Package that can create feature instances. At present there is one engine name EnginePandas which basically reads features from a file into a pandas dataframe and/or Numpy array. See the notebooks directory for examples.
The output of the engine can subsequently be used as input to Neural Net Models created by the m0d3ls package.
# Define Features
card = ft.FeatureSource('Card', ft.FEATURE_TYPE_STRING)
merchant = ft.FeatureSource('Merchant', ft.FEATURE_TYPE_STRING)
# Create TensorDefinition.
td = ft.TensorDefinition('Features', [card, merchant])
# Build a Pandas DataFrame from the TensorDefinition
with en.EnginePandas(num_threads=1) as e:
df = e.df_from_csv(td, './file.csv', inference=False)
- Pandas
- Numpy
- Numba