Is there a way to use a FlyteFile with an already uploaded path, in place of e.g. a

nn.Module

or

Tensor

when executing a workflow via

FlyteRemote

? The

FlyteFile[PYTORCH_FORMAT]

trick isn't working, though that may be something on my end. Should it? This works with

StructuredDataset

by setting the

uri

parameter as a remote path. So say I have a model uploaded at

<gs://my-bucket/my_location.pt>

and a workflows with

nn.Module

as an input parameter I'd like to run it with

flyte.execute(flyte.get_workflow(...), inputs={'model': FlyteFile(path='<gs://my-bucket/my_location.pt')})>

Or some such. Is this possible?

😄