Metadata-Version: 2.1
Name: mambavision
Version: 1.0.7
Summary: MambaVision: A Hybrid Mamba-Transformer Vision Backbone
Home-page: https://github.com/NVlabs/MambaVision
Author: Ali Hatamizadeh
Author-email: ahatamiz123@gmail.com
License: NVIDIA Source Code License-NC
Keywords: pytorch pretrained models mamba vision transformer vit
Classifier: Programming Language :: Python :: 3.7
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Software Development
Classifier: Topic :: Software Development :: Libraries
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Requires-Python: >=3.7
Description-Content-Type: text/markdown
License-File: LICENSE

# MambaVision: A Hybrid Mamba-Transformer Vision Backbone

Official PyTorch implementation of [**MambaVision: A Hybrid Mamba-Transformer Vision Backbone**](https://arxiv.org/abs/2407.08083).


[![Star on GitHub](https://img.shields.io/github/stars/NVlabs/MambaVision.svg?style=social)](https://github.com/NVlabs/MambaVision/stargazers)

[Ali Hatamizadeh](https://research.nvidia.com/person/ali-hatamizadeh) and
[Jan Kautz](https://jankautz.com/). 

For business inquiries, please visit our website and submit the form: [NVIDIA Research Licensing](https://www.nvidia.com/en-us/research/inquiries/)

--- 

MambaVision demonstrates a strong performance by achieving a new SOTA Pareto-front in
terms of Top-1 accuracy and throughput. 

<p align="center">
<img src="https://github.com/NVlabs/MambaVision/assets/26806394/79dcf841-3966-4b77-883d-76cd5e1d4320" width=62% height=62% 
class="center">
</p>

We introduce a novel mixer block by creating a symmetric path without SSM to enhance the modeling of global context: 


<p align="center">
<img src="https://github.com/NVlabs/MambaVision/assets/26806394/295c0984-071e-4c84-b2c8-9059e2794182" width=32% height=32% 
class="center">
</p>



MambaVision has a hierarchical architecture that employs both self-attention and mixer blocks:

![teaser](./mambavision/assets/arch.png)


## 💥 News 💥

- **[07.14.2024]** We added support for images with any resolution for MambaVision (no change in code needed) ! 

- **[07.12.2024]** [Paper](https://arxiv.org/abs/2407.08083) is now available on arXiv !

- **[07.11.2024]** [Mambavision pip package](https://pypi.org/project/mambavision/) is released !

- **[07.10.2024]** We have released the code and model checkpoints for Mambavision !

## Quick Start

### Classification

We can import pre-trained MambaVision models with **1 line of code**:

```bash
pip install mambavision
```

A pretrained MambaVision model with default hyper-parameters can be created as in:

```python
>>> 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:

```python
>>> 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 
``` 
## FAQ

1. Does MambaVision support processing images with any input resolutions ? 

Yes ! you can pass images with any arbitrary resolutions without the need to change the model.


2. Can I apply MambaVision for downstream tasks like detection, segmentation ? 

Yes ! we are working to have it released very soon. But employing MambaVision backbones for these tasks is very similar to other models in `mmseg` or `mmdet` packages.

3. I am interested in re-implementing MambaVision in my own repository. Can we use the pretrained weights ? 

Yes ! the pretrained weights are released under [CC-BY-NC-SA-4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/). Please submit an issue in this repo and we will add your repository to the README of our codebase and properly acknowledge your efforts. 

## Results + Pretrained Models

### ImageNet-1K
**MambaVision ImageNet-1K Pretrained Models**

<table>
  <tr>
    <th>Name</th>
    <th>Acc@1(%)</th>
    <th>Acc@5(%)</th>
    <th>Throughput(Img/Sec)</th>
    <th>Resolution</th>
    <th>#Params(M)</th>
    <th>FLOPs(G)</th>
    <th>Download</th>
  </tr>

<tr>
    <td>MambaVision-T</td>
    <td>82.3</td>
    <td>96.2</td>
    <td>6298</td>
    <td>224x224</td>
    <td>31.8</td>
    <td>4.4</td>
    <td><a href="https://drive.google.com/file/d/1zE8czwSTG5ogcsb93A95o_F3rlYf8R1G/view?usp=sharing">model</a></td>
</tr>

<tr>
    <td>MambaVision-T2</td>
    <td>82.7</td>
    <td>96.3</td>
    <td>5990</td>
    <td>224x224</td>
    <td>35.1</td>
    <td>5.1</td>
    <td><a href="https://drive.google.com/file/d/1KNJVRRBUSqOq7ZxqH1mDth4wWL5f1SFq/view?usp=sharing">model</a></td>
</tr>

<tr>
    <td>MambaVision-S</td>
    <td>83.3</td>
    <td>96.5</td>
    <td>4700</td>
    <td>224x224</td>
    <td>50.1</td>
    <td>7.5</td>
    <td><a href="https://drive.google.com/file/d/1XoSctKJgRI6OMmYmdKOoTzvnoOtfqI64/view?usp=sharing">model</a></td>
</tr>

<tr>
    <td>MambaVision-B</td>
    <td>84.2</td>
    <td>96.9</td>
    <td>3670</td>
    <td>224x224</td>
    <td>97.7</td>
    <td>15.0</td>
    <td><a href="https://drive.google.com/file/d/1wR2UeFzSmNjrC3jqJgp4IOGvYhlO9QYw/view?usp=sharing">model</a></td>
</tr>

<tr>
    <td>MambaVision-L</td>
    <td>85.0</td>
    <td>97.1</td>
    <td>2190</td>
    <td>224x224</td>
    <td>227.9</td>
    <td>34.9</td>
    <td><a href="https://drive.google.com/file/d/1YfA9K_ZbZcoLCif-ltLWCvj2pQCvb4bJ/view?usp=sharing">model</a></td>
</tr>

<tr>
    <td>MambaVision-L2</td>
    <td>85.3</td>
    <td>97.2</td>
    <td>1021</td>
    <td>224x224</td>
    <td>241.5</td>
    <td>37.5</td>
    <td><a href="https://drive.google.com/file/d/1fw1bo_oNtIImScW38DFJIPKnRt6GrlS9/view?usp=sharing">model</a></td>
</tr>

</table>

## Installation

We provide a [docker file](./Dockerfile). In addition, assuming that a recent [PyTorch](https://pytorch.org/get-started/locally/) package is installed, the dependencies can be installed by running:

```bash
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 <batch-size-per-gpu
``` 

Here `--model` is the MambaVision variant (e.g. `mambavision_tiny_1k`), `--checkpoint` is the path to pretrained model weights, `--data_dir` is the path to ImageNet-1K validation set and `--batch-size` is the number of batch size. We also provide a sample script [here](./mambavision/validate.sh). 


## Star History

[![Star History Chart](https://api.star-history.com/svg?repos=NVlabs/MambaVision&type=Date)](https://star-history.com/#NVlabs/MambaVision&Date)


## Licenses

Copyright © 2024, NVIDIA Corporation. All rights reserved.

This work is made available under the NVIDIA Source Code License-NC. Click [here](LICENSE) to view a copy of this license.

The pre-trained models are shared under [CC-BY-NC-SA-4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/). If you remix, transform, or build upon the material, you must distribute your contributions under the same license as the original.

For license information regarding the timm repository, please refer to its [repository](https://github.com/rwightman/pytorch-image-models).

For license information regarding the ImageNet dataset, please see the [ImageNet official website](https://www.image-net.org/). 

## Acknowledgement
This repository is built on top of the [timm](https://github.com/huggingface/pytorch-image-models) repository. We thank [Ross Wrightman](https://rwightman.com/) for creating and maintaining this high-quality library.  
