constant_data_fingerprint#

pymc_marketing.mmm.evaluator_cache.constant_data_fingerprint(idata)[source]#

Return a content fingerprint of a DataTree’s fit-time constant_data.

Each constant_data variable is compared by name, dims, shape, dtype and a digest of its values, and each coordinate by name, dims and a digest of its labels. Since channel_data is included, wrappers differing only in it, or only in the order of its labels, compile separately, while the summary factory and MMM.incrementality share an entry because they wrap one fit-time dataset.

Parameters:

idata (DataTree) – Fitted inference data whose constant_data group is fingerprinted when present.

Return type:

tuple or None

Returns:

One entry per data variable and per coordinate. None when idata has no constant_data group.