---
library_name: "fastforks"
pipeline_tag: "image-segmentation"
license: "mit"
tags:
- "segmentation"
- "semantic-segmentation"
- "cuda"
- "pytorch"
- "fastforks"
---
# OneFormer (fastforks)

OneFormer (Jain et al. 2023): Mask2Former-style universal segmentation conditioned on a task token; Swin backbone, run here for the semantic task.

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/shi-labs/oneformer_ade20k_swin_large>.

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

Input: one RGB image. Output: per-pixel class labels (logits with `logits=True`).

## Variants

Weights are the authors' files (not hosted here). SHA-256 shortened to 12 hex digits; the full hashes are in [`card.toml`](https://github.com/kleinicke/fastforks/blob/main/src/fastforks/segmentation/models/oneformer/card.toml).

| variant | weights (author link) | SHA-256 | weights licence | commercial use |
|---|---|---|---|---|
| `oneformer-swinl-ade20k@1` | [oneformer-swinl-ade20k.config.json](https://huggingface.co/shi-labs/oneformer_ade20k_swin_large/resolve/4a5bac8e64f82681a12db2e151a4c2f4ce6092b2/config.json)<br>[oneformer-swinl-ade20k.bin](https://huggingface.co/shi-labs/oneformer_ade20k_swin_large/resolve/4a5bac8e64f82681a12db2e151a4c2f4ce6092b2/pytorch_model.bin) | `27452b656a46`<br>`c0b2fe11dfec` | MIT | yes |

## 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://huggingface.co/shi-labs/oneformer_ade20k_swin_large/blob/4a5bac8e64f82681a12db2e151a4c2f4ce6092b2/README.md>

## Input size

any H x W >= 1, resized or padded inside the model
Tested up to (portable, fp16, RTX 5060 Ti 16 GB): `oneformer-swinl-ade20k` 1988x2880 (8.4 GiB peak allocated, 14.9 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 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| `oneformer-swinl-ade20k` | 307 | 205 | 139 | 100.0 | 2.1x | 1.4x | 7395 / 7792 | 2054 / 2476 | 2054 / 2476 |

### Startup and first call

| model | load (s): original | portable | kernels | first call (s): original | portable | kernels |
| --- | --- | --- | --- | --- | --- | --- |
| `oneformer-swinl-ade20k` | 5.08 (512x512) | 4.10 | 3.27† | 0.81 (512x512) | 0.15† | 0.11† |

### 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) |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| `oneformer-swinl-ade20k` | 1024x2048 | 139 | 138.6 | 0.4 | 83.3 | 140 | 100.0 | 2054 / 2476 | 2046 / 2474 | 3258 |

### Fixed-size speed and peak memory, all models

| model | mode | 480x640 | 720x1280 | 1080x1920 | 12 MP 3000x4000 |
| --- | --- | --- | --- | --- | --- |
| `oneformer-swinl-ade20k` | portable | 86.2 ms / 1295 / 1940 | 120.8 ms / 2220 / 3960 | 138.9 ms / 3772 / 6702 | OOM |
| `oneformer-swinl-ade20k` | kernels | 59.0 ms / 1295 / 1858 | 85.4 ms / 2222 / 3854 | 103.3 ms / 3772 / 6642 | OOM |

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

Hardware, protocol and caveats: [benchmarks](https://github.com/kleinicke/fastforks/blob/main/docs/benchmarks.md) (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| <= 0.001 and p99 <= 0.01 (absolute error of the output `logits`).

## Usage

```python
from fastforks import segmentation

model = segmentation.load("oneformer-swinl-ade20k")  # mode="auto": the preferred backend from recorded measurements
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: `oneformer-swinl-ade20k`.

## Kernels backend

`segmentation.load("oneformer-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.

<!-- staged from fastforks badda9aa (2026-11-06) -->