Depth map horizontal wrap problem

[PhilFM]

Hi VGGT Mac developers, fantastic work! I've been trying out vggt-mps on my machine using single images. I can see that the depth maps all have a wrap effect where part of the depths on the right hand side appear on the left, with most of the depth pixels shifted to the right. I've attached an example image I used and the result.

The command line I used was:

PYTORCH_ENABLE_MPS_FALLBACK=1 vggt reconstruct ~/Downloads/20251001_121239.jpg

Not sure whether this is a bug in the main VGGT code, something specific to Mac or the fact that I'm using a MacBook Pro 4 (only up 3 Pro 3 are listed as supported). Here's my system:

MacBook Pro 4
MacOS Version 15.6.1 (24G90)
Python 3.10.18

Python environment:

certifi            2025.8.3
charset-normalizer 3.4.3
contourpy          1.3.2
cycler             0.12.1
einops             0.8.1
filelock           3.19.1
fonttools          4.60.0
fsspec             2025.9.0
hf-xet             1.1.10
huggingface-hub    0.35.1
idna               3.10
Jinja2             3.1.6
kiwisolver         1.4.9
MarkupSafe         3.0.2
matplotlib         3.10.6
mpmath             1.3.0
networkx           3.4.2
numpy              1.26.1
opencv-python      4.11.0.86
packaging          25.0
pillow             11.3.0
pip                25.2
pyparsing          3.2.5
python-dateutil    2.9.0.post0
PyYAML             6.0.2
regex              2025.9.18
requests           2.32.5
safetensors        0.6.2
scipy              1.15.3
setuptools         80.9.0
six                1.17.0
sympy              1.14.0
timm               1.0.20
tokenizers         0.22.1
torch              2.3.1
torchvision        0.18.1
tqdm               4.67.1
transformers       4.56.2
typing_extensions  4.15.0
urllib3            2.5.0
uv                 0.8.22
vggt-mps           2.0.0

Execution log:

REPO_PATH= /Users/philip.fraser/Code/External/vggt-mps/src/repo/vggt
Here REPO_PATH exists
============================================================
🔮 VGGT 3D Reconstruction
============================================================
📸 Found 1 images

📂 Loading images...
  • 20251001_121239.jpg
MODEL_CONFIG: {'name': 'VGGT-1B', 'huggingface_id': 'facebook/VGGT-1B', 'local_path': PosixPath('/Users/philip.fraser/Code/External/vggt-mps/src/models/vggt_model.pt'), 'model_size': '5GB', 'parameters': '1B'}
get_model_path(): /Users/philip.fraser/Code/External/vggt-mps/src/models/vggt_model.pt
MODEL_CONFIG: {'name': 'VGGT-1B', 'huggingface_id': 'facebook/VGGT-1B', 'local_path': PosixPath('/Users/philip.fraser/Code/External/vggt-mps/src/models/vggt_model.pt'), 'model_size': '5GB', 'parameters': '1B'}

🚀 Initializing VGGT on mps

🔄 Processing images...
📂 Loading model from: /Users/philip.fraser/Code/External/vggt-mps/src/models/vggt_model.pt
✅ Model loaded successfully!
/Users/philip.fraser/Code/External/venvVGGT/lib/python3.10/site-packages/torch/nn/functional.py:4072: UserWarning: The operator 'aten::_upsample_bicubic2d_aa.out' is not currently supported on the MPS backend and will fall back to run on the CPU. This may have performance implications. (Triggered internally at /Users/runner/work/pytorch/pytorch/pytorch/aten/src/ATen/mps/MPSFallback.mm:13.)
  return torch._C._nn._upsample_bicubic2d_aa(input, output_size, align_corners, scale_factors)
Calculating depth depth_tensor shape= (1, 1, 392, 518, 1)
image hw= (518, 691)
intrisics= [[[[493.38388   0.      345.5    ]
   [  0.      490.1879  259.     ]
   [  0.        0.        1.     ]]]]

📊 Creating visualizations...

============================================================
✅ Reconstruction complete!
📁 Results saved to: outputs
  • input_views.png
  • depth_maps.png
  • 3d_reconstruction.png
  • point_cloud.ply
============================================================