting} Kedro!"


join_statements_node = node(
    join_statements, inputs="my_salutation", outputs="my_message"
)

# Assemble nodes into a pipeline
pipeline = Pipeline([return_greeting_node, join_statements_node])

# Create a runner to run the pipeline
runner = SequentialRunner()

# Run the pipeline
print(runner.run(pipeline, data_catalog))