README.md

---
license: llama3
tags:
- vision
- image-text-to-text
language:
- en
pipeline_tag: image-text-to-text
---

# LLaVa-Next Model Card

The LLaVA-NeXT model was proposed in [LLaVA-NeXT: Stronger LLMs Supercharge Multimodal Capabilities in the Wild
](https://llava-vl.github.io/blog/2024-05-10-llava-next-stronger-llms/) by Bo Li, Kaichen Zhang, Hao Zhang, Dong Guo, Renrui Zhang, Feng Li, Yuanhan Zhang, Ziwei Liu, Chunyuan Li.
These LLaVa-NeXT series improves upon [LLaVa-1.6](https://llava-vl.github.io/blog/2024-01-30-llava-next/) by training with stringer language backbones, improving the
performance.

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

## Model description

LLaVa combines a pre-trained large language model with a pre-trained vision encoder for multimodal chatbot use cases. LLaVA NeXT Llama3 improves on LLaVA 1.6 BY:
- More diverse and high quality data mixture
- Better and bigger language backbone

Base LLM: [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct)
  
![image/png](https://cdn-uploads.huggingface.co/production/uploads/62441d1d9fdefb55a0b7d12c/FPshq08TKYD0e-qwPLDVO.png)

## Intended uses & limitations

You can use the raw model for tasks like image captioning, visual question answering, multimodal chatbot use cases. See the [model hub](https://huggingface.co/models?search=llava-hf) to look for
other versions on a task that interests you.

### How to use

To run the model with the `pipeline`, see the below example:

```python
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="llava-hf/llama3-llava-next-8b-hf")
messages = [
    {
      "role": "user",
      "content": [
          {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/ai2d-demo.jpg"},
          {"type": "text", "text": "What does the label 15 represent? (1) lava (2) core (3) tunnel (4) ash cloud"},
        ],
    },
]

out = pipe(text=messages, max_new_tokens=20)
print(out)
>>> [{'input_text': [{'role': 'user', 'content': [{'type': 'image', 'url': 'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/ai2d-demo.jpg'}, {'type': 'text', 'text': 'What does the label 15 represent? (1) lava (2) core (3) tunnel (4) ash cloud'}]}], 'generated_text': 'Lava'}]
```


You can also load and use the model like following:

```python
from transformers import LlavaNextProcessor, LlavaNextForConditionalGeneration
import torch
from PIL import Image
import requests

processor = LlavaNextProcessor.from_pretrained("llava-hf/llama3-llava-next-8b-hf")
model = LlavaNextForConditionalGeneration.from_pretrained("llava-hf/llama3-llava-next-8b-hf", torch_dtype=torch.float16, device_map="auto") 

# prepare image and text prompt, using the appropriate prompt template
url = "https://github.com/haotian-liu/LLaVA/blob/1a91fc274d7c35a9b50b3cb29c4247ae5837ce39/images/llava_v1_5_radar.jpg?raw=true"
image = Image.open(requests.get(url, stream=True).raw)

# Define a chat histiry and use `apply_chat_template` to get correctly formatted prompt
# Each value in "content" has to be a list of dicts with types ("text", "image") 
conversation = [
    {

      "role": "user",
      "content": [
          {"type": "text", "text": "What is shown in this image?"},
          {"type": "image"},
        ],
    },
]
prompt = processor.apply_chat_template(conversation, add_generation_prompt=True)

inputs = processor(images=image, text=prompt, return_tensors="pt").to(model.device)

# autoregressively complete prompt
output = model.generate(**inputs, max_new_tokens=100)

print(processor.decode(output[0], skip_special_tokens=True))
```

-----------
From transformers>=v4.48, you can also pass image url or local path to the conversation history, and let the chat template handle the rest.
Chat template will load the image for you and return inputs in `torch.Tensor` which you can pass directly to `model.generate()` 

```python
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://www.ilankelman.org/stopsigns/australia.jpg"}
            {"type": "text", "text": "What is shown in this image?"},
        ],
    },
]

inputs = processor.apply_chat_template(messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors"pt")
output = model.generate(**inputs, max_new_tokens=50)
```

### Model optimization

#### 4-bit quantization through `bitsandbytes` library

First make sure to install `bitsandbytes`, `pip install bitsandbytes` and make sure to have access to a CUDA compatible GPU device. Simply change the snippet above with: 

```diff
model = LlavaNextForConditionalGeneration.from_pretrained(
    model_id, 
    torch_dtype=torch.float16, 
    low_cpu_mem_usage=True,
+   load_in_4bit=True
)
```

#### Use Flash-Attention 2 to further speed-up generation

First make sure to install `flash-attn`. Refer to the [original repository of Flash Attention](https://github.com/Dao-AILab/flash-attention) regarding that package installation. Simply change the snippet above with: 

```diff
model = LlavaNextForConditionalGeneration.from_pretrained(
    model_id, 
    torch_dtype=torch.float16, 
    low_cpu_mem_usage=True,
+   use_flash_attention_2=True
).to(0)
```
### Training Data
- 558K filtered image-text pairs from LAION/CC/SBU, captioned by BLIP.
- 158K GPT-generated multimodal instruction-following data.
- 500K academic-task-oriented VQA data mixture.
- 50K GPT-4V data mixture.
- 40K ShareGPT data.


### BibTeX entry and citation info

```bibtex
@misc{li2024llavanext-strong,
    title={LLaVA-NeXT: Stronger LLMs Supercharge Multimodal Capabilities in the Wild},
    url={https://llava-vl.github.io/blog/2024-05-10-llava-next-stronger-llms/},
    author={Li, Bo and Zhang, Kaichen and Zhang, Hao and Guo, Dong and Zhang, Renrui and Li, Feng and Zhang, Yuanhan and Liu, Ziwei and Li, Chunyuan},
    month={May},
    year={2024}
}
```