No sky seg when using "depth-anything/DA3NESTED-GIANT-LARGE"

Hi, thanks for the work! I'm using DA3NESTED-GIANT-LARGE model but get no sky segmentation output. Here's my script

```python
model = DepthAnything3.from_pretrained("depth-anything/DA3NESTED-GIANT-LARGE").to("cuda")
print(f"Using model {model.model_name} on device {model.device}")

args.batch_size = min(len(images), args.batch_size)
if args.overlap_size is None:
args.overlap_size = max(0, args.batch_size // 3)

idx_groups = []
for start_idx in range(0, len(images), args.batch_size - args.overlap_size):
end_idx = min(start_idx + args.batch_size, len(images))
idx_group = list(range(start_idx, end_idx))
idx_groups.append(idx_group)
if end_idx == len(images):
break

depth_maps = []
processed_intrinsics = []
conf_maps = []


if args.time:
# warm-up
x = torch.randn(4096, 4096, device='cuda')

def bench():
torch.cuda.synchronize()
out = x @ x
torch.cuda.synchronize()

# warm-up runs
for _ in range(10):
bench()
import time
start_time = time.time()
for group_idx in tqdm(range(len(idx_groups)), desc="Processing image groups"):
gt_H, gt_W = images[0].shape[0], images[0].shape[1]

selected_indices = idx_groups[group_idx]
selected_images = [images[i] for i in selected_indices] # N * 3 * H * W
selected_extrinsics = extrinsics[selected_indices]
selected_intrinsics = intrinsics[selected_indices]
predictions = model.inference(
image=selected_images,
process_res=max(gt_H, gt_W),
# process_res=518,
extrinsics=selected_extrinsics,
intrinsics=selected_intrinsics,
)
```
And it turns out that `predictions.sky` is `None`. What's the problem?
Thank you for your respond!