SAVRN
Search Contact SAVRN

Open-weight model · Object detection

yolos-small

by HUST Vision Lab hustvl/yolos-small

YOLOS model fine-tuned on COCO 2017 object detection (118k annotated images). It was introduced in the paper You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection by Fang et al. and first released in this repository.

Parameters31M
Context—
Weights245.6 MB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads763.6k

Runs On

What it takes to serve yolos-small (31M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 0.1 GB 0.1 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 0.0 GB 0.0 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 0.0 GB 0.0 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.

SAVRN's Notes on yolos-small

At 31M parameters, yolos-small is small enough that the GPU question almost disappears. HUST Vision Lab trained it on COCO 2017, 118k annotated images, and the checkpoint is 246 MB on disk. At 16-bit the weights round to 0.1 GB and so does the memory it needs. The cheapest setup in our list is a single MI300X with 192 GB at $1.85 per hour on-demand, a card this detector alone could never fill. Put it on shared hardware with other small vision jobs.

Apache 2.0 lets you use it commercially, modify it and redistribute it, provided notices stay and significant changes are stated. Two points a buyer should know. The lab did not write the model card; the hosting platform's team did, so go to arXiv:2106.00666 for the method. And the training set is COCO, so those 118k images are all it has seen; run it over your own footage before you commit. The last update was 2024-05-08.

SAVRN Research, 2026-09-18

Model Card

By HUST Vision Lab, published under apache-2.0, revision 3d8f7130d3ce.

YOLOS (small-sized) model

YOLOS model fine-tuned on COCO 2017 object detection (118k annotated images). It was introduced in the paper You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection by Fang et al. and first released in this repository.

Disclaimer: The team releasing YOLOS did not write a model card for this model so this model card has been written by the Hugging Face team.

Model description

YOLOS is a Vision Transformer (ViT) trained using the DETR loss. Despite its simplicity, a base-sized YOLOS model is able to achieve 42 AP on COCO validation 2017 (similar to DETR and more complex frameworks such as Faster R-CNN).

Read the full model card (474 words)

Configuration

Architecture
YolosForObjectDetection
Layers
12
Hidden size
384
Feed-forward size
1,536
Attention heads
6
Stored precision
float32
Model type
yolos

Identity and Version

Repository
hustvl/yolos-small
Publisher
HUST Vision Lab
Task
Object detection
Modality
Image
Library
transformers
Parameters
31M parameters
Languages
Not stated by the source
Revision
3d8f7130d3ce4907cb206fe1c8485dc8fe8703de
First published
2022-04-26
Last updated
2024-05-08

Files and Weights

6 files, 245.6 MB in total. The weights are 2 files totalling 245.6 MB in bin, safetensors.

Weights2 files · 245.6 MB
Configuration2 files · 4.4 KB
Documentation1 file · 4.2 KB
Repository1 file · 1.2 KB
Every file
FileTypeSizeSHA-256
model.safetensorsWeights122.8 MB 27823dd21001
pytorch_model.binWeights122.8 MB a5fbcdd7e561
config.jsonConfiguration4.1 KB —
preprocessor_config.jsonConfiguration292 B —
README.mdDocumentation4.2 KB —
.gitattributesRepository1.2 KB —

License and Download

License
apache-2.0
Access
Open weights, no gate
Download size
245.6 MB
Download from HUST Vision Lab

Released by HUST Vision Lab through its official repository on Hugging Face. Read the license.

Built From

Memory Requirements

PrecisionWeights in memory
As published245.6 MB
16-bit0.1 GB
8-bit0.0 GB
4-bit0.0 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About yolos-small

How much GPU memory does yolos-small need?

About 0.1 GB at 16-bit and 0 GB at 4-bit: the weights (31M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run yolos-small on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use yolos-small commercially?

Yes. yolos-small is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

Similar Models

Model · Object detection

rf-detr-base

Roboflow

RF-DETR is a real-time detection transformer family introduced in RF-DETR: Neural Architecture Search for Real-Time Detection Transformers by Robinson et al. and integrated in Transformers via PR #36895. RF-DETR is an end-to-end object detection model that combines ideas from LW-DETR and Deformable DETR: a DINOv2-with-registers style ViT backbone (with an RF-DETR windowing pattern for efficient attention), a multi-scale projector between encoder and decoder, and a multi-scale deformable DETR decoder for fast convergence and strong accuracy–latency tradeoffs. You can use the raw model for object detection. See the model hub to look for all available RF-DETR models. Here is how to use this…

Open weights apache-2.0 32M parameters transformers
View model

Table Transformer (DETR) model trained on PubTables1M. It was introduced in the paper PubTables-1M: Towards Comprehensive Table Extraction From Unstructured Documents by Smock et al. and first released in this repository. Disclaimer: The team releasing Table Transformer did not write a model card for this model so this model card has been written by the Hugging Face team. The Table Transformer is equivalent to DETR, a Transformer-based object detection model. Note that the authors decided to use the "normalize before" setting of DETR, which means that layernorm is applied before self- and cross-attention. You can use the raw model for detecting the structure (like rows, columns) in tables.…

Open weights mit 29M parameters 1,024 tokens transformers
View model

Table Transformer (TATR) model trained on PubTables1M and FinTabNet.c. It was introduced in the paper Aligning benchmark datasets for table structure recognition by Smock et al. and first released in this repository. Disclaimer: The team releasing Table Transformer did not write a model card for this model so this model card has been written by the Hugging Face team. The Table Transformer is equivalent to DETR, a Transformer-based object detection model. Note that the authors decided to use the "normalize before" setting of DETR, which means that layernorm is applied before self- and cross-attention. You can use the raw model for detecting tables in documents. See the documentation for more…

Open weights mit 29M parameters transformers
View model

Model · Object detection

table-transformer-detection

Microsoft

Table Transformer (DETR) model trained on PubTables1M. It was introduced in the paper PubTables-1M: Towards Comprehensive Table Extraction From Unstructured Documents by Smock et al. and first released in this repository. Disclaimer: The team releasing Table Transformer did not write a model card for this model so this model card has been written by the Hugging Face team. The Table Transformer is equivalent to DETR, a Transformer-based object detection model. Note that the authors decided to use the "normalize before" setting of DETR, which means that layernorm is applied before self- and cross-attention. You can use the raw model for detecting tables in documents. See the documentation for…

Open weights mit 29M parameters 1,024 tokens transformers
View model

Model · Object detection

PP-DocLayoutV3_safetensors

PaddlePaddle

Unified Layout Module for PaddleOCR-VL 1.5/1.6 & GLM-OCR This is the model weights for PP-DocLayoutv3 in safetensors format. Get PaddlePaddle weights at PP-DocLayoutV3 PP-DocLayoutV3 is specifically engineered to handle non-planar document images. It can directly predict multi-point bounding boxes for layout elements—as opposed to standard two-point boxes—and determine logical reading orders for skewed and curved surfaces within a single forward pass, significantly reducing cascading errors. This model is an essential component of PaddleOCR-VL-1.5, providing crucial layout analysis for the high-precision parsing of various real-world documents in PaddleOCR-VL. This work has been accepted to…

Open weights apache-2.0 33M parameters transformers
View model

Model · Object detection

rf-detr-large

Roboflow

RF-DETR is a real-time detection transformer family introduced in RF-DETR: Neural Architecture Search for Real-Time Detection Transformers by Robinson et al. and integrated in Transformers via PR #36895. RF-DETR is an end-to-end object detection model that combines ideas from LW-DETR and Deformable DETR: a DINOv2-with-registers style ViT backbone (with an RF-DETR windowing pattern for efficient attention), a multi-scale projector between encoder and decoder, and a multi-scale deformable DETR decoder for fast convergence and strong accuracy–latency tradeoffs. You can use the raw model for object detection. See the model hub to look for all available RF-DETR models. Here is how to use this…

Open weights apache-2.0 34M parameters transformers
View model