Official PyTorch implementation of MambaVision: A Hybrid Mamba-Transformer Vision Backbone.
Ali Hatamizadeh and
Jan Kautz.
For business inquiries, please visit our website and submit the form: NVIDIA Research Licensing
Try MambaVision: 
MambaVision demonstrates a strong performance by achieving a new SOTA Pareto-front in
terms of Top-1 accuracy and throughput.
We introduce a novel mixer block by creating a symmetric path without SSM to enhance the modeling of global context:
MambaVision has a hierarchical architecture that employs both self-attention and mixer blocks:
💥 News 💥
-
[06.10.2025] The MambaVision poster will be presented in CVPR 2025 in Nashville on Sunday, June 15, 2025, from 10:30 a.m. to 12:30 p.m. CDT in Exhibit Hall D, Poster #403.
-
[06.10.2025] Semantic segmentation code and models released here !
-
[06.07.2025] Object detection code and models released here !
-
[03.29.2025] You can now easily run MambaVision in Google Colab. Try here:
-
[03.29.2025] New MambaVision pip package released !
-
[03.25.2025] Updated manuscript is now available on arXiv !
-
[03.25.2025] 21K models and code added to the repository.
-
[03.25.2025] MambaVision is the first mamba-based vision backbone at scale !
-
[03.24.2025] MambaVision-L3-512-21K achieves a Top-1 accuracy of 88.1 %
-
[03.24.2025] New ImageNet-21K models have been added to MambaVision Hugging Face collection
-
[02.26.2025] MambaVision has been accepted to CVPR 2025 !
-
[07.24.2024] MambaVision Hugging Face models are released !
-
[07.14.2024] We added support for processing any resolution images.
-
[07.12.2024] Paper is now available on arXiv !
-
[07.11.2024] Mambavision pip package is released !
-
[07.10.2024] We have released the code and model checkpoints for Mambavision !
Quick Start
Google Colab
You can simply try image classification with MambaVision in Google Colab:
Pretrained MambaVision models can be simply used via Hugging Face library with a few lines of code. First install the requirements:
pip install mambavision
The model can be simply imported:
>>> from transformers import AutoModelForImageClassification
>>> model = AutoModelForImageClassification.from_pretrained("nvidia/MambaVision-T-1K", trust_remote_code=True)
We demonstrate an end-to-end image classification example in the following.
Given the following image from COCO dataset val set as an input:
The following snippet can be used:
from transformers import AutoModelForImageClassification
from PIL import Image
from timm.data.transforms_factory import create_transform
import requests
model = AutoModelForImageClassification.from_pretrained("nvidia/MambaVision-T-1K", trust_remote_code=True)
# eval mode for inference
model.cuda().eval()
# prepare image for the model
url = 'http://images.cocodataset.org/val2017/000000020247.jpg'
image = Image.open(requests.get(url, stream=True).raw)
input_resolution = (3, 224, 224) # MambaVision supports any input resolutions
transform = create_transform(input_size=input_resolution,
is_training=False,
mean=model.config.mean,
std=model.config.std,
crop_mode=model.config.crop_mode,
crop_pct=model.config.crop_pct)
inputs = transform(image).unsqueeze(0).cuda()
# model inference
outputs = model(inputs)
logits = outputs['logits']
predicted_class_idx = logits.argmax(-1).item()
print("Predicted class:", model.config.id2label[predicted_class_idx])
The predicted label is brown bear, bruin, Ursus arctos.
You can also use Hugging Face MambaVision models for feature extraction. The model provides the outputs of each stage of model (hierarchical multi-scale features in 4 stages) as well as the final averaged-pool features that are flattened. The former is used for downstream tasks such as classification and detection.
The following snippet can be used for feature extraction:
from transformers import AutoModel
from PIL import Image
from timm.data.transforms_factory import create_transform
import requests
model = AutoModel.from_pretrained("nvidia/MambaVision-T-1K", trust_remote_code=True)
# eval mode for inference
model.cuda().eval()
# prepare image for the model
url = 'http://images.cocodataset.org/val2017/000000020247.jpg'
image = Image.open(requests.get(url, stream=True).raw)
input_resolution = (3, 224, 224) # MambaVision supports any input resolutions
transform = create_transform(input_size=input_resolution,
is_training=False,
mean=model.config.mean,
std=model.config.std,
crop_mode=model.config.crop_mode,
crop_pct=model.config.crop_pct)
inputs = transform(image).unsqueeze(0).cuda()
# model inference
out_avg_pool, features = model(inputs)
print("Size of the averaged pool features:", out_avg_pool.size()) # torch.Size([1, 640])
print("Number of stages in extracted features:", len(features)) # 4 stages
print("Size of extracted features in stage 1:", features[0].size()) # torch.Size([1, 80, 56, 56])
print("Size of extracted features in stage 4:", features[3].size()) # torch.Size([1, 640, 7, 7])
Currently, we offer MambaVision-T-1K, MambaVision-T2-1K, MambaVision-S-1K, MambaVision-B-1K, MambaVision-L-1K and MambaVision-L2-1K on Hugging Face. All models can also be viewed here.
Classification (pip package)
We can also import pre-trained MambaVision models from the pip package with a few lines of code:
pip install mambavision
A pretrained MambaVision model with default hyper-parameters can be created as in:
>>> from mambavision import create_model
# Define mamba_vision_T model
>>> model = create_model('mamba_vision_T', pretrained=True, model_path="/tmp/mambavision_tiny_1k.pth.tar")
Available list of pretrained models include mamba_vision_T, mamba_vision_T2, mamba_vision_S, mamba_vision_B, mamba_vision_L and mamba_vision_L2.
We can also simply test the model by passing a dummy image with any resolution. The output is the logits:
>>> import torch
>>> image = torch.rand(1, 3, 512, 224).cuda() # place image on cuda
>>> model = model.cuda() # place model on cuda
>>> output = model(image) # output logit size is [1, 1000]
Using the pretrained models from our pip package, you can simply run validation:
python validate_pip_model.py --model mamba_vision_T --data_dir=$DATA_PATH --batch-size $BS
Results + Pretrained Models
ImageNet-21K
| Name |
Acc@1(%) |
Acc@5(%) |
#Params(M) |
FLOPs(G) |
Resolution |
HF |
Download |
| MambaVision-B-21K |
84.9 |
97.5 |
97.7 |
15.0 |
224x224 |
link |
model |
| MambaVision-L-21K |
86.1 |
97.9 |
227.9 |
34.9 |
224x224 |
link |
model |
| MambaVision-L2-512-21K |
87.3 |
98.4 |
241.5 |
196.3 |
512x512 |
link |
model |
| MambaVision-L3-256-21K |
87.3 |
98.3 |
739.6 |
122.3 |
256x256 |
link |
model |
| MambaVision-L3-512-21K |
88.1 |
98.6 |
739.6 |
489.1 |
512x512 |
link |
model |
ImageNet-1K
| Name |
Acc@1(%) |
Acc@5(%) |
Throughput(Img/Sec) |
Resolution |
#Params(M) |
FLOPs(G) |
HF |
Download |
| MambaVision-T |
82.3 |
96.2 |
6298 |
224x224 |
31.8 |
4.4 |
link |
model |
| MambaVision-T2 |
82.7 |
96.3 |
5990 |
224x224 |
35.1 |
5.1 |
link |
model |
| MambaVision-S |
83.3 |
96.5 |
4700 |
224x224 |
50.1 |
7.5 |
link |
model |
| MambaVision-B |
84.2 |
96.9 |
3670 |
224x224 |
97.7 |
15.0 |
link |
model |
| MambaVision-L |
85.0 |
97.1 |
2190 |
224x224 |
227.9 |
34.9 |
link |
model |
| MambaVision-L2 |
85.3 |
97.2 |
1021 |
224x224 |
241.5 |
37.5 |
link |
model |
Detection Results + Models
Segmentation Results + Models
Installation
We provide a docker file. In addition, assuming that a recent PyTorch package is installed, the dependencies can be installed by running:
pip install -r requirements.txt
Evaluation
The MambaVision models can be evaluated on ImageNet-1K validation set using the following:
python validate.py \
--model <model-name>
--checkpoint <checkpoint-path>
--data_dir <imagenet-path>
--batch-size</imagenet-path></checkpoint-path></model-name>
评论