Open
Description
📚 Documentation
In the blog introducing FSDP API
fsdp_model = FullyShardedDataParallel(
model(),
fsdp_auto_wrap_policy=default_auto_wrap_policy,
cpu_offload=CPUOffload(offload_params=True),
)
it should be model
instead of model()
inside FullyShardedDataParallel
so it should be
fsdp_model = FullyShardedDataParallel(
model,
fsdp_auto_wrap_policy=default_auto_wrap_policy,
cpu_offload=CPUOffload(offload_params=True),
)
Metadata
Metadata
Assignees
Labels
No labels