---
license: mit
library_name: transformers
pipeline_tag: image-text-to-text
---

# mmMamba-linear Model Card

## Introduction
We propose mmMamba, the first decoder-only multimodal state space model achieved through quadratic to linear distillation using moderate academic computing resources. Unlike existing linear-complexity encoder-based multimodal large language models (MLLMs), mmMamba eliminates the need for separate vision encoders and underperforming pre-trained RNN-based LLMs. Through our seeding strategy and three-stage progressive distillation recipe, mmMamba effectively transfers knowledge from quadratic-complexity decoder-only pre-trained MLLMs while preserving multimodal capabilities. Additionally, mmMamba introduces flexible hybrid architectures that strategically combine Transformer and Mamba layers, enabling customizable trade-offs between computational efficiency and model performance.

Distilled from the decoder-only HoVLE-2.6B, our pure Mamba-2-based mmMamba-linear achieves performance competitive with existing linear and quadratic-complexity VLMs, including those with 2x larger parameter size like EVE-7B. The hybrid variant, mmMamba-hybrid, further enhances performance across all benchmarks, approaching the capabilities of the teacher model HoVLE. In long-context scenarios with 103K tokens, mmMamba-linear demonstrates remarkable efficiency gains with a 20.6× speedup and 75.8% GPU memory reduction compared to HoVLE, while mmMamba-hybrid achieves a 13.5× speedup and 60.2% memory savings.

<div align="center">
<img src="teaser.png" />

<b>Seeding strategy and three-stage distillation pipeline of mmMamba.</b>
<img src="pipeline.png" />
</div>

Paper: [https://hf.co/papers/2502.13145](https://hf.co/papers/2502.13145)

Code: [https://github.com/hustvl/mmMamba](https://github.com/hustvl/mmMamba)

## Quick Start Guide for mmMamba Inference

We provide example code to run mmMamba inference using the Transformers library.

### Main Dependencies for Model Inference

Below are the primary dependencies required for model inference:
- torch==2.1.0
- torchvision==0.16.0
- torchaudio==2.1.0
- transformers==4.37.2
- peft==0.10.0
- triton==3.2.0
- [mamba_ssm](https://github.com/state-spaces/mamba/releases/download/v2.2.4/mamba_ssm-2.2.4%2Bcu12torch2.1cxx11abiFALSE-cp310-cp310-linux_x86_64.whl)
- [causal_conv1d](https://github.com/Dao-AILab/causal-conv1d/releases/download/v1.5.0.post8/causal_conv1d-1.5.0.post8%2Bcu12torch2.1cxx11abiFALSE-cp310-cp310-linux_x86_64.whl)
- [flash_attn](https://github.com/Dao-AILab/flash-attention/releases/download/v2.6.0/flash_attn-2.6.0%2Bcu122torch2.1cxx11abiFALSE-cp310-cp310-linux_x86_64.whl)
(Please note that you need to select and download the corresponding .whl file based on your environment.)
- peft
- omegaconf
- rich
- accelerate
- sentencepiece
- decord
- seaborn

### Inference with Transformers

```python
import numpy as np
import torch
import torchvision.transforms as T
from decord import VideoReader, cpu
from PIL import Image
from torchvision.transforms.functional import InterpolationMode
from transformers import AutoModel, AutoTokenizer

IMAGENET_MEAN = (0.485, 0.456, 0.406)
IMAGENET_STD = (0.229, 0.224, 0.225)

def build_transform(input_size):
    MEAN, STD = IMAGENET_MEAN, IMAGENET_STD
    transform = T.Compose([
        T.Lambda(lambda img: img.convert('RGB') if img.mode != 'RGB' else img),
        T.Resize((input_size, input_size), interpolation=InterpolationMode.BICUBIC),
        T.ToTensor(),
        T.Normalize(mean=MEAN, std=STD)
    ])
    return transform

def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size):
    best_ratio_diff = float('inf')
    best_ratio = (1, 1)
    area = width * height
    for ratio in target_ratios:
        target_aspect_ratio = ratio[0] / ratio[1]
        ratio_diff = abs(aspect_ratio - target_aspect_ratio)
        if ratio_diff < best_ratio_diff:
            best_ratio_diff = ratio_diff
            best_ratio = ratio
        elif ratio_diff == best_ratio_diff:
            if area > 0.5 * image_size * image_size * ratio[0] * ratio[1]:
                best_ratio = ratio
    return best_ratio

def dynamic_preprocess(image, min_num=1, max_num=12, image_size=448, use_thumbnail=False):
    orig_width, orig_height = image.size
    aspect_ratio = orig_width / orig_height

    # calculate the existing image aspect ratio
    target_ratios = set(
        (i, j) for n in range(min_num, max_num + 1) for i in range(1, n + 1) for j in range(1, n + 1) if
        i * j <= max_num and i * j >= min_num)
    target_ratios = sorted(target_ratios, key=lambda x: x[0] * x[1])

    # find the closest aspect ratio to the target
    target_aspect_ratio = find_closest_aspect_ratio(
        aspect_ratio, target_ratios, orig_width, orig_height, image_size)

    # calculate the target width and height
    target_width = image_size * target_aspect_ratio[0]
    target_height = image_size * target_aspect_ratio[1]
    blocks = target_aspect_ratio[0] * target_aspect_ratio[1]

    # resize the image
    resized_img = image.resize((target_width, target_height))
    processed_images = []
    for i in range(blocks):
        box = (
            (i % (target_width // image_size)) * image_size,
            (i // (target_width // image_size)) * image_size,
            ((i % (target_width // image_size)) + 1) * image_size,
            ((i // (target_width // image_size)) + 1) * image_size
        )
        # split the image
        split_img = resized_img.crop(box)
        processed_images.append(split_img)
    assert len(processed_images) == blocks
    if use_thumbnail and len(processed_images) != 1:
        thumbnail_img = image.resize((image_size, image_size))
        processed_images.append(thumbnail_img)
    return processed_images

def load_image(image_file, input_size=448, max_num=12):
    image = Image.open(image_file).convert('RGB')
    transform = build_transform(input_size=input_size)
    images = dynamic_preprocess(image, image_size=input_size, use_thumbnail=True, max_num=max_num)
    pixel_values = [transform(image) for image in images]
    pixel_values = torch.stack(pixel_values)
    return pixel_values


path = 'hustvl/mmMamba-linear'
model = AutoModel.from_pretrained(
    path,
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=False,
    trust_remote_code=True).eval().cuda()
tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)

# set the max number of tiles in `max_num`
pixel_values = load_image('/path/to/image', max_num=12).to(torch.bfloat16).cuda()
generation_config = dict(max_new_tokens=1024, do_sample=True)

# pure-text conversation (纯文本对话)
question = 'Hello, who are you?'
response, history = model.chat(tokenizer, None, question, generation_config, history=None, return_history=True)
print(f'User: {question}\nAssistant: {response}')


# single-image single-round conversation (图文对话)
question = '<image>\nPlease describe the image shortly.'
response = model.chat(tokenizer, pixel_values, question, generation_config)
print(f'User: {question}\nAssistant: {response}')
```