VGGT (fastforks)

VGGT, Visual Geometry Grounded Transformer (Wang et al. 2025): cameras, depth maps and point maps of every view from unposed images in one forward pass. A DINOv2 ViT-L patch embedding feeds 24 alternating frame / global attention blocks; a camera head and two DPT heads decode the tokens.

This repo is the model card only: it holds no weights. fastforks downloads the authors' weights from the links below and checks them against the SHA-256 in the model card. Code: https://github.com/kleinicke/fastforks. Kernels for the kernels backend: https://huggingface.co/fastforks/kernels. Original code: https://github.com/facebookresearch/vggt at a288dd0.

Author weight repos: https://huggingface.co/facebook/VGGT-1B.

Task: multiview. Tier: experimental. Outputs: depth, confidence, world_points, intrinsics, cam_to_world, points_confidence. Backends: portable, kernels.

Input: several images of one scene (posed and calibrated for multi-view stereo, unposed otherwise). Output: depth maps and, where the model predicts them, cameras and 3-D points.

Variants

Weights are the authors' files (not hosted here). SHA-256 shortened to 12 hex digits; the full hashes are in card.toml.

variant	weights (author link)	SHA-256	weights licence	commercial use
vggt-1b@1	model.pt	d15bf50a8615	CC-BY-NC-4.0	no

Licence

Original-code strings, recorded as they are (research-use package; check them before any other use). Code: LicenseRef-VGGT. Files: LICENSE.

Training data: Per the paper (arXiv 2503.11651): Co3Dv2, BlendMVS, DL3DV, MegaDepth, Kubric, WildRGB, ScanNet, HyperSim, Mapillary, Habitat, Replica, MVS-Synth, PointOdyssey, Virtual KITTI, Aria Synthetic Environments, Aria Digital Twin; DINOv2 initialisation.

• Licence note: The VGGT code licence allows commercial use, but the VGGT-1B weights are CC-BY-NC-4.0; Meta's separately licensed commercial checkpoint is not packaged.

Licence evidence: https://github.com/facebookresearch/vggt/blob/a288dd0/LICENSE.txt and https://huggingface.co/facebook/VGGT-1B (license: cc-by-nc-4.0)

Input size

any H x W >= 1, resized or padded inside the model Tested up to (portable, fp16, RTX 5060 Ti 16 GB): vggt-1b 3000x4000 (8.0 GiB peak allocated, 10.1 reserved)

Speed and memory

Multiview, synthetic, 336x518

model	time (ms): original fp32	original	portable	kernels	kernels speed-up: vs original	vs portable	peak memory (MiB, allocated / reserved): original	portable	kernels
vggt-1b	2502	965	651	458	2.1x	1.4x	8688 / 9158	3691 / 4674	3702 / 4654

Startup and first call

model	load (s): original	portable	kernels	first call (s): original	portable	kernels
vggt-1b	19.37 (336x518)	4.98 (336x518)	5.99† (336x518)	1.29 (336x518)	0.69 (336x518)	0.46 (336x518)

torch.compile and TensorRT

model	input	portable ms	compile ms	compile s (cold)	TensorRT fp16 ms	ONNX + build s	kernels ms	peak MiB (alloc / reserved): portable	compile	TensorRT (device)
vggt-1b	336x518	651	556.9	74.3	54.3	150	458	3691 / 4674	3611 / 3856	2988

Fixed-size speed and peak memory, all models

model	mode	480x640	720x1280	1080x1920	12 MP 3000x4000
vggt-1b	portable	793.4 ms / 3868 / 5308	561.3 ms / 3710 / 4930	566.5 ms / 3920 / 5178	849.5 ms / 8219 / 10306
vggt-1b	kernels	559.0 ms / 3880 / 5474	398.6 ms / 3721 / 5156	404.0 ms / 3931 / 5390	615.0 ms / 8232 / 10518

Fixed-size cells: time (ms) / peak allocated / peak reserved memory (MiB).

Hardware, protocol and caveats: benchmarks (RTX 5060 Ti 16 GB, sm_120). Original = the authors' code at the precision of portable (fp16, or (fp32) where fp16 is unsupported or not finite); original fp32 = its default fp32 run. Peak memory is the torch allocator's.

Accuracy

Parity with the original code

Every variant is checked against reference outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.02 and p99 <= 0.25 (absolute error of the output depth).

Usage

from fastforks import multiview

# non-commercial weights: run with FASTFORKS_NONCOMMERCIAL=1 in the environment
model = multiview.load("vggt-1b")  # mode="auto": the preferred backend from recorded measurements
result = model(["view0.jpg", "view1.jpg", "view2.jpg"])  # the views of one scene, no cameras needed
result.depth  # [V, H, W] float32 tensor; result.kind says what the maps mean
result.backend  # what ran, e.g. Backend(mode='portable', device='cuda:0', dtype='float16')

Install with pip install fastforks (see https://github.com/kleinicke/fastforks). Other variants: vggt-1b.

Kernels backend

multiview.load("vggt-1b", mode="kernels") runs our CUDA kernels (fastbb, cudnn_hacc). It needs a prebuilt kernel for your GPU architecture (downloaded from https://huggingface.co/fastforks/kernels on first use) and a recent NVIDIA driver; if there is none it raises BackendUnavailableError with the reason and never falls back silently. mode="auto" picks kernels when it can run and logs the choice.