Quick start ml experimentation pipelines
Project description
faber
what is faber?
faber is a lightweight pipelining tool for machine learning and data projects
why though?
so our pipelines look clean - like the below
from faber import node
from data_science.nodes.quick_ml import *
ml_pipe = [
node(
prepare_ml,
['df_processed', 'ml_params'],
['df_master_train', 'df_master_test'],
),
node(
train_models,
['df_master_train', 'ml_params'],
['ml'],
),
node(
get_best_params,
['ml'],
['best_params']
),
]
and our data is connected externally and defined in a yaml file; meaning if a data location changes only one place is changed . wildcards are also supported with {{}}
df_raw:
filepath: data/00_raw/df_raw.csv
read_func: read_csv_pandas
df_processed:
filepath: data/01_master/df_processed_{{run_number}}.csv
read_func: read_csv_pandas
write_func: write_csv_pandas
output_profile:
filepath: data/02_insight/output_profile.csv
read_func: read_csv_pandas
write_func: write_csv_pandas
ml:
filepath: data/models/ml_object.pickle
read_func: read_pickle
write_func: write_pickle
install using
git clone https://github.com/pinata-brad/faber.git
cd faber
python3 setup.py sdist bdist_wheel && pip install dist/faber-0.0.3.tar.gz
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
faber-ml-0.0.6.tar.gz
(7.2 kB
view hashes)