Currently we drop all of the init functions from the Lux models. However, for SciML Applications it is quite critical to support those. Proposed solution: Store the lux model along side the equivalent simple chains version. Instead of calling params on the simplechains model call initialparameters on the lux model and then flatten it.
Currently we drop all of the init functions from the Lux models. However, for SciML Applications it is quite critical to support those. Proposed solution: Store the lux model along side the equivalent simple chains version. Instead of calling params on the simplechains model call
initialparameters
on the lux model and then flatten it.