[BUG] `PatchEmbed` returns a view tensor, triggering an FSDP2 warning

/usr/local/lib/python3.12/dist-packages/torch/distributed/fsdp/_fully_shard/_fsdp_state.py:336: UserWarning: FSDP2-wrapped module (FSDPPatchEmbed) returned a view tensor. An in-place op on this view (e.g., `x += y`) will silently drop the pre-backward hook and skip the all-gather, which can cause backward to fail or produce wrong gradients. Use out-of-place ops (`out = out + y`, not `out += y`) or `.clone()` the output before any in-place op.
  output = self._register_pre_backward_hook(output)

import torch
import torch.distributed as dist
from torch.distributed.fsdp import fully_shard
from timm.layers import PatchEmbed

deviceid = torch.cuda.current_device()
device = torch.device(f"cuda:{deviceid}")
backend = torch.distributed.get_default_backend_for_device(device)
dist.init_process_group(
    backend=backend,
    init_method="env://",
    device_id=device,
)

pe = PatchEmbed()
pe.cuda()
fully_shard(pe)

x = torch.randn(2, 3, 224, 224, device="cuda")

out = pe(x)

$ torchrun --standalone --nproc-per-node gpu timmpe_fsdp2warn.py
/usr/local/lib/python3.12/dist-packages/torch/distributed/fsdp/_fully_shard/_fsdp_state.py:336: UserWarning: FSDP2-wrapped module (FSDPPatchEmbed) returned a view tensor. An in-place op on this view (e.g., `x += y`) will silently drop the pre-backward hook and skip the all-gather, which can cause backward to fail or produce wrong gradients. Use out-of-place ops (`out = out + y`, not `out += y`) or `.clone()` the output before any in-place op.
  output = self._register_pre_backward_hook(output)