function hub (wip)
This functions hub is intended to be a centralized location for open source contributions of function components. These are functions expected to be run as independent mlrun pipeline compnents, and as public contributions, it is expected that contributors follow certain guidelines/protocols (please chip-in).
data
arc_to_parquet
download remote archive files and save to parquet
gen_class_data
generate simulated classification data according to detailed specs. Great for testing algorithms and metrics and whole pipelines.
load_datasets
download toy datasets from sklearn, tensorflow datasets, and other data external curated datasets.
open_archive
download a zip or tar archive and extract its contents into a folder (preserving the directory structure)
load_dask
define a dask cluster, load your parquet data into it
explore
describe
estimate a set of descriptive statistics on pipeline data
describe_dask
estimate a set of descriptive statistics on pipeline data that has been loaded into a dask cluster
model
aggregate
rolling aggregations on time seriesA
feature_selection
feture selection using the scikit feature-selection module
sklearn classifier
train any sklearn class has that has a fit function, including estimators, tranformers, etc...
xgb_trainer
train any one of 5 xgboost model types (classifier, regressor,...)
serve
tf1_serving
deploy a tensorflow 1.x server
tf2_serving
deploy a tensorflow 2.x server
xgb_serving
deploy any xgboost model
model_server
deploy an scikit-learn or almost any pickled model
test
model_server_tester
deploy an scikit-learn or almost any pickled model
test_classifier
test a classifier's model against help-out or new data