Error with nested models "Caffe2 uses a lazy allocation..."

I am trying to run a FSDP training built on top of metaseq:

I have a model with two other models nested inside it:

- FSDP(model);
- FSDP(submodel)
- FSDP(submodel);

However I get this error: ""RuntimeError: The tensor has a non-zero number of elements, but its data is not allocated yet. Caffe2 uses a lazy allocation, so you will need to call mutable_data() or raw_mutable"" during the forward pass of the first submodule, do you ever encountered this error? Thanks a lot!

The issue names no files or tests. Start by reproducing the nested FSDP setup during the first submodule's forward pass and inspect the reported Caffe2 lazy-allocation error; done means the cause and a verified resolution are documented.

Written by the indexing model from the issue text.