Mask2Former (fastforks)

Mask2Former (Cheng et al. 2022): Swin backbone, multi-scale deformable attention pixel decoder and a masked-attention transformer decoder; semantic logits from mask classification.

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/huggingface/transformers at c4c4605f9618.

Author weight repos: https://huggingface.co/facebook/mask2former-swin-large-ade-semantic.

Task: segmentation. Tier: stable. Outputs: logits. Backends: portable, kernels.

Input: one RGB image. Output: per-pixel class logits.

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
mask2former-swinl-ade20k@1	mask2former-swinl-ade20k.config.json mask2former-swinl-ade20k.safetensors	573b9330846e b143c144341c	CC-BY-NC-4.0	no

Licence

Original-code strings, recorded as they are (research-use package; check them before any other use). Code: Apache-2.0. Files: .

Training data: ImageNet-21k / ImageNet-22k (full ImageNet, fall 2011 release), ADE20K (SceneParsing, 150 classes).

Licence evidence: https://github.com/facebookresearch/Mask2Former/blob/9b0651c6c1d5b3af2e6da0589b719c514ec0d69a/MODEL_ZOO.md#license

Input size

any H x W >= 1, resized or padded inside the model Tested up to (portable, fp16, RTX 5060 Ti 16 GB): mask2former-swinl-ade20k 3000x4000 (14.1 GiB peak allocated, 15.1 reserved)

Speed and memory

Segmentation, synthetic, 1024x2048

model	time (ms): original fp32	original	portable	kernels	kernels speed-up: vs original	vs portable	peak memory (MiB, allocated / reserved): original	portable	kernels
mask2former-swinl-ade20k	244.17	144.79	103.88	60.44	2.4x	1.7x	3572 / 3878	1374 / 1660	1164 / 1400

Startup and first call

model	load (s): original	portable	kernels	first call (s): original	portable	kernels
mask2former-swinl-ade20k	4.86 (512x512)	3.21	3.17†	0.76 (512x512)	0.12†	0.07†

Fixed-size speed and peak memory, all models

model	mode	640x480	1280x720	1920x1080	12 MP 4000x3000
mask2former-swinl-ade20k	portable	71.9 ms / 1194 / 1912	96.1 ms / 1698 / 3182	104.1 ms / 2985 / 4500	152.8 ms / 14457 / 15480
mask2former-swinl-ade20k	kernels	45.3 ms / 1194 / 1816	60.5 ms / 1699 / 3056	68.5 ms / 2986 / 4374	126.0 ms / 14457 / 15344

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 golden 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| <= 0.001 px and p99 <= 0.01 px.

Usage

from fastforks import segmentation

model = segmentation.load("mask2former-swinl-ade20k")  # portable PyTorch on the GPU if there is one, else the CPU
result = model("photo.jpg")
result.labels  # [H, W] per-pixel class index
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: mask2former-swinl-ade20k.

Kernels backend

segmentation.load("mask2former-swinl-ade20k", mode="kernels") runs our CUDA kernels (fvm_seg, seg_up). 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.