[QST] What is the proper way to load checkpoint to merlin.models.torch.DLRMModel

[breadbread1984]

❓ Questions & Help

DLRMModel cannot load trained checkpoint properly.

Details

the torch implement of DLRMModel has an interaction layer implement with a member defined in register_buffer in condition block (see

models/merlin/models/torch/blocks/dlrm.py

Line 67 in eb1e541

self.register_buffer(

). the newly created interaction has no such member until the forward is called. therefore, the checkpoint cannot be loaded properly. what is the recommended way of loading?