Closed evamaxfield closed 6 years ago
General model is complete so closing the issue.
Custom structures must have the following:
__init__(self, obj):
# initialize the basic parameters of your introspector
# at a minimum requires `_obj` and `_validated`.
self._obj = obj
self._validated = False
@property
def obj(self):
# so users cannot edit the object
@property
def validated(self):
# so users cannot edit the validated props
def validate(self):
# single call to validate that can call other validation methods
def deconstruct(self):
# take object and create Iota, Group, and IotaGroup lists
def reconstruct(self):
# take lists of Iota, Group, and IotaGroup and create object
def package(self):
# take object and form into Quilt package
Generalizable Introspectors
Allow custom validators and parsers for any python object to Dataset object. This would allow for any data type to be consumed and processed by DSDB. Meaning, if a person wants to track a single object as it is process instead of a dataframe like dataset you can.
Example:
Dataset(pytorch_model)
, allows for tracking of epochs/ cycles. Introspector could then read the state of the model to create Iota based off it's attributes.Default: assumed DataFrame like.
introspectors.init
core